Huawei–Qualcomm Patent Deal: a reset for 5G, AI, and Networked Computing IP
Executive Summary:
On October 5th, Huawei and Qualcomm announced a multi-year, broad patent-license agreement spanning 5G, computing, artificial intelligence, and networking. The agreement provides cross-licenses to the two companies’ patent portfolios and includes Qualcomm’s purchase of certain Huawei U.S. patents in computing, AI, networking, and unspecified additional technical fields. Closing remains subject to required regulatory approvals, and neither company disclosed the financial value, patents transferred, geographic scope, royalty structure, term, or product-specific rights.
The companies framed the arrangement as consistent with fair, reasonable and non-discriminatory licensing principles. That language matters. FRAND commitments normally attach to standard-essential patents, or SEPs, whose use is unavoidable when implementing a technical standard. Yet the stated scope reaches beyond conventional cellular SEP licensing into computing, AI, and networking—areas likely to include a mixture of SEP and non-SEP implementation, architecture, semiconductor, and systems patents.
For the telecom sector, the central significance is not that Huawei and Qualcomm have suddenly become commercial partners. The companies did not announce a chip supply agreement, joint product development effort, network-equipment integration program, or relaxation of U.S. technology controls. Rather, they have reduced one category of strategic uncertainty: reciprocal exposure to patent claims across a widening set of overlapping technology markets.
At a glance:
-
The agreement includes Qualcomm’s purchase of certain Huawei US patents related to computing, AI and networking, Huawei said in a statement.
-
It is Huawei’s first patent licensing deal with Qualcomm that covers 5G technologies.
-
Huawei said the deal is expected to push the total value of its patent licensing agreements to more than $6.9 billion once completed.
-
The company said its intellectual property licensing business has generated positive revenue since 2021, reflecting years of heavy investment in research and development.
-
Huawei ramped up R&D spending in recent years, as US trade restrictions since 2019 have limited its ability to buy advanced chips and crucial software.
-
Huawei made its first licensing payment to Qualcomm in 2001, while it received its first licensing income in 2011 from Motorola.
-
The Qualcomm agreement follows a licensing deal with HP Inc announced by Huawei in August 2026.
Why the patent licensing scope matters:
The announced scope—5G, compute, AI, and networking—tracks the technological convergence reshaping telecom infrastructure.
This is the practical backdrop to the agreement. The industry is moving from a model in which cellular patent licensing could be discussed principally in terms of handsets and modem technology to one in which radios, accelerators, devices, private networks, edge systems, AI infrastructure, and enterprise networking increasingly interact.
A future 5G-Advanced or 6G deployment may incorporate AI-based radio-resource management, cloud-native network functions, edge inference, high-speed fronthaul and backhaul, specialized processors, and new interconnect architectures. Patent boundaries do not necessarily follow clean organizational boundaries between a handset modem supplier, a network-equipment vendor, a data-center platform provider, and an AI-compute company.
The Huawei–Qualcomm accord reflects that reality. It addresses IP access across the technical layers rather than treating 5G as a self-contained radio-standard licensing category.
A significant shift in Huawei’s licensing posture:
Huawei’s intellectual-property licensing business has become strategically more visible as U.S. export restrictions have constrained the company’s access to advanced chips, design tools, and other technology inputs. Reuters reported that Huawei said the Qualcomm agreement, once complete, is expected to bring the total value of its patent-licensing agreements above $6.9 billion. Huawei also said that its IP licensing business has generated positive revenue since 2021.
That does not disclose the value of the Qualcomm transaction itself. It does, however, demonstrate that Huawei views patent licensing as more than a defensive function. The company is using its accumulated R&D output as a commercial asset, especially in markets and technology segments where product sales are shaped by geopolitical restrictions or procurement barriers.
Huawei has long argued that it has made substantial standards contributions in cellular technology. In the joint announcement, Huawei highlighted its work in 4G and 5G, including polar codes, while Qualcomm emphasized its own foundational wireless technologies and the global acceptance of its 5G SEP licensing program.qualcomm+1
The mutual recognition is revealing. Qualcomm’s licensing organization, Qualcomm Technology Licensing, has historically been one of the industry’s most important SEP licensors. Huawei, meanwhile, has worked to establish itself as a major technology contributor and licensor rather than solely as a licensee of Western-originated communications IP. The new agreement positions both companies as holders and users of strategically valuable portfolios across the broader compute-network-AI stack.
The 5G dimension:
Reuters described the transaction as Huawei’s first patent licensing deal with Qualcomm that covers 5G technologies. That point is more material than it may initially appear. 5G licensing is not simply a continuation of earlier 3G and 4G licensing arrangements. It encompasses a more diverse implementation landscape:
-
Massive-MIMO and advanced antenna systems.
-
Flexible numerology and broader spectrum support.
-
Ultra-reliable and low-latency communications features.
-
Network slicing and service-based core architectures.
-
Virtualized and cloud-native RAN and core implementations.
-
Private-network and industrial deployments.
-
Fixed wireless access, integrated access and backhaul, and increasingly AI-assisted optimization.
The licensing challenge is therefore expanding in parallel with the market. In earlier cellular generations, the primary commercial unit was often a phone or modem-equipped device. In 5G, relevant implementation questions can also arise in enterprise gateways, fixed-wireless equipment, industrial modules, routers, private-network infrastructure, distributed RAN systems, and cloud-hosted network functions.
The agreement could reduce licensing friction in technology areas where both companies have extensive engineering programs and potential freedom-to-operate concerns. It does not establish a universal industry royalty benchmark, nor does it reveal either company’s licensing terms. But it removes the prospect of bilateral patent disputes becoming another source of uncertainty in 5G infrastructure and connected-computing markets.
AI and wireless networking are strategic markets:
The inclusion of AI and wireless networking should not be treated as decorative language appended to a cellular patent deal. It signals that both companies see IP value moving into the architecture around wireless connectivity.
For Qualcomm, the expansion aligns with its effort to participate more broadly in AI-enabled computing, industrial systems, edge platforms, and infrastructure—not only smartphone application processors and cellular modems. Qualcomm characterized itself in the announcement as a computing company operating across personal devices and large-scale infrastructure.
For Huawei, AI and networking patent rights are strategically important because the company remains active across telecom equipment, enterprise networking, cloud, data centers, AI hardware and software, and edge infrastructure. Its IP portfolio is therefore likely to span elements relevant to AI-enabled networks, high-performance networking, acceleration, orchestration, and systems integration.
The agreement does not specify the assets involved, making it impossible to assess whether the transferred U.S. patents concern AI accelerators, interconnect technologies, network control, edge architectures, processor design, or software implementation. It would be premature to infer a specific product strategy from the deal.
Still, Qualcomm’s decision to purchase Huawei U.S. patents indicates that it identifies value in owning—not merely licensing—selected IP assets. In practical terms, acquired patents can strengthen a company’s defensive position, broaden its licensing portfolio, support cross-licensing negotiations, or provide potential enforcement rights in relevant markets. The transaction is subject to regulatory approvals, so even this aspect remains incomplete.
Huawei R&D and Patent Policy:
According to its annual report, Huawei invested 192.3 billion yuan (US$28.7 billion) into research and development in 2025, equal to nearly 22 per cent of its total revenue. One of the world’s largest patent holders, Huawei held 165,000 active granted patents worldwide as of 2025.
In 2022, Huawei founder Ren Zhengfei urged the company’s intellectual property team to step up efforts to turn its vast pool of patents into revenue via “reasonable pricing” and to “generate an appropriate return” on its investments.
In June 2026, Huawei announced that it would start charging royalties for its Wi‑Fi 7 [aka IEEE 802.11be Extremely High Throughput (EHT)] technologies, the newest wireless standard, with the rate set at 50 US cents per unit. A key player in Wi-fi licensing, Huawei’s patent licensing deals covered 1.6 billion devices globally by the end of 2025, the firm said.
Executive Quotes:
“Qualcomm has invested in foundational wireless technologies that have enabled successive generations of mobile innovation and earned broad recognition across the global wireless industry. This agreement reaffirms industry recognition of Qualcomm’s 5G technology leadership and the success of Qualcomm’s 5G SEP licensing program,” said John Han, Executive Vice President and General Manager of Qualcomm Technology Licensing. “This agreement likewise reflects Qualcomm’s recognition of Huawei’s continued innovation and intellectual property in 5G and other technology fields.”
“Huawei’s decades of sustained investment in fundamental R&D have driven innovation and progress in mobile communications and other technology fields. Huawei’s broad contributions to the 4G/5G standards, such as the near-physical-limit signal transmission technology using polar codes, have established Huawei’s leadership in the mobile communications industry, and continuously transformed the way people communicate and live,” said Alan Fan, Huawei’s Chief Intellectual Property Officer. “This agreement not only demonstrates the value of Huawei’s innovations, but also recognizes Qualcomm’s foundational contributions to modern communication technologies.”
Conclusions:
The Huawei–Qualcomm agreement is best viewed as a strategic IP normalization deal at the intersection of 5G, AI, compute, and networking. It recognizes that the center of gravity in communications technology is shifting from isolated wireless functions toward integrated systems that combine radios, processors, AI workloads, cloud software, and high-speed networks.
Its most important immediate effect is to reduce uncertainty between two major patent holders. Its most important broader message is that the next phase of telecom innovation—and of telecom IP licensing—will not be confined to the air interface.
The deal also underscores a practical reality for 5G-Advanced and emerging 6G ecosystems: the relevant IP stack increasingly extends from spectrum-efficient radio techniques and network protocols to AI inference, accelerator architectures, cloud-native control systems, edge platforms, and data-center interconnect. Companies that treat these as separate legal, technical, or commercial domains may be working from an outdated map.
……………………………………………………………………………………………………………………………
References:
https://www.huawei.com/en/news/2026/10/qualcomm-broad-patent-agreement
Huawei, Qualcomm, Samsung, and Ericsson Leading Patent Race in $15 Billion 5G Licensing Market
Huawei and Ericsson renew global patent cross-licensing agreement
5G Specifications (3GPP), 5G Radio Standard (IMT 2020) and Standard Essential Patents
Huawei or Samsung: Leader in 5G declared Standard Essential Patents (SEPs)?
AI Debt Wave Implications: Higher interest rates with enormous AI sales required for ROI
Executive Summary:
Artificial Intelligence (AI) debt fueled infrastructure buildouts have become a broad fixed-income-market supply shock. The Wall Street Journal reports that 49% of new investment-grade bond issuance year to date in 2026 has been AI-linked—a figure that underscores how the financing of data centers, compute infrastructure and supporting power systems is reshaping bond market fundamentals and technical factors.
As the supply of long-dated bonds rises, prices fall and yields rise until investors are compensated enough to absorb them. The AI buildout is adding enormous corporate debt from hyperscalers, data-center developers, semiconductor and networking suppliers, and utilities—while also driving public borrowing for grid, power, water and transportation infrastructure. This is putting upward pressure on yields across Treasuries, corporate bonds and municipal debt. This heavy concentration of long-dated supply creates a crowding-out effect that forces non-AI issuers to compete against elevated benchmark yields and potentially stifles broader economic growth, especially for companies not involved in circular AI funding deals (see References).
AI is capital intensive (see Table 1. below). Building data-center capacity requires cash not only for GPUs and servers, but also for land, buildings, fiber, networking, cooling, power procurement, backup generation, substations, transmission and water systems. The major hyperscalers can fund part of that investment from cash flow, but they are also issuing lots of new bonds and notes to preserve liquidity and accelerate buildouts. Their suppliers, data-center partners and power providers are doing the same. Much of this borrowing is long dated—exactly where the Treasury is issuing heavily to finance federal deficits.
The result is an expanding pool of long-duration debt competing for the same institutional buyers: insurers, pension funds, mutual funds, banks, foreign investors and asset managers. When those buyers do not increase their allocations at the same pace as supply, issuers must offer higher yields. Here’s the flow chart that loops around indefinitely until there is an AI crash!:
More AI related bond supply→lower prices→higher yields→higher term premium for U.S. Treasuries→higher budget deficits to finance the increased debt→more bond supply→etc.
……………………………………………………………………………………………………………………………………………………………………………………………………………………………………………………………………………….
- 2026 Debt Issuance: Projected between $220 billion and $250+ billion for major US hyperscalers (Alphabet / Google, Amazon, Meta, Microsoft, and Oracle), with broader AI-related market debt reaching up to $570 billion.
- 2027 Forecast: Goldman Sachs projects direct hyperscaler debt issuance to nearly double to around $400 billion (Jeff Pu estimates $419 billion) as companies finance over a third of their infrastructure needs.
- Aggregate Capex: Combined capital expenditures are expected to hit roughly $940 billion in 2026 and scale past $1.3 trillion in 2027.
- Market Share: Hyperscaler investment-grade bond sales have jumped from roughly 2% of total US supply (2022–2024) to roughly 9% in 2026.
- Credit Impact: Credit spreads on a 10-year hyperscaler credit basket have widened from historical 40–75bp ranges toward 90bp+, driven by leverage concerns.
- Cash Flow Outlook: S&P Global Ratings expects all major hyperscalers to run negative free operating cash flow through 2026 and 2027, with a cash-flow inflection point not projected until 2028–2029.
Hyperscaler Funding Boom: Debt vs. Capex Projections (2026–2027):
Source: Google Gemini
……………………………………………………………………………………………………………………………………………………………………………………………………………………………………
Corporate Borrowing Negatively Impacted by AI Debt Financing:
Corporate bonds face the most direct effect. When hyperscalers, data-center operators, chip suppliers and utilities all enter the fixed income market at once, they compete for investor capital and dealer balance sheets.
The bond market adjusts through the following mechanisms:
- Larger new-issue concessions, meaning issuers must offer higher yields than comparable bonds already trading.
- Lower prices for outstanding corporate bonds as investors sell them to make room for new issues.
- Wider credit spreads, particularly for lower-rated investment-grade borrowers.
- Higher borrowing costs for non-AI companies that must compete with AI-linked supply for investor allocations.
The pressure is not limited to companies directly building AI agents or systems. A telecom operator, industrial firm, REIT or consumer company will likely pay more to borrow because investors can buy a new, liquid, highly rated hyperscaler bond at a potentially higher and attractive yield. That discourages non-AI corporate borrowing and leads to slower economic growth.
………………………………………………………………………………………………………………………………………………………………………………………………
AI Pain Points Personified:
1. Ares reports, “We documented well more than 100 digital-infrastructure financings from just the past twelve months… Despite different issuers, different structures, different rating agencies and different credit markets, all of the risk converges on just eight (AI tech) names: Meta, Oracle, Microsoft, Amazon, Google, Nvidia, and, on a look-through basis, OpenAI and Anthropic.”
2. Analysts at Goldman Sachs Group and elsewhere have calculated that more than $1 trillion has already been spent on the data center build-out since the launch of ChatGPT in late 2022. Presumably, other large American businesses would need to pay for AI tools to justify all of this investment. That money needs to come from somewhere. but who’s going to pay for those AI tools?
3. From Greg Ip of the WSJ: “Will America Spend 9% of Its GDP on AI? The Industry Is Counting on It“:
“Is it plausible that Americans will spend as much of their income on this one technology as they do on food? Roughly twice what the nation pays for all forms of energy or all computers and software? Seven times what consumers spend on phone, streaming, and internet services combined?”
“You should be skeptical. Even the most transformative inventions eventually run into the law of diminishing returns: each additional dollar a user spends yields less additional productivity (or enjoyment) than the last. That imposes a natural ceiling. The question, of course, is where that ceiling is. Whether or not you think 9% of GDP is right, you have to care, because this figure isn’t some fever dream: it is implicit in the dollars that investors and companies are committing right now.”
Chart Credit: Greg Ip, Wall Street Journal
………………………………………………………………………………………………………………..
Dire Warnings:
Nobel laureate Daron Acemoglu writes about “Distorted Intelligence at the AI Frontier” (emphasis added):
“The real problem for the frontier AI labs is not that their models are too powerful and already “misaligned” with the goals their creators set for them. Rather, it is that the models are being trained in ways that may be leading to a type of intelligence that will become less predictable. AI frontier labs are training their models in ways that may be leading to a type of distorted intelligence.
“This is what I mean by distorted intelligence. If my suspicion is correct, what we are dealing with is not a model racing toward superintelligence, but a brittle house of cards that becomes more and more likely to malfunction and collapse as we demand more from it.”
Conclusions:
The above analysis outlines critical structural risks and concerns directly related to the AI infrastructure (capex) spending boom. It references several articles which highlight the risks of circular financing loops where hardware suppliers (like Nvidia and Broadcom) fund their buyers (like Anthropic and Open AI). Additionally, hyperscalers face structural cash deficits with negative free operating cash flows projected through 2027, alongside rising sovereign risk and potential conflicts of interest from intertwined debt and leasing arrangements. This table balances the author’s views with a counter-perspective:
| IEEE Techblog Warning | The Nuanced Counter-Perspective |
|---|---|
| Circular Funding Risk: Massive vendor lending deals (like Broadcom lending Anthropic $42 billion) are artificially inflating revenue and building a brittle house of cards. | Balance Sheet Cushion: Unlike the 2000 Dot-Com crash where telecom infrastructure was built on speculative, junk-rated leverage, 2026 hyperscalers possess massive cash-generative core businesses (search, cloud, e-commerce) to subsidize their debt service even if AI returns are delayed. |
| Diminishing Returns: Society cannot plausibly spend 9% of GDP on AI, meaning demand will hit a hard ceiling and spark a debt bust. | Productivity Deflation: AI infrastructure spending may not need to justify itself through direct “software sales.” If the technology lowers the baseline operating costs of the global services economy, the ROI manifests as structural corporate margin expansion rather than retail sales. |
………………………………………………………………………………………………………………………………………………………………………………………………
References:
Curmudgeon: Caveat Emptor: Huge Debt and Circular Financing Deals Dominate AI Build-Outs (07/23)
Broadcom lending Anthropic up to $42 billion in yet another AI circular financing deal
Bain & Co: AI Infrastructure Buildout Will Require $6 Trillion Revenue by 2031 to Support Massive CAPEX
The AI Infrastructure Build-Out: A $10 Trillion Bet on Compute, Power, and Networks
Nvidia CEO Huang: AI is the largest infrastructure buildout in human history; AI Data Center CAPEX will generate new revenue streams for operators
AI risks and backlash increase; Recap of the circular loop of fake AI profits and hyperscaler markups of private AI companies
China vs U.S.: Race to Generate Power for AI Data Centers as Electricity Demand Soars
How will fiber and equipment vendors meet the increased demand for fiber optics in 2026 due to AI data center buildouts?
Expose: AI is more than a bubble; it’s a data center debt bomb
Will billions of dollars big tech is spending on Gen AI data centers produce a decent ROI?
Huge Risks for the proposed $500B AI Investments from Giant Wall Street firms
Can the debt fueling the new wave of AI infrastructure buildouts ever be repaid?
Light Counting: optical transceiver sales surge; IEEE 802.3 and ITU-T optical transport standards explained/UEC addendum
Sales of All Types of Optical Transceivers Set New Records:
Light Counting reports that the sales of optical transceivers, shown in the first figure below, illustrates the new record set in Q2 2026 and estimates for the rest of 2026. Sales of Ethernet transceivers currently account for about 80% of the total market. Shipments of “legacy products”—800G 2xDR4 and 2xFR4, as well as 400ZR/ZR+, were up sharply in Q2, by up to 3x year-over-year in terms of units, well ahead of the market research firm’s expectations. Unit shipments of 100G and 400G Ethernet transceivers also set new records. Even 10G Ethernet was up. Volume shipments of 1.6T Ethernet transceivers and 800ZR/ZR+ are off to a great start this year, but this was less surprising.
Image Credit: Light Counting
……………………………………………………………………………………………………………………………………………………………………………………………………………………………………………
Leading suppliers of optical transceivers reported record revenues and profits, despite component shortages. Here are Q2-2026 results for the top two suppliers:
- Innolight achieved a record revenue of $3.3 billion in Q2, up 192% year-over-year and up 16% sequentially. The company’s net income was $1.16 billion, up 249% year-over-year and up 40% sequentially. Gross margin increased this quarter, driven by sequential volume growth of 1.6T products.
- Eoptolink reported a record revenue of $1.8 billion in Q2, up 109% year-over-year and up 53% sequentially. The company’s net income was $698 million, up 113% year-over-year and up 74% quarter-over-quarter.
| Supplier | Estimated 2025 Sales / Segment Revenue | Market Context & Performance |
|---|---|---|
| InnoLight Technology | $5.3 Billion | Market Leader (#1); revenue surged 61% year-over-year driven by mass delivery of 800G and 1.6T modules. |
| Eoptolink | $3.5 Billion | Moved to #2 globally; registered an astonishing ~188% segment surge supplying 400G/800G optics to Amazon and Nvidia. |
| Coherent Corp. | ~$4.0 – $4.5 Billion (Data Center & Comm. Segment) | Coherent was overtaken by Eoptolink in pure transceiver rankings, but its broader communications segment generated massive revenues (including a record $1.2 Billion in Q4 2025 alone). |
| Cisco Systems | ~$1.5 – $2.0 Billion (Optical/Acacia) | Cisco captures a massive chunk of the telecom and coherent pluggable market through its internal networking sales and Acacia subsidiary. |
| Lumentum Holdings | ~$1.0 – $1.3 Billion | Reported record revenues and high double-digit growth by the end of 2025, heavily buoyed by their NeoPhotonics integration. |
| Broadcom | ~$800 Million – $1.1 Billion (Modules/Engines) | While Broadcom generates over $18B+ quarterly in overall networking/AI silicon, its specific optical transceiver and co-packaged optical engine sales represent a smaller, premium segment. |
| Accelink Technologies | ~$600 – $800 Million | Maintained a steady footprint setting new shipping records for Chinese domestic cloud providers and 5G operators. |
| Sumitomo Electric | ~$400 – $500 Million | Holds a reliable niche focused primarily on specialized telecom optical components and compact Japanese network infrastructure deployments. |
This demand is being driven primarily by AI data-center build-outs, where the scaling of accelerator clusters requires concurrent investment in compute, switching, optical interconnect, power, and cooling. Ethernet is increasingly central to scale-out fabrics and, in some architectures, to emerging scale-up designs.
The reported growth at leading suppliers illustrates how rapidly demand is propagating through the component supply chain. As note above, Innolight reported nearly 200% year-over-year growth in optical-transceiver sales during the first half of 2026. Such results are consistent with an industry that is simultaneously absorbing high volumes of 800G modules and preparing the supply chain for 1.6T transitions.
Optical Transceiver Standards:
The standards baseline spans both Ethernet and optical transport.
- For intra-data-center connectivity and some point-to-point data-center-interconnect applications, IEEE 802.3df-2024 standardized Ethernet MAC parameters for 800GbE, together with physical-layer and management specifications for 400GbE and 800GbE operation.
- IEEE P802.3dj is developing 200 Gb/s-per-lane electrical and optical interfaces for 200GbE, 400GbE, 800GbE, and 1.6TbE Ethernet operation.
- For carrier optical-transport and WDM applications, ITU-T G.709 defines the OTN digital transport hierarchy and interfaces; ITU-T G.959.1 specifies OTN physical-layer interface application codes; and ITU-T G.698.2 defines single-channel optical-interface parameters for amplified DWDM “black-link” applications.
- A carrier can transport an Ethernet PHY signal directly over a coherent wavelength. The clearest standardized example is OIF 400ZR, which carries a 400GBASE-R host signal transparently over a single coherent DWDM carrier. It is explicitly specified as a 400G BASE-R PHY interface, rather than as a full ITU-T G.709 OTUk line signal.
Optical Ethernet Standards and Product Transition:
The current optical upgrade cycle spans multiple Ethernet (IEEE 802.3) generations:
It is important to distinguish standardized Ethernet interface rates from the broader product roadmap of optical suppliers. IEEE 802.3df standardized 800 GbE, while IEEE P802.3dj addresses 1.6 TbE and the associated 200 Gb/s-per-lane physical-layer ecosystem. By contrast, 2.4T and 3.2T transceiver products may be relevant to future system architectures, but they should not be presented as ratified IEEE Ethernet rates at this stage. Please see the Addendum below for an explanation of he Ultra Ethernet Consortium.
ITU-T OTN and WDM Recommendations:
While IEEE 802.3 specifications define Ethernet MAC and PHY operation, they are not the only standards framework relevant to high-speed optical modules. In optical transport networks, ITU-T Recommendations define the digital transport structure, optical-channel parameters, WDM frequency plans, and interoperable line-side interface requirements. Optical transceivers must meet the requirements of the applicable interface standard (IEEE 802.3 or ITU-T) or implementation agreement for their intended deployment. For intra-data-center Ethernet links, this normally means IEEE 802.3 PHY specifications and applicable MSAs. For OTN, WDM, or carrier interdomain applications, the relevant requirements may additionally include ITU-T G.709, G.959.1, G.698.2, and G.694.1.
- ITU-T G.709 specifies the interfaces and digital hierarchy of the Optical Transport Network (OTN), including optical transport units and associated overhead, forward-error-correction, and multiplexing structures. An Ethernet client signal may therefore be mapped into an OTN container for transport across a metro, regional, long-haul, or subsea optical network.
- ITU-T G.959.1 specifies Optical Transport Network physical-layer interfaces, including interdomain interface requirements for single-channel and multichannel optical systems. Its application codes define parameters such as nominal bit rate, reach class, wavelength range, transmitter characteristics, receiver tolerance, and optical power budgets. The recommendation is intended to enable transverse, or multivendor, compatibility at optical interfaces crossing administrative-domain boundaries; it can also be used for intra-domain interfaces where appropriate. The current G.959.1 edition, issued in 2024 and subsequently amended, includes interfaces relevant to high-speed PAM4-based optical transport applications.
- ITU-T G.698.2 is also relevant where pluggable coherent or direct-detect modules connect into amplified DWDM systems. It specifies single-channel optical-interface parameters for amplified multichannel DWDM applications, using a “black-link” model that defines interoperability at the transmitter-side and receiver-side reference points while allowing the intervening photonic line system to remain implementation-specific. The recommendation addresses metro-oriented amplified DWDM applications and defines channel spacings based on the ITU-T G.694.1 frequency grid. Consequently, an optical transceiver used entirely within a hyperscale data-center Ethernet fabric will typically be specified primarily against IEEE 802.3 PHY requirements and relevant multi-source agreements.
In summary, a coherent pluggable used as a router-to-ROADM, router-to-transponder, or interdomain line-side interface may additionally need to comply with ITU-T OTN and DWDM interface requirements—particularly G.709, G.959.1, G.698.2, and the G.694.1 frequency-grid framework. Compliance depends on the intended application code and reference point, rather than on the module form factor alone.
Standards framing:
A concise way to explain the division of responsibility is as follows:
…………………………………………………………………………………………………………………………………………………………………………………………………………………………………………..
Cloud Capex is Through the Roof:
The magnitude of current cloud capital expenditure is exceptional and is the primary driver for the optical transport market. For example, Oracle, reported $28.5 billion in capital expenditures in its most recent quarter, up 235% year over year, against quarterly revenue of $19.3 billion, up 30%. This illustrates the extent to which an established enterprise-software supplier is being repositioned around AI infrastructure and cloud capacity. See References below for more examples of hyperscaler’s enormous capex.
Chinese hyperscalers are pursuing a similar, if less publicly transparent, investment pattern. Tencent’s capex surged 194% y-o-y in Q2, reaching $7.7 billion, up 68% sequentially. Alibaba’s capex was $9.9 billion for the quarter, up 86% year-over-year and up 156% compared to the previous quarter, used primarily to build AI compute infrastructure. Baidu’s capex for the quarter reached $1.67 billion, surging 219% year‑on‑year and 96% quarter‑on‑quarter. Alibaba and Tencent’s reported capital-expenditure percentage increases greatly exceed their revenue growth, which remained closer to the 15%–20% range. The widening gap between capital intensity and near-term revenue growth is a defining characteristic of the present AI infrastructure cycle.
Nvidia commented that it expects to see another 70% growth in capex of cloud companies for 2027—similar to the growth reported by the TOP 4 Cloud companies in 2026. Nvidia must know this well, since the company is engineering multiple financial projects to support the hyper growth.
Super-cycle, but not in a straight line:
The market is clearly in an optical-interconnect supercycle. However, the present rate of expansion is unlikely to persist indefinitely. Even under a continued high-growth scenario, the forecast assumes that growth moderates from 120% in 2026 to 76% in 2027 and to approximately 20% by 2031.
The base-case forecast adopts a soft-landing assumption: cloud capital expenditure remains strong, but procurement normalizes as the supply chain expands, deployment schedules mature, and customers improve utilization of installed infrastructure. The forecast model assumes a 50% increase in cloud capital expenditure in 2027—below the roughly 70% growth contemplated by the most bullish industry forecasts.
That assumption is deliberately more conservative because optical-component supply chains historically respond aggressively to changes in customer procurement signals. Current conditions already show evidence of this effect: suppliers are expanding capacity and product portfolios rapidly in response to exceptional demand for high-speed pluggable optics.
Historical transceiver sales data from 2004 through 2025 also indicate that the sector is cyclical. Periods of two to three years of strong growth have frequently been followed by flat or declining years as inventory, manufacturing capacity, and end-market deployment plans return to balance. In that context, a soft landing is possible, but a more volatile bumpy-ride scenario—characterized by a supply-chain correction during 2027–2030—remains plausible.
Architecture matters:
The forecast also incorporates a structural counterweight to unit demand: optical-interconnect efficiency is improving. AI-cluster architectures are evolving in ways that can reduce the number of optical transceivers required per accelerator, even as total cluster bandwidth continues to rise.
Several developments are relevant:
-
TPU-based and other tightly integrated accelerator systems can reduce the optical-interconnect requirement per XPU relative to more externally networked cluster designs.
-
Multi-plane network topologies can improve path diversity and bisection bandwidth while reducing the need for redundant interconnect resources in certain deployments.
-
Higher-radix switch silicon enables flatter fabrics, reducing the number of network tiers, hop count, and associated optical interfaces.
-
The transition from 100 Gb/s-per-lane to 200 Gb/s-per-lane signaling—under development in IEEE P802.3dj—raises port density and system bandwidth, but also changes the optics-per-bandwidth relationship.
Accordingly, the relevant demand metric is not simply accelerator count. It is the interaction among accelerator architecture, scale-up versus scale-out topology, switch radix, oversubscription design, optical reach, packaging approach, and the extent to which the deployment uses pluggable optics, linear-drive optics, linear-receive optics, co-packaged optics, or near-packaged optics.
Light Counting Report Scope:
The report covers more than 100 Ethernet optical-interconnect product categories, including:
-
100G, 200G, 400G, 800G, 1.6T, 2.4T, and 3.2T retimed transceivers.
-
Linear pluggable optics and linear-drive/linear-receive optical architectures, including LPO and LRO.
-
Co-packaged optics and near-packaged optics, including CPO and NPO.
-
Products segmented by reach, interface technology, and application.
The market analysis is segmented by end customer into cloud, telecommunications, and enterprise markets. The cloud segment is further divided by customer group—top-five U.S. cloud providers, top-five Chinese cloud providers, and other cloud operators—and by application: conventional cloud and front-end networking, AI scale-out networks, and AI scale-up networks.
Conclusions:
The Ethernet optical transceiver market is benefiting from an AI-driven infrastructure build-out that is materially stronger than expected only six months ago. Yet the same intensity that creates the 2026 upside also increases the risk of a later inventory, capacity, or capital-expenditure correction. The key analytical question for 2027–2031 is no longer whether the market will grow; it is whether the industry can transition from extraordinary demand acceleration to a sustainable deployment cadence without repeating the supply-chain overshoots that have characterized prior optical-networking cycles.
Addendum – The Ultra Ethernet Consortium (UEC):
The UEC is separate from IEEE 802.3, but its work is deliberately built on—and coordinated with—the IEEE Ethernet PHY and MAC ecosystem. UEC is not an IEEE 802.3 task force and does not itself ratify IEEE 802.3 amendments. Instead, it publishes an Ethernet-based AI/HPC networking specification across multiple layers and works with relevant standards-development organizations where changes or formal standardization are needed. UEC uses and extends Ethernet, rather than that UEC defines a separate Ethernet PHY.
Division of responsibility:
How UEC relates to IEEE 802.3:
UEC’s stated premise is an Ethernet-based, open, interoperable stack for AI and HPC. Its work is designed to retain compatibility with the broad Ethernet ecosystem while optimizing behavior above—and in selected optional cases at—the PHY/link boundary. It does not replace IEEE 802.3-defined Ethernet port rates or optical PMDs.ultraethernet+1
That means:
-
IEEE 802.3df supplies the standardized 800GbE Ethernet MAC/PHY foundation.
-
IEEE P802.3dj is developing 200 Gb/s-per-lane electrical and optical interfaces for 200GbE through 1.6TbE operation.
-
UEC defines how AI/HPC systems can use such Ethernet links more effectively: particularly host-to-network interaction, RDMA-oriented transport, congestion control, traffic distribution, reliability behavior, and large-scale fabric operation.ieee802+2
UEC Specification 1.0 was released in June 2025, with Version 1.0.2 published in January 2026. UEC describes the result as a full communications stack, with scope extending from applications and transport through switching, link behavior, optics/cables integration, management, and security. UEC 1.0 is designed to operate with Ethernet-compatible physical-layer technology. It contemplates 100 Gb/s-per-lane and 200 Gb/s-per-lane signaling, which aligns naturally with the industry transitions covered by IEEE 802.3df and P802.3dj. However, UEC does define optional Ethernet PHY and link-layer features, so it is not accurate to say that it has no Layer 1 or Layer 2 relevance whatsoever.
……………………………………………………………………………………………………………………………………………………………………………………………………………..
References:
https://www.lightcounting.com/newsletter/en/september-2026-quarterly-market-update-371
https://www.lightcounting.com/report/september-2026-ethernet-optics-372
Dell’Oro: Data Center capex grew 92% in 2Q-2026 (caveats galore)
Hyperscaler capex > $600 bn in 2026 a 36% increase over 2025 while global spending on cloud infrastructure services skyrockets
Dell’Oro: 2H2026 Data Center Capex to Accelerate due to massive AI Deployments
Alphabet’s 2026 capex forecast soars; Gemini 3 AI model is a huge success
Bain & Co: AI Infrastructure Buildout Will Require $6 Trillion Revenue by 2031 to Support Massive CAPEX
Impact of optical component shortages & bottlenecks explained + Hyperscaler’s CAPEX
Hyperscaler AI Race: Soaring Capex Wipes Out Free Cash Flow; AGI and Digital Gods
How will fiber and equipment vendors meet the increased demand for fiber optics in 2026 due to AI data center buildouts?
China’s 2030 ICT Plan focus: 50G-PON, AI Traffic and Satellite–Terrestrial Networks
Executive Summary:
China’s Ministry of Industry and Information Technology (MIIT) has issued its 15th Five-Year Plan for the Development of the Information and Communication Industry, setting a 2030 target for a next-generation communications infrastructure with nationwide coverage, high performance, and a foundation for the sector’s broader modernization by 2035.
The plan sets several quantitative goals for 2030:
-
Information and communications industry revenue of RMB 4.1 trillion.
-
Deployment of 50 5G/5G-Advanced base stations per 10,000 people.
-
95% penetration of 5G/5G-Advanced users.
-
320 million fixed-broadband subscribers with access rates of at least 1 Gbit/s.
-
Intelligent-computing capacity of 9,800 EFLOPS.
- Digital economy: Core industries aim to reach 12.5% of GDP by 2030.
- Networks: Build next-generation communications with full coverage and leading performance.
- Security: Enhance domain coordination and practical network/data security.
The strategy also emphasizes tighter satellite–terrestrial integration. MIIT calls for coordinated construction of satellite–terrestrial interconnection centers and development of direct-to-device satellite services and satellite IoT applications. It further identifies 6G base stations and terminals as development priorities, alongside stronger standards frameworks for optical communications, computing networks, and quantum communications.
China’s 50G-PON push:
The plan’s broader infrastructure objectives reinforce China’s interest in 50G-PON as a successor to today’s 10G-class PON deployments. ITU-T G.9804 defines 50G-PON as a higher-speed passive optical network supporting 50 Gbit/s downstream capacity and at least 12.5 Gbit/s upstream capacity. The technology is intended to extend fiber-access performance for residential, enterprise, cloud, industrial, and mobile-network backhaul use cases.
While operators and equipment suppliers in most markets remain in early trial and evaluation phases, China has accumulated practical deployment experience through 136 10 Gbit/s network pilots conducted last year. Those trials evaluated 50G-PON ultra-broadband access, fiber-to-the-room (FTTR) integrated with Wi-Fi 7, and optical-AI integration, as well as cloud-computing and cloud-gaming applications.
At the China International Optoelectronic Expo in Shenzhen, industry participants argued that China’s supply chain is technically positioned for larger-scale 50G-PON deployment. The remaining constraints are primarily economic and operational: system cost, terminal economics, and power consumption remain material barriers to broad commercial rollout.
AI changes the access-network equation:
Unlike several earlier optical-access upgrade cycles, 50G-PON is being positioned against a more immediate traffic and workload driver: AI. Zhang Dechao, deputy head of basic-network research at the China Mobile Research Institute, said token-based AI workloads are reshaping optical-network traffic patterns, particularly by increasing upstream and more symmetric bandwidth requirements.
Historically, consumer broadband traffic was strongly downstream-weighted. AI inference, distributed data processing, edge-cloud interaction, enterprise collaboration, and high-resolution interactive applications are changing that profile. Zhang said uplink traffic now represents roughly 40% to 50% of total traffic in some network scenarios, making symmetrical bandwidth an increasingly important design requirement. He also cited projections that AI inference could account for 25% of total network traffic by 2035 and contribute to more than 63% growth in aggregate traffic demand.
That shift matters for PON evolution. A network designed primarily for high downstream residential consumption is not necessarily optimized for AI-era workloads that require sustained upstream capacity, lower latency, more predictable performance, and closer integration between access, edge, cloud, and data-center infrastructure.
Supply-chain conflict:
AI is also creating an uncomfortable contradiction for the optical-access sector. The same AI investment cycle that is expanding demand for higher-capacity access and transport networks is redirecting capital, component capacity, and engineering resources toward AI data centers, high-speed Ethernet optics, accelerator interconnects, and cloud infrastructure.
Lin Tao, marketing director at component supplier Accelink Technologies, warned that this concentration of investment could constrain PON-specific R&D and production capacity. Suppliers may find higher near-term returns in AI-related optical modules and interconnect products than in access-network components, particularly where 50G-PON volumes remain uncertain.
The implication is straightforward: technical readiness alone will not determine the timing of 50G-PON deployment. Operators must establish sufficiently compelling service, enterprise, mobile backhaul, and AI-connectivity business cases to justify investment amid competition for optical components, manufacturing capacity, and capex.
Chinese optical suppliers nevertheless contend that core 50G-PON technology—including ASICs, optical modules, and devices—has reached maturity. Large-scale shipments of standardized customer-premises terminals are expected to begin next year, potentially marking the transition from pilots to broader commercial deployment.
References:
http://finance.people.com.cn/n1/2026/0909/c1004-40794989.html
https://www.lightreading.com/optical-networking/china-targets-1m-50g-pon-ports-by-2030
China Mobile and Huawei Turn High-Speed Railway 5G-A into a Premium Service
Key take-aways: “16th Smart City and Intelligent Economy Expo” for Huawei, Alibaba & China’s three state backed carriers
China vs U.S.: Race to Generate Power for AI Data Centers as Electricity Demand Soars
Analysis & Economic Implications of AI adoption in China
China’s telecom industry rapid growth in 2025 eludes Nokia and Ericsson as sales collapse
China’s open source AI models to capture a larger share of 2026 global AI market
China ITU filing to put ~200K satellites in low earth orbit while FCC authorizes 7.5K additional Starlink LEO satellites
China gaining on U.S. in AI technology arms race- silicon, models and research
Broadcom lending Anthropic up to $42 billion in yet another AI circular financing deal
Backgrounder:
According to Reuters, which obtained Anthropic’s IPO prospectus, the Big Tech giants are depicted in multiple ways in the IPO filing- as distribution partners, financial backers, computer suppliers, and competitors, all at once. They uncovered the clearest view yet of how circular the AI world really is, and how much an AI lab’s success depends on the giants it’s also competing against.
Reuters was able to calculate that Anthropic pays roughly 16% of every dollar it earns from its cloud partners (Amazon, Google, Microsoft) back to those same big tech partners. It counts the full value of contracts sold through a cloud marketplace as revenue, then treats the cloud platform’s cut of revenue as a marketing cost.
OpenAI does the opposite by only counting what it keeps after the cloud partner takes its share. That difference matters for understanding the big numbers floating around, and how to actually compare the two rivals’ toplines as they both eye the public market.
Broadcom’s relationship with Anthropic spans compute supply, equipment leasing and financing — giving the semiconductor company a central role in Anthropic’s infrastructure buildout. That differentiates it from other major partners and investors such as Amazon which primarily provide cloud infrastructure and distribution for Anthropic’s AI model Claude.
As part of that complex relationship, revealed in the IPO filing, Broadcom has agreed to lend Anthropic up to $42 billion to finance infrastructure spending. In turn, Anthropic stands to become the largest customer in Broadcom’s entire chip design business next year, making their relationship a prime example of the reciprocal spending that has animated AI skeptics on Wall Street, even as the AI lab readies a public offering that could see it valued at $2 trillion.
“It feels that there’s quite a concentrated bet right now on two companies being able to generate enough revenues to support all the financing that’s happened,” said Robert Leitao, managing partner of Rothschild & Co.
Anthropic disclosed that Broadcom’s role in supplying hardware and acting as a financing partner creates “potential conflicts of interest” that might affect Anthropic’s ability to access the computing power needed for its work, according to the prospectus.
The AI lab also warned that Broadcom’s decisions around pricing and hardware could affect its ability to procure enough computing infrastructure. Broadcom did not comment. Anthropic declined to comment to Reuters.
In April, Anthropic announced it was teaming up with Broadcom and Google for a deal that would see Google provide its Tensor Processing Unit (TPU) chip capacity to Anthropic, with the supply coming online in 2027. Broadcom designs Google’s TPUs that are manufactured by TSMC in Taiwan.
A Broadcom circuit board for chip testing is pictured during a lab tour as Broadcom prepares to launch new optical chip tech to fend off Nvidia in San Jose, California, U.S., September 5, 2025. Brittany Hosea-Small · REUTERS via Yahoo Finance.
Anthropic is also relying on Broadcom for equipment leasing and financing. Broadcom could designate a financing partner, and the debt instruments could be converted into Anthropic shares. Anthropic said in its filing it doesn’t expect any notes to be sold before it completes its IPO. The convertible note Anthropic would issue could finance about a third of the $125.2 billion commitment the AI lab has made for a five-year lease of tensor processing unit (TPU) computing capacity.
Other Players:
Broadcom rivals Nvidia and AMD have also provided funding to their own customers, including OpenAI and Anthropic, which the labs then used to pay for access to the companies’ high-powered chips. The concern about such circular financial schemes is that if one domino in the row falls, it will cause a chain reaction that will decimate the AI trade and, as a result, the global equities markets that have benefited from and come to rely on AI firms and hyperscalers.
As to the circularity of all of this financing, the BIS notes (via FT Alphaville):
“… circular relationships make reported demand partly endogenous to firms’ own financing decisions. For example, when a supplier finances a customer, part of the supplier’s revenue growth reflects its own capital investment, rather than organic final demand. This makes it harder for investors, lenders and supervisors to gauge what part of the current AI boom is based on organic demand.”
“The parallel with the telecom boom of the late 1990s is instructive: upstream equipment vendors financed network operators so they could buy the vendors’ equipment. This meant that part of the equipment vendors’ reported sales was being funded by the vendors themselves. For a time, as operators expanded their networks, equipment orders also expanded and vendors booked both the sales and loans as assets. However, when operators’ own revenues failed to materialize or slowed, they could neither repay the loans nor sustain the equipment purchases. Equipment vendors then sustained both financial losses and a loss of sales. Such dynamics may also play out in AI if revenue growth and end user demand fall short of firms’ expectations.”
……………………………………………………………………………………………………………………………………………………………………………………………..
- 2026 Debt Issuance: Projected between $220 billion and $250+ billion for major US hyperscalers (Alphabet / Google, Amazon, Meta, Microsoft, and Oracle), with broader AI-related market debt reaching up to $570 billion.
- 2027 Forecast: Goldman Sachs projects direct hyperscaler debt issuance to nearly double to around $400 billion (Jeff Pu estimates $419 billion) as companies finance over a third of their infrastructure needs.
- Aggregate Capex: Combined capital expenditures are expected to hit roughly $940 billion in 2026 and scale past $1.3 trillion in 2027.
- Market Share: Hyperscaler investment-grade bond sales have jumped from roughly 2% of total US supply (2022–2024) to roughly 9% in 2026.
- Credit Impact: Credit spreads on a 10-year hyperscaler credit basket have widened from historical 40–75bp ranges toward 90bp+, driven by leverage concerns.
- Cash Flow Outlook: S&P Global Ratings expects all major hyperscalers to run negative free operating cash flow through 2026 and 2027, with a cash-flow inflection point not projected until 2028–2029.
Source: Google Gemini
…………………………………………………………………………………………………………………………………………………………
References:
https://www.ft.com/content/87875b20-4081-4511-9afe-4ee389409742
Curmudgeon: Caveat Emptor: Huge Debt and Circular Financing Deals Dominate AI Build-Outs (07/23)
Bain & Co: AI Infrastructure Buildout Will Require $6 Trillion Revenue by 2031 to Support Massive CAPEX
The AI Infrastructure Build-Out: A $10 Trillion Bet on Compute, Power, and Networks
Nvidia CEO Huang: AI is the largest infrastructure buildout in human history; AI Data Center CAPEX will generate new revenue streams for operators
AI risks and backlash increase; Recap of the circular loop of fake AI profits and hyperscaler markups of private AI companies
China vs U.S.: Race to Generate Power for AI Data Centers as Electricity Demand Soars
How will fiber and equipment vendors meet the increased demand for fiber optics in 2026 due to AI data center buildouts?
Expose: AI is more than a bubble; it’s a data center debt bomb
Will billions of dollars big tech is spending on Gen AI data centers produce a decent ROI?
Huge Risks for the proposed $500B AI Investments from Giant Wall Street firms
Can the debt fueling the new wave of AI infrastructure buildouts ever be repaid?
Telcos don’t have an AI problem; they have a voice estate visibility problem
By Satish Barot, Co-founder and CTO, Klearcom with Shazia Hasnie
AI is increasing telecom interdependence
This author has spent around twenty years building telephony products, and the last few watching what happens when AI gets added to them. The models are usually fine, but the voice estates around them are not.
Explanation: The voice estate is everything a voice call touches: the number dialed, the carriers delivering it, the menu that answers, the systems behind it, and in every country/location served.
NVIDIA’s February 2026 survey of roughly one thousand telecom respondents found that sixty percent of organizations are using or evaluating generative AI, up from forty nine percent in its 2024 edition. The voice call that once met a gateway and a queue now meets speech recognition, a language model, a synthetic voice, and also identity checks.
Every one of those can be healthy while the call goes wrong. So the useful test is whether you can prove, from outside your own network, that a real customer’s call still ends the way it should.
Why AI amplifies operational complexity
Old fashioned faults announce themselves: a line drops and something logs it. AI is quieter, and it fails in ways worth naming. It gets worse quietly: a model loses accuracy on one kind of request and nothing reports an error. It’s also confidently wrong. WildASR tested seven speech recognition systems on recordings degraded the way phone audio is. When a caller is cut off mid word, by a network delay or by the system deciding they had finished, the models fill in words nobody said. Further down the line that looks like a clean answer. The same benchmark makes a broader point: robustness measured in one language can substantially mispredict behavior in another, so a release validated in one market tells you little about the next.
Speed fails at the edges, not the average. A system that usually answers in under a second, but takes two and a half seconds on one call in ten, feels broken to those callers and healthy on every voice testing team chart. Nor is the result repeatable: EVA-Bench tested twelve systems and found a wide gap between passing once and passing every time, which undoes testing against a single expected answer.
Then the ones nobody looks at: the fallback menu that takes over when the AI gives up is usually the least maintained thing you own, and it runs exactly when things go wrong. None of this turns a light red on a contact center voice dashboard.
The other exposure is a number that anybody can dial. OWASP ranks prompt injection first among risks to applications built on language models, and a voice channel is an open microphone into the instruction path itself: a caller talks the model out of what it was built and told to do. Mitigations exist and they are worth naming. Treat caller speech as untrusted input and keep the transcript out of the context that carries instructions, constrain what the model can emit to a defined set of intents and slots rather than free text, and require confirmation for anything that moves money, changes credentials, or reads account data back.
The customer journey as the real test environment
Take a composite example, assembled from patterns that recur across multi country voice estates rather than drawn from one operator. An operator runs support numbers in fourteen countries behind one AI system. On Friday the team delivers an updated model and a new synthetic voice. Both pass in testing, no issues. By Monday, the share of calls handled without an agent is up by six points, complaints are up in three countries where the release went live first, and every dashboard is green. Nobody knows why.
The new voice made the menus two seconds longer, and nobody adjusted the window in which the system listens for a caller to stop talking. Callers spoke too early, the system heard fragments, filled the gaps, and guessed just confidently enough to keep a person out of the call.
The headline number improved because failures had stopped escalating. Every part did its job, so nothing raised an alarm. What broke was the outcome, and no single part owns the outcome. People given those same recordings transcribed them easily.
Those failures are not hypothetical. Test calls run by my own team over the past two years have found a global biopharmaceutical company whose menu prompts ran into each other faster than callers could answer, a financial services provider whose speech recognition intermittently failed on the word “agent”, so the route to a human worked on some calls and failed on others, and a healthcare data provider whose misrouted number reached an AI assistant that was never meant to answer it. In every case the systems carrying the call reported nothing wrong.
From system monitoring to outcome validation
Monitoring looks inward: are my systems up? Validation looks outward. It places a real call in country and checks the result against what should have happened: is a customer in this country, on this network, right now, getting the right answer?
Monitoring only sees what you own, and these failures hide inside the part causing them. So the two work together: validation tells you something is wrong and where, monitoring tells you why. Three changes follow. Report three numbers where you currently report one: calls the AI resolved correctly, calls it escalated correctly, and calls it failed. Report the slower calls alongside the average ones, and report by country and carrier instead of one global figure.
Figure 1. A single call crosses the voice estate an operator owns and systems it does not. Monitoring covers only the shaded stage. Validation traverses the whole path from the caller’s side.
Continuous testing as an assurance layer
The industry already has vocabulary for this. TM Forum’s autonomous network levels give operators a shared scale for how much of the “operate, assure and optimize” loop runs without people, and the same NVIDIA survey places 88% of organizations at Levels 1 to 3 on a scale of 0 to 5. Moving up that scale rests on evidence about delivered outcomes.
If the journey is what breaks, the journey is what you test. That means dialing the real public number, from a real device, on a real carrier in the country the customer is calling from, rather than looping a call back inside your own data center. It means testing from the outside in, and it means doing it on a schedule, across every country and carrier your customers use, fixed and mobile.
Figure 2. Testing the journey from the outside in. A scheduled call dialed from a real device in the country the customer calls from, scored on the delivered outcome.
Check outcomes, not connections: audio clear both ways, the right menu, key presses registered, the caller reaching the right team with their details intact. Score it across many calls with a library of realistic phrases, accepting a pass rate rather than an exact match, so decline shows as a trend. Treat a model update like a network change: try it on a few numbers first, then pull it back if the pass rate drops.
Several approaches are in this category. Synthetic transaction monitoring from operator owned probes, carrier side test call generation and crowdsourced testing on real handsets all produce outside in evidence, and they trade off differently on country coverage, cost and how closely the test resembles a real customer’s call. The common requirement is that the evidence originates outside the systems being assured, on a call placed in the country where the customer is dialing from.
Practical implications
Keep human escalation as a safety valve, not a number to drive down: cutting it without checking correctness makes the metric look better while the service gets worse. Then give the outcome an owner. Gartner’s June 2025 forecast that over 40% of agentic AI projects will be canceled by the end of 2027 attributes those cancellations to escalating costs, unclear business value and inadequate risk controls. Model capability doesn’t appear anywhere on that list. This is organizational before it’s technical.
These failures are seams between systems that break without producing an error, and monitoring your own equipment cannot see them: the gap sits between what your systems report and what your customer gets. Networks have been tested end to end for decades rather than trusted part by part. Voice estates deserve the same.
References
- NVIDIA (2026) State of AI in Telecommunications: 2026 Trends. Fourth annual survey report, published 19 February 2026, based on responses from 1,038 telecom professionals worldwide. Summary of findings: Survey Reveals AI Advances in Telecom.
- Tay, G., Ma, W., Lee, J., Tang, Y., Lee, D., Yin, W., Shen, D., Meng, S., Zhu, Y., Li, M. and Smola, A. (2026) Back to Basics: Revisiting ASR in the Age of Voice Agents. Boson AI. arXiv preprint arXiv:2603.25727, 26 March 2026. Introduces the WildASR benchmark. Dataset and code: bosonai/WildASR.
- Bogavelli, T., Gauthier Melançon, G., Stankiewicz, K., Bamgbose, O., Riols, F., Nguyen, H.H., Mehndiratta, R., Brin, L.D., Marinier, J., Subramani, H., Madamala, A., Nemala, S.K. and Sunkara, S. (2026) EVA-Bench: A New End-to-end Framework for Evaluating Voice Agents. arXiv preprint arXiv:2605.13841, 13 May 2026, revised 27 May 2026. DOI: 10.48550/arXiv.2605.13841.
- OWASP GenAI Security Project (2025) OWASP Top 10 for LLM Applications 2025. OWASP Foundation. Prompt injection is listed first, as LLM01.
- Gartner (2025) Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027. Press release, Sydney, 25 June 2025.
- TM Forum (2025) Autonomous Networks Framework v2.0.0 (IG1218F). Introductory Guide, published 20 March 2025, TM Forum Approved 9 May 2025. Levels of autonomy 0 to 5. Evaluation criteria for assigning a level are in Autonomous Network Levels Evaluation Methodology (IG1252).
-
Bain & Co: AI Infrastructure Buildout Will Require $6 Trillion Revenue by 2031 to Support Massive CAPEX
………………………………………………………………………………………………………………………………………………………………………..
About the author:
Satish Barot is co-founder and CTO of Klearcom. He has spent around twenty years building telephony products and leads the engineering team behind Klearcom’s voice testing platform.
Bain & Co: AI Infrastructure Buildout Will Require $6 Trillion Revenue by 2031 to Support Massive CAPEX
Introduction:
Bain & Company estimates that the AI industry will need to generate approximately $6 trillion in annual revenue by 2031 to support the capital intensity (CAPEX) of the global AI infrastructure buildout, including investment in data centers, accelerated computing, memory, networking and power systems. In its latest technology report, released September 29th, Bain projects that new AI-enabled products could account for roughly $4.2 trillion of that revenue requirement. The firm identifies AI-driven innovation across search, advertising, autonomous systems and physical AI as major prospective sources of value creation.
Enterprise adoption could contribute a further $1 trillion to $1.4 trillion annually through productivity gains in software engineering, sales, marketing, customer service and IT operations. Bain characterizes “absorption speed”—the rate at which enterprises operationalize AI—as the emerging competitive variable, with leading AI labs investing more than $9.75 billion in engineering models intended to accelerate enterprise deployment and integration.
“The debate today is fixated on employee productivity. The economics of AI infrastructure demand trillions in new revenue beyond productivity gains. What the industry needs is a wave of innovation that will dwarf what mobile and cloud unlocked,” said David Crawford, chairman of Bain’s global technology practice and lead author of the report.
At a Glance:
- AI infrastructure investment is racing ahead, but generating enough economic value to justify it will require trillions of dollars in new AI-driven revenue.
- Productivity gains from existing enterprise and consumer applications won’t be enough; entirely new markets must emerge to close the funding gap.
- The winners will be those that create breakthrough AI applications that transform industries and expand the global economy.
Consumer AI services, including subscription and advertising-supported offerings, are projected to contribute between $200 billion and $400 billion in annual revenue. This segment is central to the commercial scaling of AI because it extends AI-enabled services to billions of users through consumer platforms, devices and digital-service ecosystems.
“New products and uses that don’t exist today will enable new markets and opportunities from abundant intelligence – these may include drug discovery, mental health and energy generation,” Bain said.
Bain forecasts annual AI infrastructure spending of up to $1.5 trillion by 2031. The figure encompasses greenfield data-center construction, expansion of existing facilities, and continuing investment in GPUs, memory, network fabrics and related infrastructure. Bain assumes that capital expenditure could represent approximately one-quarter of total AI-industry revenue—“an ambitious but reasonable percentage based on trends among cloud providers.” On that basis, the firm calculates that the AI market would need to approach $6 trillion in annual revenue to sustain the projected infrastructure investment cycle.
“The unprecedented speed and scale of the AI buildout, with billions flowing into chips, data centers, networks and power systems, have focused attention on the challenge of building capacity. But the more important question may be whether enough economic value can be created to justify it.”
Consider the scale of investment and the gap between that and the revenue model necessary to fund it.
- The arms race among hyperscalers (Microsoft, Google, Amazon, Meta, and Oracle) is accelerating: Their capital expenditures could reach $780 billion in 2026, nearly five times the level of just three years earlier.
- Leading-edge AI data centers today are approaching 1 gigawatt (GW) of power capacity. By 2027, many are expected to approach 2 GW facilities, with 9 GW campuses emerging by the end of the decade.
The scale of individual AI data-center projects illustrates the infrastructure requirements underlying those projections. Bain said AI data-center size and cost are increasing rapidly, with leading facilities approximately doubling in scale every 12 to 16 months. Meta Platforms’ Prometheus data center in Ohio, for example, had approximately 600 MW of capacity and an estimated cost of $24 billion in 2025, according to Epoch AI. The facility is projected to reach as much as 2 GW of capacity and cost approximately $80 billion by 2027. Epoch AI projects that Prometheus could reach 5 GW by 2029, with costs of up to $175 billion, and 9 GW by 2030, at an estimated cost of $200 billion.
Such growth places data-center infrastructure squarely within the telecommunications and network-infrastructure domain. Multi-gigawatt AI campuses require high-density optical interconnects, large-scale Ethernet or InfiniBand fabrics, low-latency east-west traffic engineering, high-capacity metro and long-haul connectivity, and resilient access to electric generation and transmission capacity. The infrastructure challenge consequently extends beyond data-center construction to the coordinated scaling of semiconductor supply chains, transport networks, power systems, cooling infrastructure and specialized technical labor.
Bain identified electric-grid capacity, access to GPUs and other critical infrastructure components, talent availability, workforce retention, public acceptance and regulatory requirements as important factors shaping the pace and geography of AI data-center deployment. Resource consumption, noise and community impacts are also becoming material considerations in project planning and approval processes.
Governments in the UAE, Saudi Arabia, the European Union, South Korea and the United States are supporting the expansion of AI and data-centre infrastructure, Bain said. The firm described data centres as increasingly important to technology innovation, economic development and national sovereignty.
The Bain report stated: “Capital needs for data infrastructure will remain high … bottlenecks in power, semiconductors, and other inputs carry large capital needs of their own, opening additional entry points for investors. And as sovereign infrastructure becomes a bigger part of national strategies, partnerships offer both a way in and geographic diversification.”
…………………………………………………………………………………………………………………………………………………………………………………………………………………….
Greg Ip of the WSJ: Believers in the artificial-intelligence boom need to take a very close look at this number: 9% of GDP. That is how much American businesses and consumers eventually have to spend per year on the services of companies like Anthropic and OpenAI to justify the staggering sums being committed to the technology right now.
Is it plausible that Americans will spend as much of their income on this one technology as they do on food? Roughly twice what the nation pays for all forms of energy or all computers and software? Seven times what consumers spend on phone, streaming, and internet services combined?
You should be skeptical. Even the most transformative inventions eventually run into the law of diminishing returns: each additional dollar a user spends yields less additional productivity (or enjoyment) than the last. That imposes a natural ceiling. The question, of course, is where that ceiling is. Whether or not you think 9% of GDP is right, you have to care, because this figure isn’t some fever dream: it is implicit in the dollars that investors and companies are committing right now.
The figure is courtesy of Columbia University finance professor Stijn Van Nieuwerburgh. His paper, first presented at the Brookings Institution, showed that the AI build-out is now bigger than any investment boom in American history. His more intriguing, and sobering, statistic is how much revenue AI would have to garner to justify that boom.
Conclusions:
- Dramatic innovation will be required to deliver the revenue necessary to fund the gap.
- The economics required to generate ROI from AI infrastructure are demanding trillions in new revenue, not just cost savings.
- The industry needs a wave of application innovation comparable with what mobile and cloud unlocked, not just productivity gains on existing workflows.
- The infrastructure is being built ahead of the demand curve, and funding it sustainably will require adding approximately 1% to the annual global GDP growth rate.
- The question is whether the applications arrive in time to pay for it.
……………………………………………………………………………………………………………………………………………………………………………………………………………….
References:
https://www.bain.com/insights/topics/technology-report/
https://www.wsj.com/tech/ai/will-america-spend-9-of-its-gdp-on-ai-the-industry-is-counting-on-it-3501bb4f?st=HkbAZs&reflink=desktopwebshare_permalink&mc_cid=65590ec418&mc_eid=5917efd771
The AI Infrastructure Build-Out: A $10 Trillion Bet on Compute, Power, and Networks
Dell’Oro: Data Center capex grew 92% in 2Q-2026 (caveats galore)
Dell’Oro: 2H2026 Data Center Capex to Accelerate due to massive AI Deployments
PwC: Global AI data center spending to hit $31.6tn by 2050; Role of full stack orchestration layer explained
Nvidia CEO Huang: AI is the largest infrastructure buildout in human history; AI Data Center CAPEX will generate new revenue streams for operators
AI risks and backlash increase; Recap of the circular loop of fake AI profits and hyperscaler markups of private AI companies
China vs U.S.: Race to Generate Power for AI Data Centers as Electricity Demand Soars
How will fiber and equipment vendors meet the increased demand for fiber optics in 2026 due to AI data center buildouts?
Expose: AI is more than a bubble; it’s a data center debt bomb
Will billions of dollars big tech is spending on Gen AI data centers produce a decent ROI?
Huge Risks for the proposed $500B AI Investments from Giant Wall Street firms
Can the debt fueling the new wave of AI infrastructure buildouts ever be repaid?
Meta’s Petal Subsea Cable to Bring Petabit-Class Optics to the Atlantic
Executive Summary:
Meta has announced Petal, a 7,000-km subsea cable connecting the U.S. and France that is designed to carry 1 petabit per second, or 1,000 Tbps, which is roughly twice the capacity Meta attributes to the current top transoceanic systems. If delivered as announced, it would be the first ocean-spanning subsea system engineered for an aggregate design capacity of 1 petabit/s (1,000 Tbps), It is a planned system—not an operational one. Petal is expected to enter service in 2029. At a glance:
- Petal, the next step in Meta’s subsea innovation, will be the first subsea cable to deliver petabit capacity at transoceanic distances, connecting France and the United States over approximately 7,000 km (4,300 mi).
- Expected to enter service in 2029, it will be the first subsea cable system to deploy multi-core fiber technology at scale, doubling the capacity per fiber without a proportional increase in power or physical infrastructure.
- Petal will be built in partnership with NEC and Sumitomo Electric Industries, with support on the French landing from Orange.
- As AI and cloud workloads grow, moving enormous amounts of data between global data centers becomes increasingly important.
- Owning more of the underlying network infrastructure can give hyperscalers greater control over capacity, reliability and future expansion.
Image Credit: Meta
……………………………………………………………………………………………………………………………………………………………………………………………………………………………
Detailed Discussion:
Petal is a roughly 7,000-km / 4,300-mile transatlantic cable connecting the United States and France. Meta announced it on September 21, 2026, saying it will be built with subsea-system supplier NEC, fiber supplier Sumitomo Electric Industries, and with Orange supporting the landing on France’s Atlantic coast.
At 1 Pbps, the raw line-rate equivalent is 125 TB/s—before protocol, FEC, framing, and operational overhead. That is a system-level aggregate capacity, not a claim that any one customer, application, wavelength, or AI training job receives a sustained 1 Pbps connection.
The technical breakthrough:
The key is not merely higher baud rates or more efficient coherent DSP. Petal’s headline architectural change is space-division multiplexing inside the optical fiber itself: two independently usable cores in each fiber.
A conventional single-mode fiber has one light-guiding core. In a two-core fiber, two separate cores sit within the same cladding, creating two spatial channels per fiber. Meta says Petal will be the first system to deploy multi-core fiber at transoceanic distance and that the approach doubles capacity without a proportional rise in physical infrastructure or power.
Conceptually: System capacity=(fiber pairs)×(spatial paths per fiber)×(usable spectrum)×(spectral efficiency)\{System capacity}
Petal attacks the second term. Rather than trying to keep extracting more bits per hertz from a single optical core, it introduces another spatial path in the same fiber. This matters because modern submarine systems are encountering increasingly difficult tradeoffs among:
-
Fiber-pair count and cable diameter.
-
Repeater count, electrical feed limits, and wet-plant power.
-
Amplifier noise and achievable spectral efficiency.
-
Modulation reach over a 7,000-km amplified path.
-
Manufacturing, installation, and repair complexity.
The practical attraction is clear: add spatial capacity before attempting a disproportionate increase in per-core spectral efficiency. This is broadly analogous to the industry’s movement toward more fiber pairs in modern submarine cables, but implemented here through a multi-core-fiber approach.
Why two cores—not many?
Multi-core fiber is not a new research subject, but deploying it in a long-haul undersea system is a materially different engineering proposition from demonstrating it in a laboratory or terrestrial trial. For a transoceanic cable, the relevant questions include:
-
Whether inter-core crosstalk remains acceptably low across the full submerged route and lifetime.
-
Whether repeater/amplifier architecture can amplify both cores efficiently and reliably.
-
Whether field splicing, branching, terminal equipment, fault isolation, and repairs are operationally manageable.
-
Whether yield, mechanical reliability, and cost work at industrial cable-production scale.
-
Whether the added spatial dimension produces capacity gains without undermining the wet plant’s power and reliability economics.
A two-core design is a conservative first operational step relative to more ambitious multi-core approaches. It aims to create meaningful spatial multiplication while keeping the fiber, repeater, and marine-system engineering tractable.
Why it matters for AI:
“AI workload” should not be interpreted as a single workload continuously transmitting a petabit per second over the Atlantic. The more credible rationale is that AI turns inter-data-center transport into a more strategic and less fungible infrastructure layer.
AI raises the value of predictable global capacity:
Large AI clusters are often concentrated where power, land, chips, and data-center construction capacity are available. The data, users, safety systems, content pipelines, training artifacts, model checkpoints, and inference services are global. That creates several high-bandwidth flows:
-
Replication and synchronization of massive data repositories across regions.
-
Model checkpoint and artifact movement among research, training, and serving locations.
-
Distributed training and experimentation, where the bandwidth and latency penalty must be judged against the value of pooling scarce accelerators.
-
Movement of data for preprocessing, evaluation, fine-tuning, and global content or telemetry analysis.
-
Serving-plane transport among regional inference clusters, content-delivery systems, and core application infrastructure.
-
Resilience capacity, enabling a large region to shift traffic or recover faster following a fiber fault, landing-station incident, power disruption, or data-center impairment.
For Meta, Petal also supports the broader reality that its global consumer platforms, data centers, content systems, and AI infrastructure require persistent, high-volume transatlantic connectivity. Meta explicitly frames the project in terms of rising global connectivity demand, while associating the capacity increase with a broader digital-services buildout.
Latency still constrains AI architecture:
A U.S.–France subsea path of roughly 7,000 km has an unavoidable propagation floor. Light travels through fiber at about two-thirds of the speed of light in vacuum, so the one-way physical propagation component alone is on the order of 35 ms, with actual end-to-end latency higher after route geometry, terrestrial legs, switching, and equipment delay.
That means Petal is highly valuable for bulk transport, replication, inference backhaul, data movement, and resilience. It does not eliminate the architectural preference to keep tightly synchronized, latency-sensitive distributed training within a metro, campus, or regional geography. For training workloads with frequent all-reduce operations, the speed-of-light constraint remains decisive.
The likely consequence is not “one worldwide AI supercomputer.” It is a fabric of large regional AI clusters connected by increasingly enormous interregional and intercontinental pipes.
Economic and network significance:
Petal is another indicator that hyperscalers are becoming direct builders and de facto strategic operators of global submarine infrastructure—not merely anchor tenants buying capacity from consortium cables or wholesale carriers.
Meta says it has invested in more than 20 subsea cable projects and cites Project Waterworth as part of that broader effort. This is strategically important for several reasons.
Capacity control:
Owning or controlling cable capacity gives a hyperscaler more freedom to engineer traffic, schedule upgrades, reserve restoration capacity, and match network expansion to data-center deployment. It reduces exposure to capacity scarcity on high-demand corridors and can improve economics relative to repeatedly purchasing long-term capacity leases.
Route diversity and resilience:
A new direct U.S.–France route can improve route diversity, although diversity is real only if the cable’s landing stations, terrestrial backhaul, marine path, and network interconnection are genuinely differentiated from existing failure domains. A cable does not create resilience merely by being new; it must avoid common choke points and be integrated into a broader mesh with restoration options.
Supply-chain positioning:
Petal strengthens the positioning of NEC and Sumitomo Electric in the strategically important subsea market. At the same time, it shows that advanced fiber technology—not just transponder generation or more fiber pairs—is again becoming a major differentiator in submarine-system design.
The utility model is changing:
Traditional cable consortia typically divided ownership and capacity among telecom operators. Hyperscaler investment has shifted the market toward private or hyperscaler-led systems optimized around cloud, content, and AI traffic rather than generalized carrier demand. Petal intensifies that shift: the network’s economic center of gravity is moving toward companies that own both the workload and the data-center footprint.
Important caveats:
The announcement is technically consequential, but it is important not to overstate it.
-
It is a design target, not deployed capacity. Petal is expected to enter service in 2029; its 1-Pbps performance has not yet been demonstrated in an operational transoceanic cable.about.fb
-
“1 Pbps” is aggregate system capacity. It is not a single end-to-end flow, and it should not be treated as equivalent to application throughput.
-
The precise terminal-line-system details remain undisclosed. Meta has not publicly specified the usable optical spectrum, individual wavelength rates, modulation formats, amplifier configuration, repeater spacing, fiber-pair count, landing points, spectrum allocation, or upgrade roadmap. Those details determine how the headline capacity is realized in practice.
-
AI is an important demand driver, but not the only one. The cable will carry a mixture of Meta traffic: consumer application traffic, content systems, cloud-like internal workloads, data replication, inference-related flows, and capacity reserved for protection and growth.
-
“No proportional increase” is not “no increase.” Meta says Petal can transfer twice the data without a proportional increase in power or physical infrastructure. That is a meaningful efficiency claim, but it does not mean the system avoids higher absolute power, equipment, manufacturing, or deployment requirements.
Conclusions:
Petal’s real importance is that it moves multi-core fiber from an advanced optical concept toward transoceanic commercial deployment. The project indicates that the next major submarine-capacity step may come not simply from better coherent optics or wider spectrum, but from adding spatial channels within the wet plant while constraining cable size and power-feed requirements.
For AI infrastructure, Petal is best understood as a global data-center interconnect and capacity-control asset. It will not overcome latency physics or make transatlantic synchronous training universally practical. But it can make it far easier for Meta to move large data sets, replicate state, balance workloads, support cross-region inference and services, and operate a more resilient global AI and application fabric. If it reaches service in 2029 at its stated performance, it will establish a consequential new benchmark for subsea-system architecture.
……………………………………………………………………………………………………………………………………………………………………………………………
References:
Announcing Petal, a First-of-its-Kind Transoceanic Subsea Cable
Inside Petal: Building the World’s First Petabit-Class Transoceanic Subsea Cable
https://tech.facebook.com/engineering/2022/02/economic-impact-subsea-cables/
TechCrunch: Meta to build $10 billion Subsea Cable to manage its global data traffic
Hyperscalers Dominance of Subsea Cable Capacity to Increase in the AI Era
Fiber Optic Networks & Subsea Cable Systems as the foundation for AI and Cloud services
Subsea cable systems: the new high-capacity, high-resilience backbone of the AI-driven global network
FCC updates subsea cable regulations; repeals 98 “outdated” broadcast rules and regulations
Echo and Bifrost: Facebook’s new subsea cables between Asia-Pacific and North America
NEC completes Patara-2 subsea cable system in Indonesia
Meta’s “Iris” AI Chip for MTIA: Implications for Telecom-Grade Optical Networking, DCI and High Capacity Ethernet Fabrics
Google Cloud announces TalayLink subsea cable and new connectivity hubs in Thailand and Australia
Ericsson signs MoU’s with Murata and Sumitomo to progress next-gen mobile network infrastructure
Introduction:
Ericsson has signed separate memoranda of understanding with Murata Manufacturing and Sumitomo Electric Industries to investigate component, radio, and network technologies for next-generation mobile infrastructure.
The collaborations are intended to link Japanese component and materials expertise with Ericsson’s RAN and system-integration capabilities, with an emphasis on technologies that could improve radio-network performance, lower energy consumption, and strengthen the resilience and trustworthiness of communications infrastructure for the 6G and AI era. Key takeaways:
- Ericsson, Murata Manufacturing and Sumitomo Electric Industries to explore advanced technologies aimed at enhancing network performance and improving product energy efficiency in the 6G and AI era
- Ericsson’s new R&D Center in Yokohama to provide a platform for closer collaboration with Japan’s technology ecosystem
- Collaborations to combine the advanced technology capabilities of Murata and Sumitomo Electric Industries with Ericsson’s global R&D capabilities and mobile network leadership
Focus on RAN innovation:
The MoUs provide a framework for joint technology assessment, engineering knowledge exchange, and proof-of-concept work. The companies will examine how advances in components, interconnects, radio hardware, and related technologies can be incorporated more effectively into future RAN platforms and programmable network architectures.
The work is expected to address a central 6G engineering challenge: translating innovation at the component level into system-level gains in capacity, coverage, energy efficiency, reliability, and operational flexibility. That includes evaluating technologies that can support increasingly software-defined, AI-assisted, and cloud-integrated mobile networks without compromising the performance and determinism required at the radio edge.
Linking components to systems:
Murata brings capabilities in electronic components and RF-related technologies, while Sumitomo Electric contributes expertise spanning optical communications, connectivity, materials, and related infrastructure technologies. Ericsson contributes global RAN research, network-architecture expertise, and experience in the design and deployment of programmable mobile networks.
Rather than positioning the agreements as product-development commitments, the MoUs establish an exploratory basis for identifying technical areas where Japanese component leadership can be aligned with Ericsson’s end-to-end network and RAN roadmap. The intended outcome is to accelerate the transition of promising technologies from component-level research into scalable network solutions relevant to both Japan and international markets.
Yokohama R&D role:
The collaborations will draw on Ericsson’s global R&D organization and its Yokohama R&D Center, which began operations in April 2026. The center develops advanced radio hardware and software for Japanese and global markets, including programmable network technologies, next-generation mobile systems, and open network architectures.
The Yokohama facility also serves as a collaboration point for customers and technology partners and supports Ericsson’s participation in international standardization. Its role is therefore broader than local product development: it connects Japanese technology innovation with Ericsson’s global research programs, ecosystem activities, and standards engagement.
Strategic significance:
For Ericsson, the agreements reinforce a long-term industrial commitment to Japan at a time when 6G research is increasingly focused on the interaction among advanced semiconductors and components, radio systems, AI-native network functions, cloud platforms, and energy-efficient infrastructure.
For the wider industry, the significance lies in the effort to close the gap between high-value component innovation and deployable network architecture. Future 6G systems will depend not only on new spectrum bands and radio techniques, but also on advances in RF front ends, antenna and packaging technologies, optical and electrical interconnects, power efficiency, and software-controlled RAN platforms. These collaborations position Ericsson, Murata, and Sumitomo Electric to evaluate where such technologies can deliver measurable system-level value in future mobile networks.
In August, NTT DoCoMo selected Ericsson’s RAN Compute platform for its networks in Japan, including for its 5G operations. The deployment is expected to improve network quality, and allow for some AI-native and programmable networks with an eye on long-term software evolution.
Executive Quotes:
Hiroshi Izumitani, Executive Vice President (Board Member), Director, Communication & Sensor Business Unit at Murata Manufacturing Co., Ltd., says:
“Murata has sought to address the challenges faced by its customers and society, contribute to solving them through technology, and support the advancement of culture. Through this collaboration, we aim to deepen our understanding of the needs and system requirements for next-generation communications infrastructure and explore the potential of our technologies to deliver value.”
Hirotake Iwadate, Executive Officer and General Manager, Transmission Devices Division, Sumitomo Electric Industries, Ltd., says:
“Sumitomo Electric recognizes the importance of long-term collaboration in advancing communications technologies. Through this collaboration with Ericsson, we look forward to deepening technical dialogue, exchanging expertise, and exploring opportunities for future innovation in communications and digital infrastructure technologies.”
Chafic Nassif, President of Ericsson Northeast Asia, says:
“Japan is home to many of the world’s most advanced component technologies. Through closer collaboration with Murata and Sumitomo Electric Industries, Ericsson aims to combine system-level knowledge from global network deployments with Japan’s deep component expertise, creating opportunities to accelerate innovation and expand the application of these technologies across communications, AI infrastructure, data centers, automotive systems and other advanced industries.”
………………………………………………………………………………………………………………………………………………………………………………..
References:
https://www.telecoms.com/5g-6g/ericsson-teams-up-with-murata-and-sumitomo-for-next-gen-networks
Yokohama announced as site of new Ericsson Japan R&D Center
4.8 GHz to 4.9 GHz frequency band uses & Verizon Wireless experimental license for testing ISAC with Ericsson
Ericsson and MediaTek Demonstrate 3GPP-Based GNSS RTK Positioning with Sub-30cm Accuracy Over a Commercial 5G Network
AT&T’s 600 MHz Deployment with Ericsson: Turning Low-Band Spectrum Into Coverage and Uplink Capacity
Ericsson leads SK Telecom AI RAN vs. NVIDIA’s GPU centric AI RAN Alliance
AT&T/Ericsson Demonstrate 5G-Based ISAC for Drone Detection at World Cup Stadium
Google’s Project Suncatcher: Satellite Orbit Validation of AI Accelerator Compute and Thermal Management
Executive Summary:
On Thursday, October 1st, Google plans to launch an experimental satellite designed to assess whether AI inference workloads can operate correctly in low Earth orbit. The spacecraft, designated MVP, is a technology demonstrator for Project Suncatcher, Google’s research initiative exploring space-based, solar-powered AI infrastructure.
Last month, technicians in protective suits and hairnets inspected, handled and tested the refrigerator-sized satellite commissioned by Google. They evaluated its deployable solar panels, which will unfold after launch and orient toward the sun. The spacecraft then underwent vibration testing to determine whether launch loads could damage its onboard processors or compromise mechanical assemblies. Technicians also applied witness marks across fasteners to identify any loosening during the test.
The satellite passed the vibration test: its fasteners remained secured, and its chips showed no apparent damage. James Manyika, Google’s senior vice president for research, described the outcome as “great,” while noting that orbital operations remain the more consequential test.
Project Suncatcher seeks to evaluate the technical viability of placing AI-compute infrastructure in space, where photovoltaic power is potentially abundant and uninterrupted by terrestrial weather or nighttime cycles. On Oct. 1, the MVP spacecraft is scheduled to launch aboard a SpaceX Falcon 9 from Vandenberg Space Force Base near Santa Barbara, California. Google provided The New York Times with an early inside look at the project, which would have appeared largely science fictional only a year ago.
Elon Musk, Jeff Bezos, Sam Altman and others have pledged support for orbital data centers, but the concept remains constrained by significant technical and economic barriers. These include launch cost, radiation tolerance, thermal management, intersatellite communications, orbital operations and eventual spacecraft disposal. At the same time, mounting local opposition to terrestrial data-center construction, together with power-grid, land-use and transmission constraints, has increased industry interest in off-planet computing infrastructure.
Google is not launching a data center. MVP is an experimental precursor intended to validate selected subsystem and operational assumptions. The spacecraft carries four tensor processing units (TPUs), specialized AI accelerators whose aggregate compute capability is approximately comparable to that of a single data-center server. Its solar-array system will provide roughly 1 kW of power, broadly comparable to the consumption of a household hair dryer.
That power budget is sufficient to evaluate how Google’s hardware performs under orbital radiation, vacuum and thermal conditions. The spacecraft will process simple AI queries and is intended to operate for approximately one year, although it is expected to remain in orbit for as long as six years before orbital decay causes atmospheric reentry and burnup.
Google tested A.I. chips at Crocker Nuclear Laboratory in Davis, Calif., with a particle accelerator known as a cyclotron. The goal was to test whether the chips could survive radiation in space.
Photo Credit…Jason Henry for The New York Times
………………………………………………………………………………………………………………………………………………………………….
Mr. Manyika emphasized that Google’s expectations are measured. “We don’t expect, to be perfectly frank, that we’ll have anything usefully operational in the next few years,” he said, comparing the mission to the company’s early efforts to build driverless cars. “Remember how Google was researching for like 15 years, before anything showed up? I think this is going to look like that.”
Scaling from one technology-demonstration satellite to a distributed orbital-computing system would require substantial capital and years of development, according to Brandon Lucia, a professor of electrical and computer engineering at Carnegie Mellon University. “If you do this on a large scale, there are additional engineering problems,” he said. “That is uncharted waters.”
From concept to flight test:
Project Suncatcher originated with Blaise Agüera y Arcas, a Google vice president and AI researcher who leads a team focused on intelligence research. Approximately three years ago, he attended a gathering of entrepreneurs and AI researchers centered on the escalating energy requirements of AI systems. He left convinced that space-based computing could eventually provide access to large-scale solar generation.
The idea “has been on my mind since I was kid,” he said. “There are longstanding ideas in science fiction about using stars for computation.”
Mr. Agüera y Arcas subsequently presented the concept to Mr. Manyika, who was initially skeptical but agreed to investigate whether AI processors could survive the radiation environment of space and be cooled effectively in vacuum.
In February 2025, Google began exposing AI chips to radiation at the Crocker Nuclear Laboratory in Davis, California. There, a cyclotron subjected the chips to radiation doses intended to approximate five years of space exposure. Radiation can induce “bit flips”—single-event errors that alter a circuit’s binary state from zero to one or from one to zero. Such faults can degrade or interrupt computation and can be particularly consequential in AI accelerators, memory systems and control electronics.
Google’s tests produced encouraging results. The company found that restarting the chips could generally clear the observed bit flips, suggesting that reset and recovery mechanisms may mitigate at least some radiation-induced errors. The test does not, however, eliminate the broader need for fault tolerance, error detection and recovery across a space-qualified computing system.
In May 2025, Mr. Agüera y Arcas joined a meeting arranged by Mr. Manyika to present the project to Sundar Pichai, Google’s chief executive. Sergey Brin, Google’s co-founder, also attended.
Mr. Brin and Mr. Pichai quickly greenlit the project. “OK, so this is a good idea,” Mr. Brin had said, according to Mr. Agüera y Arcas. “Let’s talk about how we’re doing it.”
Google has not disclosed Project Suncatcher’s budget. The company has said it expects orbital data-center costs to approach terrestrial data-center costs in the mid-2030s, assuming continuing reductions in launch costs. That assumption is central to the commercial premise: spacecraft hardware, launch, insurance, operations, networking and replacement cycles must collectively become competitive with land, power, cooling, grid interconnection and construction costs on Earth.
Satellite platform and thermal design:
Google contracted with Planet Labs, a satellite-imagery provider in which it had previously invested, to develop spacecraft capable of carrying its AI processors. James Mason, Planet Labs’ chief space officer, said discussions with Mr. Brin about performing computing tasks in space had occurred over several years, although the concept had previously appeared more distant.
“Back then, it seemed further off,” Mr. Mason said. “That was really before large language models took off and A.I. demand really started going exponential.”
Planet Labs agreed to launch two Google satellites in 2027. Google subsequently sought an earlier on-orbit demonstration and accepted additional programmatic risk to accelerate the schedule, according to Eric Stevens, a director of systems engineering at Planet Labs. To meet that timeline, Google integrated its AI chips into an existing Planet Labs satellite platform and initiated qualification testing.
Thermal management is among the program’s most consequential engineering challenges. AI accelerators produce substantial heat during computation, while convection-based cooling systems—including conventional fans—cannot operate in vacuum. Heat must instead move through conductive paths and be rejected through radiation.
Google’s design uses a layered thermal architecture. TPU devices are mounted on a green motherboard, above which sits thermal interface material—a compliant, pale-green compound supplied in sheets and intended to improve heat transfer between the chips and the adjacent metallic heat-spreading structure. Aluminum and copper layers conduct heat away from the motherboard to a radiator panel, which rejects thermal energy into space.
The initial system will operate in duty cycles rather than continuously. Travis Beals, Google’s senior director of product management for Project Suncatcher, said the chips can operate for approximately 15 minutes before they must be shut down to cool. Within those intervals, the processors will handle short inference requests for Google’s Gemini AI system.
The scaling challenge:
Google’s roadmap extends beyond the MVP mission. The company plans to launch two additional satellites next year and has developed concepts for constellations of more than 80 spacecraft flying in close formation and communicating with one another while processing AI workloads. Google is also evaluating the prospect of a purpose-built spacecraft approximately the length of a soccer field.
The key question is not whether a few AI accelerators can operate in orbit, but whether an orbital compute system can scale economically and reliably. A commercially useful architecture would need to solve several interdependent issues:
-
Radiation hardening, fault detection, redundancy and recovery for processors, memory, networking and spacecraft-control systems.
-
Continuous thermal rejection at substantially higher compute densities than the MVP demonstration.
-
High-capacity intersatellite links and ground connectivity capable of moving model inputs, outputs and potentially model parameters.
-
Autonomous fleet management, precise formation flying, collision avoidance and debris-risk mitigation.
-
Launch, replacement and disposal economics that compete with terrestrial data-center construction and power procurement.
-
A sustainable operating model for systems whose computing resources, maintenance cycles and network topology are inherently orbital rather than terrestrial.
The MVP mission does not resolve those issues, but it should generate operational data on the foundational constraints: radiation effects, thermal behavior, processor reliability, power availability and the feasibility of serving simple AI inference requests from orbit.
“If, five years from now, everything we’ve done has worked perfectly, it probably means we’ve not taken enough risk and we’ve not learned as much as we could,” Mr. Beals said. “If we’re really successful with this in the long run, this will ultimately be boring and people won’t think anything at the fact that their Gemini query might be getting served in space.”
References:
https://www.nytimes.com/2026/09/24/technology/google-suncatcher-ai-data-center-space.html
