Nokia’s ECE Divestiture: Impact on Private 5G Market & Disaggregation
Executive Summary:
On October 13th, Nokia said it will sell its Enterprise Campus Edge (ECE) business to AURELIUS, a private-equity investor focused on corporate carve-outs. The planned transaction is expected to close in the first half of 2027, subject to customary closing conditions. Financial terms were not disclosed.
ECE is a substantial part of Nokia’s enterprise private-wireless franchise. Its portfolio spans private LTE and 5G networking through the Digital Automation Cloud, industrial edge and micro-edge computing, industrial devices, and software applications for AI-enabled industrial use cases.
Nokia’s ECE business serves nearly 900 enterprise customers in manufacturing, mining, logistics, airports, ports, and power generation. Its digitalization platform connects industrial assets and processes operational-technology (OT) data on premises, positioning data security, sovereignty, automation, worker safety, and sustainability as core value propositions. The unit reportedly employs about 1,000 people and generates more than €170 million in annual revenue, yet Nokia concluded that its enterprise-campus economics did not fit its post-restructuring priorities in fixed networks, mobile infrastructure, defense, data centers, and AI.
Strategic Rationale:
For Nokia, the sale is consistent with its stated objective of concentrating capital and resources on businesses where it believes it can achieve the greatest long-term differentiation and value creation. Nokia Chief Corporate Development Officer Konstanty Owczarek characterized the divestment as portfolio sharpening, while emphasizing continuity for ECE customers and employees during the transition.
For AURELIUS, the attraction is a globally deployed private-5G and industrial-edge platform with an established enterprise customer base, including Fortune 500 accounts. Fabian Steger, Managing Director of AURELIUS Funds IV and V, said the firm sees scope to expand ECE’s position in private 5G and industrial edge, particularly in AI-powered industrial automation.
Operational implications:
ECE will operate as a standalone company under AURELIUS ownership. Nokia said ECE employees are expected to transfer with the business at closing, in accordance with applicable local employment processes, and that the two companies will cooperate to preserve customer and employee continuity.
The carve-out removes a business with meaningful enterprise exposure from Nokia’s portfolio, but it does not signal an exit from enterprise connectivity altogether. Rather, it reflects Nokia’s broader restructuring under CEO Justin Hotard: prioritizing fixed and mobile network infrastructure, defense, and AI-oriented network opportunities, while shedding or reviewing non-core units.
Nokia’s earlier strategic review encompassed ECE alongside FWA CPE, site implementation and outside plant, and microwave radio units—businesses that collectively generated approximately €900 million in revenue and a €100 million operating loss over the preceding 12 months.
Why It Matters:
The transaction underscores a widening split in the private-networks market between diversified infrastructure vendors and specialist operators focused on industrial digitalization. ECE’s on-premises OT edge, private 5G, and industrial AI software assets give AURELIUS a platform positioned for manufacturing automation, mission-critical campus connectivity, and sovereign data processing.
For Nokia, the deal further advances its pivot toward network infrastructure aligned with AI-driven traffic growth, data-center interconnect, optical and IP networking, and defense communications. The company has set a target of raising annual operating profit to €2.7 billion–€3.2 billion by 2028, from €2 billion in 2024.
Private 4G/5G Market Assessment:
Private cellular remains a real and growing market: industry estimates cited by TelecomTV value 4G/5G private networks at about $4 billion in 2025, rising to $7.2 billion by 2028, while Berg Insight projects growth from $2.4 billion in 2025 to $12 billion by 2030. The issue is not demand for deterministic, licensed-spectrum industrial connectivity; it is the cost and complexity of selling, integrating, and supporting highly customized campus networks at scale.
Nokia’s move therefore removes a major end-to-end competitor from the broad enterprise campus segment, but not from private wireless altogether. Nokia says it will continue selling radios for private deployments with Nokia and third-party cores, while retaining mission-critical capabilities in defense, rail, and utility-grid communications.
Market Impact:
The deal should intensify competition among vendors that can deliver private 5G as a repeatable product rather than a systems-integration project. It’s summarized in this table:
| Market segment | Likely effect of Nokia’s ECE sale |
|---|---|
| Enterprise private-5G specialists | Celona, Kajeet, Anterix-adjacent ecosystem players, and similar vendors gain a larger addressable base as ECE becomes a standalone competitor rather than part of a diversified Nokia portfolio. Omdia’s 2026 assessment already treats Celona, Ericsson, HPE, Nokia, Samsung, and ZTE as leading end-to-end private-5G vendors. |
| Mobile operators | Nokia’s reduced direct enterprise engagement may reduce channel conflict with operators selling private 5G to the same factories, ports, and campuses—an issue Fierce Network identifies as part of Nokia’s rationale. |
| Systems integrators |
A private-equity-owned ECE is likely to be more open to third-party radios, cores, edge compute, and integration partners, expanding routes to market for component and software suppliers.
|
| Industrial OT and AI vendors | ECE’s emphasis on on-premises OT data, sovereignty, and AI-enabled automation validates the shift from “private 5G as connectivity” toward private 5G as an industrial automation platform. |
| Hyperscalers and cloud providers | The sale reinforces the view that enterprise edge value is migrating toward application orchestration, data pipelines, and AI workloads—areas where cloud-native entrants can compete without owning the RAN. |
Conclusions:
Nokia’s sale of ECE will likely make the private-5G market more fragmented, more multi-vendor, and more focused on industrial outcomes rather than network equipment alone. That favors AI wireless specialists such as Celona, systems integrators like Kyndryl, and silicon innovators such as EdgeQ—but only if they can convert openness into lower total cost of ownership, faster deployment, and proven operational reliability.
The ECE sale removes one major integrated competitor from the enterprise campus. It does not remove the requirement that private 5G operate with industrial-grade determinism, security, and lifecycle assurance. The near-term risk for ECE is transition uncertainty. Its customers will want assurance that product roadmaps, support arrangements, and contractual commitments survive the ownership change.
Longer term, however, AURELIUS’s stated plan to expand ECE into AI-powered industrial automation could make it a more focused, and potentially more aggressive, competitor in manufacturing, logistics, mining, ports, and energy.
References:
https://www.nokia.com/newsroom/nokia-to-sell-enterprise-campus-edge-business-to-aurelius/
https://www.telecoms.com/telecoms-infrastructure/nokia-sells-off-enterprise-campus-edge-business
SNS Telecom & IT: Private 5G for the Defense Sector to hit $2.5 Billion by 2029
SNS Telecom & IT: Private 5G Market to Reach $6.6 Billion as Physical AI Takes Hold
GSA: Global private mobile networks exceed 2,000 worldwide; Ericsson Private 5G from Verizon Business extends beyond U.S.
Private 5G networks move to include automation, autonomous systems, edge computing & AI operations
OneLayer Raises $28M Series A funding to transform private 5G networks with enhanced security
Verizon partners with Nokia to deploy large private 5G network in the UK
EdgeQ’s breakthrough demos and partnerships showcased at MWC 2024 & MWC 2023
EdgeQ Samples World’s 1st Software-Defined 5G Base Station-on-a-Chip
SoC start-up EdgeQ comes out of stealth mode with 5G/AI silicon for 5G private networks/IIoT
Microsoft Maia 300 Accelerator Chip: Co-Designing AI Inference for Improved Performance per Watt
Microsoft’s third-generation custom AI accelerator silicon (made by TSMC) combines a multi-process-node chiplet architecture with higher-density racks, targeting improved compute efficiency and lower infrastructure costs.
Executive Summary:
Microsoft announced its Maia 300 AI accelerator on October 12, 2026, outlining plans for multi-gigawatt deployments beginning in 2027. The company is positioning the third generation of its custom AI silicon as a key component of Azure’s infrastructure for inference workloads, including coding, long-context reasoning, and multi-step AI agent tasks.
Maia 300 reflects a broader shift in AI infrastructure design: optimizing the complete computing system rather than maximizing accelerator performance in isolation. Its chiplet architecture, high-bandwidth memory subsystem, and system-level integration are intended to improve throughput, power efficiency, and the economics of delivering AI services at scale.
Microsoft claims that Maia 300 delivers approximately 40% more tokens per megawatt than leading GPU-based systems. However, independent validation will require comparable workloads, clearly defined power boundaries, and measurements of throughput, latency, and output quality. The accelerator’s significance will ultimately depend on whether its architectural choices translate into lower costs per useful AI task under real-world operating conditions.
Azure Maia 200 now in production; Azure Maia 300 coming in 2027. Image credit: Microsoft
…………………………………………………………………………………………………………………………………………………………………………………………….
At a Glance- Key capabilities of the Maia 300 system:
Less data movement, more useful compute: Moving data efficiently, rather than raw compute power, is increasingly the bottleneck in large inference systems. Maia 300’s software-defined dataflow architecture and AI-assisted kernel optimization work together to maximize locality, reduce data movement and improve utilization across compute, memory and networking resources. By treating dataflow as a first-class design principle, Maia 300 delivers a significant performance-per-watt advantage over traditional GPU architectures.
Higher memory bandwidth within every watt: As AI workloads become increasingly context and memory-bound, Maia 300 delivers industry-leading HBM bandwidth per watt. Eight HBM4 stacks provide 18.0 TB/s of peak memory bandwidth within a 950-watt TDP, enabling efficient execution of large models and long-context workloads. Specialized kernel support uses on-die SRAM as a software-managed scratchpad, reducing pressure on HBM.
Chiplet architecture for enhanced efficiency: A new multi-chiplet architecture separates compute, memory-controller and I/O functions, improving efficiency, scalability and cost structure. Compute chiplets are built on TSMC’s N2P process node to maximize performance-per-watt.
Larger context, richer intelligence: By designing for advanced low-precision formats, including enhanced MXFP4, Maia 300 reduces memory and bandwidth requirements for both computation and KV-cache storage. The result is larger context windows, higher throughput and improved model efficiency within the same power envelope.
Scale-up networking at datacenter scale: Maia 300 treats networking as a seamless extension of the software-defined dataflow architecture. Rather than moving data through separate communication layers, dataflow operations span compute, memory and network resources as a single system. High-bandwidth interconnects between Maia accelerators are optimized for tensor parallelism, while a custom AI transport delivers performant and reliable scale-up communication across accelerators and racks. The AI-optimized transport leverages Ethernet-based networking infrastructure to enable a unified topology that scales efficiently from a single rack to datacenter-scale deployments.
Higher AI throughput from every rack: Maia 300 rack systems bring together 96 accelerators within a 167-kilowatt power envelope, delivering 27.6 TB of HBM capacity and 1.7 PB/s of aggregate HBM bandwidth across a high-speed scale-up domain. That density allows Microsoft to serve more simultaneous AI interactions and larger workloads from each rack, improving datacenter capacity and the economics of expanding AI services.
………………………………………………………………………………………………………………………………………………………………………………………….
Chiplet Architecture and Process Technology:
Maia 300 uses a chiplet architecture that partitions functions across multiple dies manufactured using different semiconductor process technologies. According to the reported design, the primary compute chiplets use TSMC’s advanced 2 nm-class technology, while other components use a 3 nm process. Other components, including those handling data movement between the processor and the rest of the system, use TSMC’s less expensive 3-nanometer process.
This approach differs from the Nvidia AI accelerator architecture described in the Reuters article, which combines a large central compute die with surrounding high-bandwidth memory chips. Microsoft’s design assigns different functions to different process nodes rather than manufacturing all components using the most advanced process.
This approach reflects an important principle in accelerator design: not every component benefits equally from the most advanced manufacturing process. Allocating compute-intensive functions to leading-edge technology while implementing other functions on different process nodes can improve manufacturing flexibility and potentially reduce overall costs.
However, process-node selection is only one aspect of the design. The performance of a chiplet-based accelerator also depends on die-to-die interconnect bandwidth, communication latency, packaging efficiency, power delivery, and thermal management. These factors determine how effectively the individual components operate as a unified processor.
For AI inference, this system-level perspective is particularly important. Increasing arithmetic throughput provides limited benefit when memory access or data movement becomes the dominant bottleneck. The architectural challenge is therefore to balance compute resources, memory bandwidth, and communication paths to sustain high utilization across representative workloads.
Memory Bandwidth and Inference Performance:
Memory architecture is central to the performance of large language model inference. Maia 300 incorporates eight high-bandwidth memory (HBM4) stacks, with Microsoft reporting peak memory bandwidth of 18.0 TB/s and a thermal design power of 950 W per accelerator.
These specifications provide useful indicators of the accelerator’s intended operating envelope, but peak bandwidth alone does not establish application performance. Actual results depend on memory access patterns, model architecture, batch size, sequence length, cache behavior, and the efficiency of the software stack.
Inference also encompasses distinct computational phases. During prompt processing, or prefill, accelerators process input tokens and may be constrained primarily by compute throughput. During autoregressive token generation, or decode, performance can become more sensitive to memory bandwidth, model-weight movement, and the size of the key-value (KV) cache used to retain context.
Long-context reasoning and agentic workloads introduce additional demands on memory capacity, data movement, and repeated model execution. Consequently, evaluating Maia 300 requires more than peak compute specifications: it requires sustained measurements across realistic inference workloads, including both throughput and latency.
Microsoft’s emphasis on inference-oriented design is therefore significant. The practical advantage will depend on how effectively the architecture converts available compute and memory resources into useful output tokens while meeting application-level latency requirements.
Rack Density, Power, and System-Level Efficiency:
Microsoft plans to deploy Maia 300 in racks containing 96 accelerators, with reported power consumption of approximately 167 kW per rack. For comparison, Nvidia’s GB300 rack configuration has been reported to contain 72 GPUs and consume approximately 142 kW. These figures illustrate the increasing power density of AI infrastructure, but they do not establish comparative performance or energy efficiency. Accelerator counts, rack configurations, and power measurements must be interpreted in the context of the complete system, including memory, networking, power conversion, and cooling.
Dividing reported rack power by accelerator count yields approximately 1.74 kW per Maia 300 accelerator position and 1.97 kW per GB300 GPU position. These are normalized rack-level figures, not measurements of individual accelerator power consumption. Differences in system configuration and measurement boundaries also limit their value as a direct comparison.
Microsoft separately claims that Maia 300 delivers approximately 40% more tokens per megawatt than leading GPU-based systems. This is a potentially important improvement because energy consumption is a major component of the cost of operating large-scale AI infrastructure. Nevertheless, the claim cannot be evaluated rigorously without identifying the comparison systems, workloads, software configurations, and power boundaries.
“It’s basically delivering more intelligence for every watt, every dollar, and every byte of memory by optimizing across the entire system,” Rani Borkar, president of Azure hardware systems and infrastructure at Microsoft, told Reuters.
A meaningful comparison should report aggregate token throughput alongside time to first token, inter-token latency, and sustained utilization. It should also distinguish accelerator or rack power from total facility power, which includes cooling and other datacenter overheads.
Ultimately, the relevant metric is not simply how many accelerators fit into a rack or how much power each rack consumes. It is how much useful inference work the complete system delivers for a given energy budget, while satisfying latency and quality requirements.
Hardware–Software Co-Design and Infrastructure Economics:
Custom AI silicon gives Microsoft greater control over the interaction between accelerator architecture, inference software, networking, and Azure’s datacenter infrastructure. This control can enable workload-specific optimization, improved resource utilization, and greater flexibility in deploying models across different hardware platforms.
However, custom silicon does not automatically produce lower costs. Its economics depend on manufacturing yields, packaging costs, software maturity, deployment scale, utilization, and the engineering effort required to optimize models and inference runtimes. Hardware specialization can also introduce trade-offs in flexibility as models and workloads evolve.
Microsoft’s reported plans for Maia 300 deployments spanning multiple gigawatts beginning in 2027 indicate the scale at which these considerations become important. The company also expects capital expenditures to exceed $175 billion in calendar year 2026, reflecting a broader investment in AI infrastructure rather than spending attributable solely to Maia 300.
At this scale, small improvements in sustained throughput per watt can have substantial operational consequences. Higher throughput may increase service capacity within a fixed power envelope, while lower energy consumption per token may reduce operating costs. Yet the financial benefit depends on actual utilization, workload mix, hardware lifecycle, and the total cost of operating the infrastructure.
The strategic objective is therefore broader than reducing dependence on third-party accelerators. It is to improve control over the cost, capacity, and performance of AI services through coordinated hardware and software design.
Deployment Scale and Validation Challenges:
Microsoft’s planned Maia 300 deployment raises several engineering questions that cannot be resolved by specifications alone.
First, how does the accelerator perform across different model sizes, context lengths, and inference patterns? Second, can its reported efficiency advantage be sustained under production concurrency and latency constraints? Third, how do networking, memory utilization, and software optimization affect performance when hundreds or thousands of accelerators operate as part of a larger service?
Independent evaluation should compare Maia 300 against clearly identified competing systems using equivalent models, output-quality requirements, precision settings, and inference software configurations. Measurements should include tokens per second, time to first token, inter-token latency, energy per generated token, and cost per million tokens. Agentic workloads may additionally require cost and latency measurements per successfully completed task.
These results should be reported with explicit power boundaries and sufficient methodological detail to permit reproduction. Without such evidence, peak specifications and vendor-reported efficiency figures remain useful indicators of design objectives, but not definitive proof of superior system performance.
Conclusions:
Microsoft Maia 300 illustrates how AI accelerator development is moving beyond the pursuit of peak compute performance toward system-level optimization. Its chiplet architecture, HBM4 memory subsystem, and integration with Azure’s inference infrastructure reflect the importance of balancing computation, memory bandwidth, data movement, and power consumption.
The central question is whether these architectural choices deliver sustained improvements in useful inference throughput, latency, and cost under real production conditions. Microsoft’s reported 40% advantage in tokens per megawatt is a promising claim, but its significance depends on transparent and comparable measurements.
As AI deployments grow, the competitive advantage of custom silicon will increasingly be determined by the efficiency of the complete system—not the specifications of the accelerator alone. Maia 300’s long-term importance will depend on whether Microsoft can translate hardware–software co-design into measurable gains in service capacity, energy efficiency, and the cost of delivering useful AI results.
…………………………………………………………………………………………………………………………………………………………………………………………………………………………………………..
References:
Reuters, “Microsoft plans huge deployment of its own AI chips, with an eye on saving costs,” Oct. 12, 2026.
Microsoft, official Maia 300 announcement and technical architecture materials, Oct. 12, 2026.
Nvidia, GB300 rack architecture and system power specifications. Cite the specific product documentation used to establish the 72-GPU configuration and 142 kW power figure.
TSMC, semiconductor process technology documentation. Cite the relevant process-node documentation to substantiate the manufacturing-node assignments reported for Maia 300.
Google’s TPU Business Outpaces Rivals as Hyperscalers Accelerate Custom AI Silicon Strategies
AWS to deploy AI inference chips from Cerebras in its data centers; Anapurna Labs/Amazon in-house AI silicon products
Huawei’s Ascend Silicon Roadmap Extends From AI Chips to Cluster-Scale Infrastructure
Custom AI Chips: Powering the next wave of Intelligent Computing
The AI Infrastructure Build-Out: A $10 Trillion Bet on Compute, Power, and Networks
Analysis: Nvidia–MediaTek Deepen AI Platform Collaboration -What it Means for AI RAN & DCI
Will Google Cloud’s AI and data analytics revenue +TPU IP licensing income offset huge AI CAPEX to produce a decent ROI?
OpenAI, Anthropic, and Their Hyperscaler Partners: Huge Financial Risks and Open-Weight Competition
Hyperscaler AI Revenue Hurdle:
In an analysis released last month, Jared Bernstein estimated that the big-spending hyperscalers need $13-$18 trillion in revenues over the next 10 years to justify their AI-infrastructure investments. This is, by definition, speculative work, requiring judgment calls on technical aspects of the analysis, such as discount rates, depreciation schedules, taxes, and revenue paths. So, we were happy to see that several other analysts and researchers have, independently, been working on the same problem. Callum Williams of the Economist, whose produced his own excellent analysis on the topic, put together a comparison of these different estimates, including theirs.
What’s striking is how close each of the estimates are: firms will need between $2.4-3.0 trillion in annual incremental AI revenues by 2030 to justify their big spending; this is about 5-6 times more than they are currently making under the most general possible assumptions. They may clear that hurdle. Ryan Cummings told Axios:
“Can all these people eventually turn this into something profitable? I think they will…I’m not sure who’s going to do that, but I do think there will be trillions of dollars of profits that are up for grabs in the future, but just not in this super-accelerated timeline that is required to justify the investments.”
Hyperscaler Debt & Capex Projections (2026–2027):
- 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.
………………………………………………………………………………………………………………………………………………………………………………………………………..
Hidden Risks for the AI Lab Giants:
Anthropic and OpenAI belong to the group of frontier AI model developers commonly known as “the AI labs.” Their financial performance provides a more direct view of AI business economics than that of the hyperscalers discussed above. Unlike diversified technology companies, whose established businesses can offset AI-related spending and complicate profitability analysis, these firms concentrate their operations on developing and commercializing AI. Their results therefore offer a clearer indication of whether revenue from AI products and services can support the costs of model development, computing infrastructure, and commercialization.
Public disclosures about these privately held companies are consequently valuable. Recent reporting on Anthropic and the Financial Times (FT) [paywall] June 16, 2026 report on OpenAI [1.] provide insight into their revenues, spending, and losses. OpenAI’s reported results also illustrate the importance of distinguishing operating costs from noncash accounting charges when assessing the economics of its underlying business. Substantial losses do not, by themselves, establish that AI is a financial bubble. They do, however, reinforce concerns about whether revenue growth can justify the sector’s spending commitments and valuations.
Note 1. The referenced June 16, 2026, FT report covers OpenAI’s 2025 results and distinguishes its approximately $39 billion net loss from losses of approximately $8 billion after excluding specified noncash items. It does not directly substantiate the stated $1.60 loss per revenue dollar.
AI proponents point to Amazon, Uber, and Tesla as examples of companies that sustained years of losses before becoming profitable. The relevant comparison, however, is not simply whether a company loses money, but the magnitude of those losses relative to revenue and its prospects for improving margins. As shown in the chart below, Anthropic and OpenAI reportedly lost more than $1.76 and $1.60, respectively, for every dollar of revenue in the year examined—more than twice Uber’s loss-to-revenue ratio in 2016 and more than five times Amazon’s in 1999. These comparisons underscore the scale of the challenge: rapid revenue growth must ultimately translate into a sustainable cost structure, rather than continued dependence on external financing.
Underestimated Competition:
Despite their prominence in the AI industry, the frontier-model developers face questions about whether their competitive position can support sustained profitability. They depend heavily on external providers of computing infrastructure, and their ability to convert enterprise adoption into sufficient revenue and margins remains uncertain. Customers are also scrutinizing AI costs more closely, particularly as resource-intensive AI agents increase computing requirements and operating expenses. Efforts to control those costs could constrain the labs’ revenue growth even as AI usage expands.economist
Competition from open-weight models adds to that pressure. AT&T, for example, reported increasing the share of its AI use served by open models from 20 percent in May to 40 percent by September, with savings of up to 80 percent compared with earlier in the year. Such models can offer enterprises greater deployment flexibility and reduce reliance on proprietary-model providers. Although model weights may be available without charge, deployment, inference, integration, and ongoing operation are not cost-free. The competitive issue for OpenAI and Anthropic is therefore whether their proprietary offerings deliver enough additional value to justify a price premium.
The AI labs also lack the diversified earnings base available to hyperscalers. Alphabet, for example, can draw on its established advertising business to support investments in Google DeepMind, custom processors, and data centers. OpenAI and Anthropic have no comparable non-AI profit pool. Until their AI businesses generate sufficient cash flow, they remain dependent on external capital to finance expansion and absorb losses—a funding model whose sustainability depends on investor confidence as well as technical progress.
……………………………………………………………………………………………………………………………………………………………………
| Dimension | Proprietary AI Labs (OpenAI / Anthropic) | Cloud Hyperscalers (Google / Microsoft / AWS) | Open-Weight Alternatives (Meta / Enterprise In-House) |
|---|---|---|---|
| Financial Engine | Purely AI-dependent; high loss-to-revenue ratios. | Diversified profit pools (Ads, Cloud, SaaS). | Subsidized by core business (e.g., Meta’s ad network). |
| Pricing Leverage | High premium; under constant pressure to justify value. | Tied to broader ecosystem infrastructure and compute sales. | Zero licensing costs; infrastructure/deployment costs only. |
| Strategic Vulnerability | High infrastructure dependence and lack of safety cushion. | Massive aggregate ROI requirements to justify $T infrastructure bets. | Cost of long-term deployment, fine-tuning, and specialized engineering. |
………………………………………………………………………………………………………………………………………………………………………………………
References:
https://econjared.substack.com/p/a-budding-consensus-on-ai-revenues
https://econjared.substack.com/p/ai-companies-are-in-a-race-against
AI Debt Wave Implications: Higher interest rates with enormous AI sales required for ROI
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
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?
Perspective & Analysis of ITU-R Workshop on IMT-2030 (6G) terrestrial radio interfaces: February 10, 2027 in Geneva, Switzerland
Executive Summary:
ITU-R WP 5D will hold a one-day IMT‑2030 Workshop in February 2027 during its 54th meeting (February 9-18, 2027 in Geneva, Switzerland). The tentative date of the workshop is February 10, 2027.
There will be presentations from independent evaluation groups, potential IMT-2030 terrestrial radio interface technology proponents, and WP 5D experts explaining the IMT-2030 process, requirements, and evaluation methodology and criteria.
Following the Recommendation ITU‑R M.2160, which defines the framework and overall objectives of the future development of IMT for 2030 and beyond, WP 5D has completed the following draft new Reports related to the IMT-2030 terrestrial radio interfaces:
- “Minimum requirements related to technical performance for IMT-2030 radio interface(s)” (Document 5/116).
- “Guidelines for evaluation of radio interface technologies for IMT-2030” (Document 5/119)
WP 5D is still working on the following draft new Report/revision of Document, which are expected to be completed in October 2026:
- “Requirements, evaluation criteria and submission templates for the development of IMT-2030.”
- Revision 1 of Document IMT-2030/2 on “Submission, evaluation process and consensus building for IMT-2030.”
From the 54th meeting of WP 5D in February 2027, Step 3 of the IMT-2030 developing process “Submission/Reception of the Radio Interface Technology (RIT)/Set of RIT (SRIT) proposals and acknowledgement of receipt” will start in accordance with Document IMT-2030/2(Rev.1). Therefore, in conjunction with the 54th meeting, WP 5D will be holding a workshop on IMT-2030 focusing on the terrestrial radio interfaces.
The objectives of the workshop are as follows:
- to provide information from WP 5D on the IMT-2030 development process, including minimum technical performance requirements and evaluation guidelines;
- to invite proponents of IMT-2030 RIT/SRIT to present current and future development aspects of their technologies;
- to invite registered independent evaluation groups to present information about their evaluation group and planned actions; and in particular;
- to facilitate dialogue between ITU-R WP 5D and potential proponents of technology and evaluation groups.
Further information on the invitation for submission of proposals for candidate radio interface technologies for the terrestrial components of the radio interface(s) for IMT-2030 and invitation to participate in their subsequent evaluation can be found in Add. 2 to Circular Letter 5/LCCE/115.
The IMT-2030 Workshop program can be found on the “Web page for the IMT-2030 submission and evaluation process” (under “Workshops on IMT-2030”). Parties interested in the workshop are kindly requested to check the website before the 54th WP 5D meeting, which includes further detailed information about the workshop registration.
………………………………………………………………………………………………………………………..
Perspective & Analysis:
The February 10, 2027 ITU-R workshop marks the practical launch of the IMT-2030 terrestrial radio-interface submission and evaluation phase (step 3). For the first time, candidate Radio Interface Technology (RIT) and Set of Radio Interface Technology (SRIT) will be submitted by proponents (e.g. 3GPP via ATIS).
ITU-R WP 5D will be moving from broad 6G vision-setting into a process in which proposed air interfaces must be documented, evaluated, compared, and eventually consolidated into the IMT-2030 family.
Recommendation ITU-R M.2160 established the high-level IMT-2030 framework. This workshop puts operational machinery around it. WP 5D has already completed draft reports covering minimum technical-performance requirements and radio-interface evaluation guidelines, while the remaining submission requirements, evaluation criteria, templates, and the revised IMT-2030/2 process document are expected to be completed in October 2026. That sequencing is significant: the rules and scoring framework should be substantially settled before candidates publicly explain how their technologies intend to meet them.
A standards-process checkpoint:
The agenda is structured around three constituencies whose interaction will determine the credibility of the IMT-2030 process:
-
WP 5D will explain the technical thresholds, evaluation methodology, submission procedures, IPR treatment, consensus-building process, and global-circulation steps.
-
Potential RIT/SRIT proponents will have an early opportunity to signal submission plans, technical maturity, and intended differentiators. The anticipated 20–30-minute slots imply substantive positioning rather than brief corporate announcements.
-
Independent evaluation groups will identify their participants, technical scope, modelling and simulation capabilities, and evaluation plans. Their role is central: an IMT process is more defensible when claims are reproducibly tested outside the submitting organizations.
The workshop therefore provides an early readout on a question that matters more than headline peak-rate claims: whether candidate 6G proposals can form a technically coherent, globally evaluable standard rather than a collection of regional or vendor-specific ambitions.
What to watch for:
The most revealing material is likely to be found in the gap between the performance requirements and proponent presentations. Watch for clarity on spectrum assumptions, channel models, deployment scenarios, energy efficiency, coverage, mobility, latency/reliability trade-offs, and implementation complexity. A credible candidate should explain not merely that it can meet an isolated target, but under which assumptions and with what architectural cost.
Equally important is the distinction between an RIT and an SRIT. An RIT is a candidate radio interface; an SRIT is a set of complementary radio interfaces. The SRIT construct gives the process flexibility to recognize that one uniform air interface may not be optimal across all future terrestrial operating conditions. But it also raises an interoperability and fragmentation test: any multi-component proposal must demonstrate a coherent system rationale rather than becoming a vehicle for aggregating incompatible technologies. The workshop’s dialogue between proponents and evaluators should expose how seriously that issue is being addressed.
Strategic implications:
For 3GPP and its ecosystem, the event should be viewed as an external ITU-R milestone rather than the moment when detailed specifications are invented. The commercial and technical center of gravity for a likely cellular submission will remain in the relevant standards-development work; however, ITU-R determines whether a candidate fits the global IMT-2030 framework and can achieve broad international recognition. The workshop creates a public forum for aligning those paths before formal evaluation hardens positions.
For operators and vendors, the event may reveal the balance that future IMT-2030 submissions strike between evolutionary cellular engineering and more ambitious 6G claims. Strong emphasis on measurable performance, transparent assumptions, and evaluator readiness would indicate process discipline. Conversely, presentations dominated by aspirational capabilities without testable evaluation paths would underscore that the technology is still pre-standard research rather than a mature candidate interface.
What Won’t Be Covered in this Workshop:
- IMT-2030 Frequency Arrangements are unknown. That topic can’t be discussed until after WR 27 – October 18 thru November 12, 2027 in Shanghai, China when the IMT-2030 frequencies will be specified.
- 6G core network architecture/specs will be exclusively done within 3GPP, with no ITU-R or ITU-T involvement.
…………………………………………………………………………………………………………………………………………………………..
References:
https://www.itu.int/en/ITU-R/study-groups/rsg5/rwp5d/Pages/ws-20270210.aspx
https://www.itu.int/en/ITU-R/study-groups/rsg5/rwp5d/imt-2030/pages/default.aspx
ITU-R WP 5D Timeline for submission, evaluation process & consensus building for IMT-2030 (6G) RITs/SRITs
Roles of 3GPP and ITU-R WP 5D in the IMT 2030/6G standards process
IMT-2030 (“6G”) Minimum Technology Performance Requirements for Radio Interface Technologies
ITU-R M.[IMT-2030.EVAL] & ITU-R M.[IMT-2030.SUBMISSION] reports: Evaluation & Submission Guidelines for 6G RIT/SRITs (6G)
Sept 2026 3GPP meeting updates for 5G Advanced and 6G planning (with timeline for submisson to ITU-R WP 5D via ATIS)
3GPP approves timelines for Release 21 which will specify 6G RAN, Core and 5G Advanced
Dell’Oro: Global telecom capex increased ~5% YoY; other reports show uneven telco capex
According to a recently published report from Dell’Oro Group, worldwide telecom capex increased about 5% year-over-year (YoY) in the first half of 2026. The stronger-than-expected start to the year follows several years of declining investment and marks an improvement relative to expectations at the beginning of 2026.
Telecom equipment trends remain closely aligned with capex. Aggregate manufacturer revenues across the six telecom equipment programs tracked by Dell’Oro Group—Broadband Access, Microwave Transport, Optical Transport, Mobile Core Network (MCN), Radio Access Network (RAN), and High End Router & Aggregation—also increased approximately 5% YoY in 1H2026.
“The first half was stronger than we expected, but the improving near-term trajectory does not materially change the longer-term capex story,” said Stefan Pongratz, Vice President at Dell’Oro Group. “Operators are in a stronger capacity position following the 5G and fiber investment cycles, and the focus is gradually shifting from coverage toward capacity, modernization, automation, and efficiency. At the same time, improving operator revenues are helping to reduce capital intensity ratios even as network investments remain flat.”
Additional highlights from the September 2026 Telecom Capex report:
- Despite the stronger start to 2026, Dell’Oro Group made only negligible revisions to its longer-term forecast. Worldwide telecom capex is projected to grow at a 0 to 1% CAG between 2025 and 2030, reflecting a more mature investment environment following the major 5G and fiber coverage cycles.
- Capital intensity, which peaked at 18% in 2022, is projected to decline to around 14 percent in 2028, before increasing modestly as early 6G investment emerges.
- Telecom equipment revenues are projected to grow at a 2 to 3% CAGR between 2025 and 2030, outpacing CSP capex, partly reflecting incremental demand from cloud providers.
The Dell’Oro Group Telecom Capex Report provides in-depth coverage of more than 50 telecom operators, highlighting carrier revenue, capital expenditure, and capital intensity trends. The report provides actual and 5-year forecast details by carrier, by region, by country (United States, Canada, China, India, Japan, and South Korea), and by technology (wireless/wireline). To purchase this report, please contact us by email at [email protected]
…………………………………………………………………………………………………………………………………….
Other Voices:
1. Telecoms capex worldwide: trends and forecasts 2020–2032, Rupert Wood, September 15, 2026.
The publicly available summary makes two relevant observations:
-
Telecoms capex has declined since 2022, and whether it will rebound remains an open question.
-
“The useful economic life of telecoms assets is lengthening, and for that reason capex is slowing.”
Its coverage includes eight global regions, 13 individual countries and the EU, with spending broken down by funding source—operators, infrastructure companies, content and application providers, and enterprises—and by network asset type. It also distinguishes infrastructure from technology, new from legacy networks, and growth from maintenance spending.analysysmason
This report provides an independent structural explanation for subdued investment, rather than another operator-specific snapshot. Its inclusion of nontraditional investors also helps clarify why equipment demand and operator capex can diverge. However, the public summary does not disclose an aggregate growth forecast; I would not attribute a numerical forecast to Analysys Mason without access to the report.
2. Primary financial reports from Deutsche Telekom, Orange and Nokia add useful independent context: investment trends differ substantially by network operator and geography, capital intensity can fall even while spending rises, and equipment suppliers increasingly benefit from cloud demand—not just carrier network investment.
- Deutsche Telekom’s cash capex excluding spectrum fell 4.9% to €7.8 billion in the first half of 2026, primarily reflecting the timing of German fiber investment, while group revenue increased 2.4%.
- Orange’s economic capex rose 2.7% on a comparable basis to €3.2 billion, driven by higher investment in Africa and the Middle East; spending excluding that region declined 2.4%. Orange’s revenue grew faster than investment, allowing its comparable capex-to-revenue ratio to edge down to 15.2%.report.
3. Network Equipment supplier results also highlight a source of equipment demand beyond traditional carrier budgets. Nokia reported first-half sales growth of 6%, while second-quarter Optical Networks and IP Networks sales increased 20% and 16%, respectively, at constant currency. Sales to AI and cloud customers more than doubled, and orders from those customers reached €2.8 billion in the quarter. These results support a distinction between a selective recovery in operator investment and stronger demand for optical and IP infrastructure serving AI and cloud networks.
……………………………………………………………………………………………………………………………………………………………………….
References:
https://www.linkedin.com/feed/update/urn:li:activity:7513319856398610432/
https://www.analysysmason.com/research/content/regional-forecasts-/telecoms-capex-forecast/
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
Dell’Oro: Global telecom CAPEX declined 10% YoY in 1st half of 2024
Dell’Oro: 6G RAN Capex to reach $500 billion by 2034 + Counterpoint
Deutsche Telekom: Device-to-Cloud AI Traffic Won’t Force RAN Upgrades or Increase CAPEX
Hyperscaler AI Race: Soaring Capex Wipes Out Free Cash Flow; AGI and Digital Gods
SpaceX Moves to Acquire Nationwide 800 MHz Spectrum for Starlink Mobile D2D Service
Introduction:
SpaceX announced an agreement on Thursday to acquire Grain Management’s nationwide 800 MHz spectrum portfolio, expanding the spectrum resources available for its Starlink Mobile direct satellite-to-device (D2D) service [1.]. The transaction would give SpaceX a low-band complement to its existing global 2 GHz mid-band spectrum holdings and strengthen its position as a potential competitor to established U.S. wireless operators. The agreement covers 100% of Space X’s nationwide 800 MHz spectrum portfolio. Financial terms were not disclosed, and the acquisition remains subject to Federal Communications Commission (FCC) approval.
Note 1. Starlink’s satellite connectivity plans extend beyond cellphones to broader direct-to-device end points. IoT support is explicitly part of its published roadmap; smartwatch support has yet to be confirmed for general capability. Starlink’s “Direct to Cell” branding describes a cellular satellite-access system, not a smartphone-only endpoint strategy. Starlink’s official Direct to Cell page states:
“In addition to expanding mobile coverage, Direct to Cell will enable ubiquitous Internet of Things (IoT) connectivity outside of terrestrial coverage, connecting millions of devices across critical global industries.”
………………………………………………………………………………………………………………………………………………………………
Low-band spectrum complements 2 GHz capacity:
The acquisition would add a coverage layer to Starlink Mobile’s spectrum strategy. While its existing 2 GHz spectrum is intended to support higher-bandwidth services, the 800 MHz portfolio would provide more favorable propagation characteristics, including improved penetration through obstacles such as foliage and building materials.
Those characteristics could help address signal attenuation—a significant constraint for satellite-to-smartphone connectivity. However, favorable low-band propagation should not be equated with assured indoor service; the announcement does not establish the coverage or performance that Starlink Mobile would deliver inside buildings.
The acquisition could expand SpaceX’s ability to provide mobile coverage from orbit, but it does not establish that the company intends to dispense with terrestrial infrastructure.
Regulatory approval and competitive implications:
Separately, the FCC approved Starlink Mobile’s Gen2 constellation application this week, according to the supplied Reuters report. That authorization permits SpaceX to launch 15,000 satellites optimized for operation in the 2 GHz band globally. The constellation authorization and the proposed 800 MHz acquisition are distinct regulatory matters: approval of the former does not constitute approval of the spectrum transaction.
Shares of T-Mobile US, Verizon and AT&T each fell more than 5% in extended trading today. The declines reflected investor concerns that SpaceX’s expansion into mobile connectivity could intensify price competition and pressure incumbent operators’ market share in the mature U.S. wireless market.
The strategic significance is the combination of additional low-band spectrum and an expanded satellite constellation. Whether those assets translate into a competitive alternative to terrestrial mobile networks will depend on regulatory approval and the coverage, capacity and service performance SpaceX ultimately delivers.
Technical implications: propagation, capacity and handset compatibility:
The engineering significance of the acquisition extends beyond improved signal penetration. Satellite-to-smartphone services must accommodate propagation distances far greater than those in terrestrial cellular networks, while maintaining reliable communication with ordinary handsets. Operating frequency, satellite altitude, antenna design and beam geometry therefore jointly determine the achievable coverage and service performance. Lower-frequency spectrum offers a useful propagation advantage, but it does not eliminate those system-level constraints.
Capacity is a separate issue. The description of 800 MHz as a coverage layer and 2 GHz as a capacity layer is a useful shorthand—not an intrinsic division between the two frequency bands. Available bandwidth and the ability to form and reuse satellite spot beams matter alongside propagation. Narrower beams can improve spatial isolation and help manage satellite-link impairments, but they require larger antennas; maintaining coverage with smaller beam footprints can also require more beams or additional satellites. Consequently, nationwide spectrum holdings should not be confused with nationwide terrestrial-equivalent capacity.
Satellite operation requires more than a compatible frequency band:
Direct-to-device systems must compensate for the propagation delay and Doppler shift associated with rapidly moving low-Earth-orbit satellites. In systems serving unmodified LTE smartphones, the network performs that compensation. Residual timing and frequency errors vary across the satellite beam and can disrupt initial access, while longer round-trip delays can reduce the efficiency of retransmission protocols. Low-band spectrum does not remove these protocol constraints.ericsson
Handset compatibility also requires a distinction between two approaches. Satellite systems using a conventional LTE air interface can serve existing smartphones in supported terrestrial cellular bands. By contrast, 3GPP Release 17 NR-NTN introduces satellite-specific capabilities, including handset-based timing and frequency compensation using the device’s location and satellite orbital information. These capabilities require compatible chipsets and devices; support for an 800 MHz cellular band alone does not establish support for NR-NTN. The spectrum acquisition therefore should not be characterized as a 5G NTN deployment without disclosure of the intended air interface and device requirements.ericsson
Nationwide spectrum helps—but does not eliminate interference constraints:
Nationwide exclusive access to terrestrial mobile spectrum can simplify satellite deployment by reducing domestic co-channel coexistence problems. However, satellite beams do not stop at national borders. Operations still require interference controls and coordination to protect terrestrial systems in neighboring countries. Depending on the authorization and deployment model, those controls can include geographic exclusion zones and restrictions on satellite transmission parameters.
Indoor Coverage is Highly Unlikely:
Improved propagation is an engineering advantage; reliable indoor service remains a performance claim that needs supporting specifications or measurements. The technical evidence instead emphasizes the interdependence of frequency, antenna design, satellite geometry and protocol behavior.
………………………………………………………………………………………………………………………………………………
Grain Management Backgrounder:
“For nearly twenty years, we have built Grain to see opportunity where others see complexity,” said David Grain, Founder and Chief Executive Officer of Grain Management. “Our spectrum expertise allows us to connect the strategic value of these assets with the technologies and operators that can realize their potential. This agreement with SpaceX brings that capability to bear at extraordinary scale, with the potential to change where and how Americans connect.”
Grain combines deep sector experience, strategic insight, and capital to unlock spectrum value, advance its productive use, and help operators meet evolving connectivity needs. Grain’s acquisition of the nationwide 800 MHz portfolio from T-Mobile in August 2026 extends that approach – the firm acquired the portfolio in exchange for cash and Grain’s 600MHz spectrum, then pursued multiple use cases that would expand connectivity for American consumers and create value for the industry, Grain’s investors, and the people whose futures they support.
“Spectrum is a finite resource with an expanding role in the economy,” Grain added. “As terrestrial and space-based networks converge, we see a new horizon globally for our spectrum strategy, building on our experience to address larger industry challenges and help shape the next generation of connectivity.”
…………………………………………………………………………………………………………………………………………………………
Separately, Space X today signed a fleet-wide deal with American Airlines to equip over 1,000 mainline aircraft with Starlink internet access. Starlink will bring fast, reliable Wi-Fi to customers throughout their journey, making it easier to stay connected from gate to gate. Whether responding to emails, catching up on messages, streaming favorite shows, browsing the web or keeping the family entertained, customers can enjoy a seamless online experience in the air.
“Our customers have told us that dependable connectivity can make a meaningful difference in how they spend their time in the air,” said American’s Chief Customer Officer Heather Garboden. “Whether they are livestreaming a favorite show, gaming, collaborating in real time on an important work deal, scrolling social media or online shopping, Starlink will give our customers countless options to make the most of their time while traveling. Just as importantly, it provides the foundation for future innovations that will further enhance the customer experience.”
“We’re thrilled to bring Starlink across American’s mainline fleet,” said Lauren Dreyer, Vice President of Starlink Business Operations at SpaceX. “Customers will have fast, reliable internet from gate to gate, whether they are streaming, working, or staying connected to family and friends throughout their travel.”
……………………………………………………………………………………………………………………………………………….
References:
Starlink Mobile: NTN–Terrestrial Convergence, Network Capacity, and the Limits of Disruption
FT: SpaceX considering Starlink Direct-to-Consumer mobile service & terrestrial cellular network infrastructure in the U.S.
Ookla: Starlink a viable competitor for hybrid 5G/NTN services due to network performance improvements and larger coverage area
Ookla: D2D satellite connectivity surged 24.5% during last 9 months; Starlink’s footprint expansion leads the way
US Mobile’s new bundle combines its multi-network mobile service with Starlink residential internet
Direct-to-Device (D2D) satellite network comparison: Starlink V2 (Starlink Mobile) vs “Satellite Connect Europe”
Blue Origin announces TeraWave – satellite internet rival for Starlink and Amazon Leo
China ITU filing to put ~200K satellites in low earth orbit while FCC authorizes 7.5K additional Starlink LEO satellites
Starlink doubles subscriber base; expands to to 42 new countries, territories & markets
Amazon Leo (formerly Project Kuiper) unveils satellite broadband for enterprises; Competitive analysis with Starlink
Deutsche Telekom: Device-to-Cloud AI Traffic Won’t Force RAN Upgrades or Increase CAPEX
Introduction:
Deutsche Telekom is in the process of “upgrading mid-range plans,” according to Dhananjay Mirchandani, DT’s senior VP for group controlling and the group’s next CFO from May 1, 2027. “I cannot recall a single conversation in which Alex Jenbar [CTO of Telekom Deutschland] … or somebody else approached us and said, ‘in addition to whatever we currently have on our roadmaps for RAN modernization, that there is an incremental requirement for an investment to be able to prepare ourselves for additional uplink/AI-related traffic,'” he said. “[That’s] just to give you a sense of the degree of confidence we have in terms of our own capex planning, specifically related to mobile network capacity,” he added.
His assessment on medium-term mobile capacity is bound to be sobering for its radio access network (RAN) vendors in Germany: Ericsson, Huawei and Nokia.
The rise of agentic and physical AI could upend decades of mobile network design that dedicates most bandwidth to download speeds (i.e., downlink) and much less to the data channels for uploads (i.e., uplink). It’s not certain how or when traffic patterns will change in a big way, but telcos are talking about preparing for a wave of AI-fueled traffic coming from smartphones, smart glasses or robots.
The Reality Check – No Incremental RAN Capex:
During an investor event, financial analysts directly pressed Deutsche Telekom’s incoming CFO, Srinivas Mirchandani, on whether capital intensity would escalate after 2027 to handle the upstream traffic generated by physical AI, wearables, and smartphone-based LLM queries.
• The Stand: Mirchandani honestly noted that he could not recall a single internal conversation where the CTO of Telekom Deutschland (Alex Jenbar) requested incremental investment beyond their existing radio access network (RAN) modernization roadmaps to handle AI uplink traffic.
• The Baseline Numbers: DT’s long-term guidance (established at its 2024 Capital Markets Day) mandates that capital spending outside the U.S. and spectrum will sit at 21% of service revenue by 2027. While DT is trading slightly above that now, it is entirely due to heavy fiber-optic rollouts in Germany, not mobile radio expansions.
| Attribute | Deutsche Telekom (Europe Group) | T-Mobile US |
|---|---|---|
| Capex Stance | Rigid cap on mobile intensity; excess savings diverted strictly to German landline fiber infrastructure. | Confirmed network architecture expansion funded via standardized 5G-Advanced cycles. |
| AI Traffic Philosophy | Believes traffic peaks are manageable within existing, scheduled modernization roadmaps. | Views the network explicitly as the “connective tissue for physical AI and AI wearables.” |
| Uplink Solutions | Delaying heavy structural changes; emphasizing spectrum efficiency under legacy boundaries. | Implementing Uplink Carrier Aggregation, Uplink MIMO, and Transmit Switching under 5G-Advanced. |
- Delaying Capex via Software: Panelists from across the telco ecosystem—including Turkcell and Deutsche Telekom’s own group partnering division—noted that AI is actually delaying capex cycles rather than accelerating them. Operators are deploying AI algorithms inside the radio access network to optimize power consumption, conduct predictive capacity planning, and boost edge throughput. In short: AI software is making existing hardware last longer.
- The Opex ROI Conundrum: Telcos are seeing immediate financial returns from “AI for the network” (e.g., DT projects €2.5 billion in cumulative cost savings by 2030 through AI automation). However, the business model for a “network for AI”—where consumers or enterprise clients pay a premium for high-speed, low-latency uplink to feed remote LLMs—remains entirely unproven.
- The 6G Boundary Wall: Based on this author’s carrier checks, most Tier-1 network operators now view massive, structural upstream re-architecting as a “6G-era investment narrative” rather than an immediate 5G sub-6GHz demand. Until dedicated enterprise use cases validate moving further up the value chain, hardware vendors will have to wait for a significant revenue lift from device-to-cloud AI traffic.
Source: Deutsche Telekom, Photo: Norbert Ittermann
…………………………………………………………………………………………………………….
Conclusions:
Deutsche Telekom (DT) indicates that rising device-to-cloud traffic volumes will not necessitate premature radio access network (RAN) upgrades. Technical analyses suggest that upcoming capacity demands driven by agentic AI may not immediately accelerate RAN capital expenditure.
DT does not see AI uplink traffic driving an incremental increase to its RAN capex plans in the medium term. DT is holding a rigid perimeter around its long-term financial guidance. By 2027, capital intensity (excluding the U.S. and spectrum) is strictly capped at 21% of service revenues. Excess capital is not being funneled into expanding cell site capacities for upstream AI traffic; it is being aggressively diverted toward landline fiber-optic buildouts in Germany.
-
- Uplink Carrier Aggregation (CA): Allows the network to combine multiple frequency bands (e.g., mid-band and low-band) exclusively for the upstream path. This maximizes the utilization of already-deployed spectrum without requiring operators to acquire or build new macro tower infrastructure.
- Uplink MIMO & Transmit Switching: Enhances data throughput from consumer AI wearables and smartphones back to the cloud by multiplying the data paths between the device and the tower antenna arrays. [
- Network Slicing: Under 5G SA, carriers can allocate a virtualized, highly efficient “slice” of the existing spectrum explicitly for low-latency AI queries. Slicing can improve resource allocation and protect selected service requirements, but it is not a substitute for sufficient uplink capacity. However, it isolates upstream data paths without needing incremental physical hardware upgrades.
References:
https://www.openranforum.com/home
AI-RAN and Agentic AI get real: Ericsson, Nokia, Verizon & other operators enter into a new network automation era
Cisco Execs: New “Network Supercycle” as Agentic AI Workloads Reshape Telecom Infrastructure
Cisco report: Agentic AI to reshape WAN traffic, AI inference will be ~25% of total traffic by 2035
Nokia’s AI Applications Study: “Physical AI” may require RAN redesign to support high‑volume, low‑latency uplink traffic
STL Partners webinar: Agentic AI needed for RAN autonomy & efficiency
The Financial Trap of Autonomous Networks: Scaling Agentic AI in the Telecom Core
AT&T’s 600 MHz Deployment with Ericsson: Turning Low-Band Spectrum Into Coverage and Uplink Capacity
SoftBank’s Transformer AI model boosts 5G AI-RAN uplink throughput by 30%, compared to a baseline model without AI
T‑Mobile achieves record 5G Uplink speed with 5G NR Dual Connectivity
Telstra achieves 340 Mbps uplink over 5G SA; Deploys dynamic network slicing from Ericsson
Finland’s Elisa, Ericsson and Qualcomm test uplink carrier aggregation on 5G SA network
ITU-R WP5D & 3GPP work schedules for development of IMT-2030 terrestrial radio interface (RIT/SRIT) recommendations
Explanation:
This post is an update of an IEEE Techblog published October 20, 2025. It also includes the 3GPP Release 21 timeline, which was previously published (see References below). This author expects that the proposed IMT 2030 RIT/SRIT specs submitted will be primarily from 3GPP (contributed to ITU-R WP5D by ATIS). However, any ITU member is invited to submit proposals.
ITU-R Working Party (WP) 5D Backgrounder:
WP 5D is responsible for the vision and overall system aspects of the terrestrial component of IMT, comprising IMT-2000, IMT-Advanced, IMT-2020 and IMT-2030; has the prime responsibility within ITU-R for issues related to the terrestrial component of IMT, including technical, operational and spectrum related issues to meet the objectives of current and future IMT systems; is the lead group for the overall maintenance of existing and the development of new Reports/Recommendations on IMT; is responsible for studies related on aspects supporting the evolving deployment of IMT including aspects such as regulatory impacts and operational matters within the purview of Study Group (SG) 5; will continue to work closely with Working Parties 4B and 4C on issues related to the satellite component of IMT; will continue to work closely with other Working Parties on issues relevant for IMT systems; for the study period between WRC-23 and WRC-27, WP 5D has the lead responsibility for the preparatory work on WRC-27 agenda item (AI) 1.7 and collaborates closely with WP 4C on agenda item 1.13 and in particular WP 5D will perform those studies on agenda item 1.13 that were assigned to WP 5D by CPM27-1.
IMT 2030 (6G) Backgrounder:
- Resolution ITU-R 65-1 on the “Principles for the process of future development of IMT-2020 and IMT-2030” outlines the essential criteria and principles that will be used in the process of developing the Recommendations and Reports for IMT-2020 and IMT-2030, including Recommendation(s) for the radio interface specification.
- Recommendation ITU-R M.2160, “Framework and overall objectives of the future development of IMT for 2030 and beyond” identifies six usage scenarios for IMT-2030 and envisions a broad variety of capabilities, tightly coupled with intended usage scenarios and applications for IMT-2030, resulting in a great diversity/variety of requirements. Recommendation ITU-R M.2160 also identifies the capabilities of IMT-2030, recognizing that they will have different relevance and applicability for the different use cases and scenarios addressed by IMT-2030, some of which are currently not foreseen. In addition, IMT-2030 can be applied in a variety of scenarios, and therefore different test environments are to be considered for evaluation purposes.
- A test environment is defined as the combination of usage scenario and geographic environment as described in Report ITU-R M.[IMT-2030.EVAL].
- IMT-2030 Minimum Technology Performance Requirements for Radio Interface Technologies is a report which outlines 20 technical performance requirements (TPR). Seven of them are new and specific to describe the 6G performances. Those IMT 2030 technical performance requirements will be used as unified requirements to evaluate the 6G radio interfaces (RITs/SRITs).
- WP 5D completed the draft IMT-2030 minimum technical performance requirements in February 2026 and the draft evaluation guidelines in June 2026. Both have been submitted to ITU-R Study Group 5 for approval in December 2026.
Image Credit: ITU-R WP5D
………………………………………………………………………………………………………………………………………………………………………………………………
ITU-R WP 5D Work Plan for IMT-2030 RIT/SRITs:
Submission of IMT 2030 RIT/SRIT proposals will begin at 54th meeting of Working Party (WP) 5D (currently planned for February 2027). The final deadline for submissions is 12 calendar days prior to the start of the 59th meeting of WP 5D in February 2029. The evaluation of the proposed RITs and SRITs by the independent evaluation groups and the consensus-building process will be performed throughout this time period and thereafter. The detailed schedule and development process are shown in the two figures below.
………………………………………………………………………………………………………………………………………………………………………………
……………………………………………………………………………………………………………………………………………………………………….
Follow on Process for IMT-2030 Final Approval (author’s conjecture):
- The IMT 2030 RIT/SRITs development are scheduled to be completed at WP5D Meeting #63 in June 2030.
- At the next 5D meeting (#64), likely to be in October 2030 (but not yet scheduled), any draft revisions and specification updates will be reviewed and finalized.
- Following completion in WP 5D, the draft IMT 2030 RIT/SRITs and IMT 2030 Frequency Arrangements [1.], recommendations would proceed to ITU-R SG5 (Terrestrial Services) for adoption and then approval under the applicable ITU-R procedure. The dates and approval route remain to be confirmed. However, final approval is likely to be in November or December of 2030.
Note 1: These two IMT 2030 companion recommendations should be finalized and approved as a package (that didn’t happen with IMT 2020 because the ITU-R M.1036-7 Frequency Arrangements recommendation was approved after ITU-R M.2150 was finalized and approved).
………………………………………………………………………………………………………………………………………………………………………………………………
3GPP Release 21 (6G)Timeline:
The crucial ITU-R/3GPP schedule relationship:
Here is how the ITU-R submission deadline relates to the 3GPP Release 21 freeze:.
…………………………………………………………………………………………………………………………………………………………….
References:
https://www.itu.int/en/ITU-R/study-groups/rsg5/rwp5d/imt-2030/pages/default.aspx
ITU-R WP 5D Timeline for submission, evaluation process & consensus building for IMT-2030 (6G) RITs/SRITs
Roles of 3GPP and ITU-R WP 5D in the IMT 2030/6G standards process
IMT-2030 (“6G”) Minimum Technology Performance Requirements for Radio Interface Technologies
ITU-R M.[IMT-2030.EVAL] & ITU-R M.[IMT-2030.SUBMISSION] reports: Evaluation & Submission Guidelines for 6G RIT/SRITs (6G)
Sept 2026 3GPP meeting updates for 5G Advanced and 6G planning (with timeline for submisson to ITU-R WP 5D via ATIS)
3GPP approves timelines for Release 21 which will specify 6G RAN, Core and 5G Advanced
Warning: 6G core network must not repeat the 5G SA 3GPP architecture specs vs lack of interoperable standards
FCC plans 6G spectrum auctions before IMT 2030 frequencies have been identified and without a 6G frequency arrangement standard
How NTIA “Call to Action for 6G Leadership and Security” might influence 6G/IMT-2030 standards and 3GPP specifications
Comparing AI Native mode in 6G (IMT 2030) vs AI Overlay/Add-On status in 5G (IMT 2020)
AI-Native 6G RAN in Practice: Research and Validation Insights from 6G-MIRAI-HARMONY
AI-Native 6G RAN in Practice: Research and Validation Insights from 6G-MIRAI-HARMONY
By Shammi Thakur, Research Director, MarkNtel Advisors with Shazia Hasnie, PhD
Introduction:
A wireless AI model can perform well in the laboratory and still face a very different test once the radio environment changes. However, propagation conditions shift as users move. Traffic loads vary. Devices and carrier frequencies differ. Hardware introduces imperfections that a clean simulation may not capture. Distributed RANs add another complication because a decision made at one network element can affect conditions elsewhere.
A recent 5G trial shows why the distinction matters. In August 2026, SoftBank and Ericsson reported vendor results from a single commercial-network validation of an AI-native scheduler for link adaptation on SoftBank’s 5G network in Japan. The companies evaluated spectral efficiency, downlink user throughput, robustness, and stability under live network conditions. Average gains across the evaluated areas were about 10% for both spectral efficiency and downlink user throughput, while the highest reported gains reached approximately 25% and 50%, respectively. [1]
Commercial-network testing adds something a controlled simulation cannot: exposure to the variability that the system is expected to handle.
IEEE ComSoc TechBlog has already examined the move from AI as an overlay in 5G toward AI-native 6G, along with the developing IMT-2030 requirements and the role of AI in future RANs. [2][3] This article takes a narrower view. It examines how AI-native RAN functions can be trained, coordinated, and tested when wireless conditions, hardware, data, and network topology all change.
6G-MIRAI-HARMONY provides a concrete research setting for that question. The EU-Japan program connects realistic channel and hardware modeling, AI-based physical-layer functions, distributed coordination, AI-ready RAN architecture, and common approaches to data management, benchmarking, and validation. [4][5]
The Research Challenge for AI-Native 6G RAN:
Training data quietly becomes part of an AI model’s operating assumptions. Conventional wireless algorithms are generally built around assumptions that engineers can state explicitly. Machine learning adds another dependency: the examples used during training and evaluation influence what the model learns to recognize as normal.
A model can consequently look robust while remaining sensitive to conditions absent from its training environment.
Wireless conditions can change without the underlying network function changing. Users move through different propagation environments. Traffic loads rise and fall. Carrier frequencies and device characteristics vary. Hardware can introduce quantization, RF impairments, and converter limitations.
Distributed RANs add another layer. Measurements used by one function may come from elsewhere in the network. Beamforming and scheduling decisions can affect later decision cycles. Fronthaul capacity can determine how much information reaches another processing point before a coordinated action has to be taken.
An AI function therefore cannot always be judged as an isolated block. Network behavior around the function can influence the result just as much as the model architecture itself.
6G-MIRAI-HARMONY: Research Approach and Scope:
6G-MIRAI began on April 1, 2025, and runs for 36 months through March 2028 under Horizon Europe and the Smart Networks and Services Joint Undertaking. Grant Agreement 101192369 lists total project costs of €3,188,935 and an EU contribution of €2,979,966.50. Ericsson France is the coordinating organization. [5]
The European consortium includes Ericsson France, Fraunhofer, Telefónica, IS-Wireless, Sequans, Apple Technology Engineering, KU Leuven, CNIT, and the University of Pisa. The associated HARMONY work in Japan brings in organizations including the University of Tokyo, Kyocera, NEC Networks & System Integration, Shimizu, Tokyo University of Agriculture and Technology, and KDDI Research. [6]
Five Research and Technology Items connect the technical work:
- Realistic models and datasets covering channel, hardware, and measurement conditions.
- AI/ML for physical-layer components, including learning-based approaches to wireless functions.
- AI-driven coordination for distributed and cell-free deployments.
- AI-ready 6G-RAN architecture covering where intelligence, data, and computing are placed.
- Data management, testing, and validation for common scenarios and reproducible evaluation. [4][5]
The sequence matters because each part affects the next. Data conditions influence AI training. AI functions interact with distributed coordination. Architecture determines where models can run and what information they can access. Benchmarking and validation determine how the resulting measurements can be compared.
Realistic Models and AI-Based Physical-Layer Functions:
The wireless environment represented in an experiment becomes part of the AI system whether researchers intend it or not.
6G-MIRAI-HARMONY’s methodology combines link- and system-level modeling with analytical evaluation and later laboratory- or prototype-oriented validation. QuaDRiGa is identified for spatially and temporally consistent channel generation, while Sionna supports differentiable communication-system modeling. Measurement data and experimental RAN data provide additional routes toward less idealized evaluation conditions. [4]
Three parts of the modeling chain deserve particular attention:
- The channel: Propagation, mobility, and spatial variation determine the conditions presented to the model.
- The hardware: RF impairments, quantization, finite-resolution converters, and other non-idealities can alter physical-layer behavior.
- The data: Training and test datasets determine which situations the model encounters and which remain unseen.
Those conditions matter for functions such as channel estimation, CSI prediction and compression, neural receivers, precoding, power control, and link adaptation.
Independent research is addressing similar issues. A 2026 study of CNN-based hybrid precoding for cell-free massive MIMO evaluates imperfect CSI, phase noise, channel aging, molecular absorption, and transceiver hardware nonidealities in a THz simulation environment. The reported results include spectral-efficiency gains against several conventional approaches, while the study also measures the effect of individual hardware impairments on performance. [7]
Physical-layer intelligence therefore has to be tested against the radio conditions it is expected to encounter. An improvement under one channel or hardware assumption does not establish the same improvement under another.
Distributed Intelligence for Cell-Free Massive MIMO:
Distributed access points can work together, but the information needed to coordinate them is distributed as well. Beamforming decisions affect interference seen by other users. Scheduling changes the traffic mix entering later decision cycles. Power-control decisions can alter the conditions under which subsequent optimization takes place.
Fronthaul places a direct limit on how much information can be exchanged. Fully centralized processing can require substantial CSI and data transfer as the number of access points and users grows. Practical designs therefore have to balance cooperation against available transport and processing resources.
Independent cell-free massive MIMO research continues to address that scalability problem. A 2025 IEEE Access study describes network overloading and scalability as major design challenges and evaluates a joint pilot and access-point assignment method in systems tested with up to 150 access points. [8] The paper notes that fully centralized processing can create significant fronthaul signaling and computational demands as deployments expand.
Synchronization adds another requirement. Coherent joint processing depends on suitable phase and frequency alignment across distributed access points. A 2024 IEEE Open Journal of the Communications Society study examines phase misalignment among access points in cell-free massive MIMO and evaluates a partially coherent transmission framework designed to reduce the need for network-wide phase alignment. [9]
AI-driven coordination has to operate inside those constraints. Improving beamforming or scheduling is only part of the problem when the information required for that decision is delayed, compressed, or unavailable at the point where the model runs.
Designing an AI-Ready 6G RAN:
Where the model runs can be as important as what the model does.
A latency-sensitive control loop may need processing close to the network element generating the measurements. Another workload may benefit from a larger computational environment with access to a wider network view. Data access, inference latency, model updates, and control interfaces all affect that choice.
Four practical questions follow:
- Where does inference run?
- Which measurements reach the model?
- How frequently are those measurements updated?
- What happens when the required information sits elsewhere in the RAN?
O-RAN already has work addressing part of this problem. Its WG2 specifications cover the Non-Real-Time RAN Intelligent Controller, A1 and R1 interfaces, AI/ML learning and deployment, and the AI/ML workflow. [10]
O-RAN’s current work does not define the final architecture for 6G. It does show where AI lifecycle and deployment questions are already being addressed in today’s RAN architecture. 6G-MIRAI-HARMONY takes those concerns into a research setting that also includes distributed and potentially disaggregated 6G architectures. [4]
Common Data, Benchmarking and Validation:
Performance figures become difficult to compare when the experiments behind them use different conditions. Two studies might report higher spectral efficiency while using different datasets, channel assumptions, training procedures, hardware models, and test conditions. Without those details, the numbers are difficult to interpret.
The 6G-MIRAI-HARMONY D3.1 deliverable addresses this problem directly. It provides the project’s first consolidated framework for benchmarking and validation, data-management principles, and common scenarios. The document places particular emphasis on repeatability, traceability, comparability, data quality, and reuse. It also defines KPI/KVI considerations and protocols for training, evaluation, and generalization. [11] The evaluation chain is shown in Figure 1.
Figure 1. The evaluation chain for AI-native RAN research.
The scenario defines the conditions represented in the experiment. The dataset determines what the model sees. The model is evaluated through selected metrics and then tested under stated validation conditions.
Changing one part of that chain can change the result. An unseen propagation condition may expose a generalization problem. A realistic hardware model may reduce the gain seen with ideal hardware. A different metric may favor a different approach. Benchmarking and validation also answer different questions. Benchmarking establishes common conditions for comparing approaches. Validation examines whether the resulting system behaves as required under defined conditions.
D3.1 is an existing project output. Later deliverables are intended to add datasets and results covering common scenarios, validation, and benchmarking. Those later outputs should remain described as project work rather than completed findings until the corresponding evidence is published. [11][12]
From Research Validation to Standardization:s
ITU-R’s IMT-2030 work has already moved beyond the initial framework.
Recommendation ITU-R M.2160 established the framework and overall objectives for IMT-2030 in November 2023. The framework includes six usage scenarios, among them Artificial Intelligence and Communication (AIAC), and identifies ubiquitous intelligence as one of its overarching design principles. [13]
In February 2026, ITU-R Working Party 5D completed the draft report on minimum technical performance requirements for IMT-2030 radio interfaces. The draft defines 20 minimum technical performance requirements and provides a consistent basis for specification and evaluation. The draft has been submitted to Study Group 5 for consideration in December 2026. [14] In June 2026, Working Party 5D also completed the draft guidelines for evaluating candidate IMT-2030 radio interface technologies. The guidelines define three evaluation methods—simulation, analytical evaluation, and inspection—and specify seven test environments. The draft has likewise been submitted to Study Group 5 for approval in December 2026. [14]
ITU-R’s candidate radio-interface technology submission process is scheduled to begin in February 2027, with the candidate-submission window extending through February 2029. [14] 3GPP is pursuing AI/ML work within the RAN in parallel. TR 38.745, “Study on Artificial Intelligence (AI)/Machine Learning (ML) for NG-RAN Phase 3,” was created for Release 20 on September 5, 2025, and was placed under change control on March 19, 2026. [15]
Research results do not become standards simply because an algorithm performs well in one experiment. The useful evidence includes the conditions behind the result: the scenario, dataset, metrics, validation method, limitations, and reproducibility of the test.
Conclusions:
AI-native 6G RAN will require more than putting machine learning inside individual RAN functions.
Propagation changes. Hardware is imperfect. Network intelligence becomes distributed. Fronthaul and synchronization can limit coordinated processing. Architecture determines where data and computation are available. Training and evaluation conditions affect what an AI result actually shows.
6G-MIRAI-HARMONY brings these issues into one research program by connecting realistic wireless models, AI-based PHY techniques, cell-free coordination, AI-ready RAN architecture, and common approaches to data, benchmarking, and validation. The work points to a practical requirement for AI-native RAN research: performance results need enough information around them to show how they were produced, where they hold, and what changes when the underlying conditions change.
References:
- SoftBank Corp. and Ericsson, “SoftBank Corp. and Ericsson conduct Japan’s first trial of Ericsson AI in RAN on a 5G commercial network,” August 20, 2026.
https://www.ericsson.com/en/press-releases/2/2026/softbank-corp-and-ericsson-conduct-japans-first-trial-of-ericsson-ai-in-ran-on-5g-commercial-network - IEEE ComSoc Technology Blog, “Comparing AI Native mode in 6G (IMT 2030) vs AI Overlay/Add-On status in 5G (IMT 2020),” January 15, 2026.
https://techblog.comsoc.org/2026/01/15/comparing-ai-native-mode-in-6g-imt-2030-vs-ai-overlay-add-on-status-in-5g-imt-2020/ - IEEE ComSoc Technology Blog, “IMT-2030 (‘6G’) Minimum Technology Performance Requirements for Radio Interface Technologies,” March 17, 2026. https://techblog.comsoc.org/2026/03/17/update-imt-2030-6g-minimum-technology-performance-requirements/
- 6G-MIRAI-HARMONY, “Concept and Methodology,” 2026.
https://6g-mirai-harmony.eu/concept-and-methodology/ - European Commission, CORDIS, “Machine Intelligence based Radio Access Infrastructure (6G-MIRAI),” Grant Agreement No. 101192369.
https://cordis.europa.eu/project/id/101192369 - 6G-MIRAI-HARMONY, “Consortium and Objectives,” 2026.
https://6g-mirai-harmony.eu/consortium-and-objectives/ - Tadele A. Abose and Thomas O. Olwal, “Hardware-Impairment-Aware CNN-Based Hybrid Precoding for Cell-Free Massive MIMO Systems Under Imperfect CSI in Terahertz-Enabled 6G Networks,” Telecom, vol. 7, no. 3, 2026.
https://www.mdpi.com/2673-4001/7/3/70 - Mhammad B. Hmayed and Ali H. Bastami, “Scalable Cell-Free Massive MIMO System: Joint Optimal Pilot and Access Point Assignment,” IEEE Access, vol. 13, pp. 201778–201788, 2025. DOI: 10.1109/ACCESS.2025.3638655
https://ieeexplore.ieee.org/document/11271199 - Unnikrishnan Kunnath Ganesan, Tung Thanh Vu, and Erik G. Larsson, “Cell-Free Massive MIMO With Multi-Antenna Users and Phase Misalignments: A Novel Partially Coherent Transmission Framework,” IEEE Open Journal of the Communications Society, vol. 5, pp. 1639–1655, 2024.
DOI: 10.1109/OJCOMS.2024.3373170.
https://ieeexplore.ieee.org/document/10459246 - O-RAN Alliance, “WG2: Non-Real-time RAN Intelligent Controller and A1,” Technical Work Group scope.
https://www.o-ran.org/technical-groups/wg2 - 6G-MIRAI-HARMONY, “D3.1 — Definition of Initial Common Scenarios, Data Management and Benchmarking Methodology,” 2026.
https://6g-mirai-harmony.eu/wp-content/uploads/2026/02/6G-MIRAI_Deliverable_D3.1_Scenarios_Data_Bench_v1.0.pdf - 6G-MIRAI-HARMONY, “Deliverables, Publications and Workshops,” 2026.
https://6g-mirai-harmony.eu/deliverables-publications-and-workshops/ - ITU-R Recommendation M.2160-0, “Framework and overall objectives of the future development of IMT for 2030 and beyond,” November 2023.
https://www.itu.int/rec/R-REC-M.2160-0-202311-I/en - ITU-R, “IMT towards 2030 and beyond (IMT-2030),” current development, evaluation, and submission process.
https://www.itu.int/en/ITU-R/study-groups/rsg5/rwp5d/imt-2030/pages/default.aspx - 3GPP, TR 38.745, “Study on Artificial Intelligence (AI)/Machine Learning (ML) for NG-RAN Phase 3,” Release 20.
https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=4445
Author Bio:
Shammi Thakur is Research Director at MarkNtel Advisors, with more than 15 years of experience in strategic market intelligence, industry forecasting, and competitive analysis. He leads research initiatives across global technology and communications markets, overseeing the development of market studies and strategic research frameworks. His work focuses on translating complex industry developments into evidence-based insights, with particular experience in emerging technologies, market evolution, and the forces shaping next-generation communication systems.
Website: MarkNtel Advisors
Warning: 6G core network must not repeat the 5G SA 3GPP architecture specs vs lack of interoperable standards
Introduction:
As the global telecom ecosystem pivots from 5G/IMT 2020 towards normative studies for IMT-2030 (6G) in 3GPP Release 20 and Release 21, the air interface (IMT 2030 RIT/SRITs forthcoming recommendation) naturally commands the spotlight. Discussions are dominated by IMT 2030 Technical Performance Requirements like sub-millisecond latencies, Integrated Sensing and Communication (ISAC), and AI-native physical layers. Yet network architects must confront a sobering historical reality: a brilliant radio access network is utterly paralyzed without a functional, universally interoperable core network.
If the telecom industry intends to realize the true commercial and operational promises of 6G, we must urgently dismantle the architectural precedent set during the 5G Standalone (SA) core (lack of) standardization cycle. By bypassing traditional global transport gatekeepers and prioritizing abstract logical modeling over concrete implementation realities, the industry traded the promise of an open, multi-vendor cloud ecosystem for a return to legacy vendor lock-in.
The ITU-T Bypass and the 5G SA Core Network Standardization Gap:
The fragmentation of the 5G SA core network can be traced directly to a structural and geopolitical power struggle.
- Historically, ITU-T Study Group 13 held the global mandate for establishing international standards for non-radio, architectural, and transport layers of all next generation networks. This governance ensured deep multi-vendor interconnectivity and data-plane uniformity across sovereign networks.
- During the development of the 5G Service-Based Architecture (SBA), 3GPP effectively insulated the core network design from ITU-T oversight. 3GPP executives told this author that they did not trust ITU-T to generate the 5G core network standards.
- Instead, the 3GPP specifications (such as TS 23.501 and TS 23.502) were kept entirely in-house and on completion were rubber-stamped by ETSI, 3GPP’s venue host. While this bypass accelerated time-to-market, it created a catastrophic standards-to-implementation gap. 3GPP defined network functions—such as the Access and Mobility Management Function (AMF), Session Management Function (SMF), and User Plane Function (UPF)—as highly abstract, logical blocks.
- While the control plane mandated modern tools like HTTP/2 and RESTful OpenAPIs, this framework merely established interface syntax. It completely omitted the exhaustive behavioral guardrails, edge-case definitions, and low-level realization mechanics required to build a production-ready, cloud-native core. That disconnect led to different network operator implementations of the 5G SA Core, depending on the vendor(s) they selected.
The Illusion of Interoperability:
The practical consequence of this abstraction is well known to any tier-1 network operator that has attempted a multi-vendor 5G SA core network deployment. On paper, the 5G SBA is modular. In reality, an operator cannot reliably procure an AMF from Vendor A, an SMF from Vendor B, and a UPF from Vendor C, connect them via standard 3GPP interfaces, and achieve a stable, carrier-grade network. The specifications are simply too loose.
Crucial underlying implementation details—such as cloud-native container orchestration pipelines, state database synchronization, database persistence layers, and hardware-accelerated UPF data planes—were left entirely outside the scope of the standard. To fill this vacuum, major infrastructure vendors engineered proprietary software logic beneath the “open” APIs. Consequently, multi-vendor core deployments required immense, custom, and cost-prohibitive systems integration efforts. For most operators, the path of least resistance was a return to single-vendor silos.
When 3GPP kept the Service-Based Architecture (SBA) work entirely in-house, they defined the network functions (like AMF, SMF, and UPF) as abstract, logical blocks rather than concrete blueprint specifications:
-
- The Interface Gap: 3GPP dictated the use of HTTP/2 and RESTful OpenAPIs for the control plane. While that sounds open, it merely defined the syntax, not the exhaustive behavior required when edge cases or multi-vendor implementations collided.
- No Implementation Blueprint: Crucial components required to construct an actual, production-ready 5G SA core—such as cloud-native container orchestration pipelines, underlying state database synchronization, database persistence layers, and low-level User Plane Function (UPF) acceleration mechanics—were entirely omitted.
The Practical Outcome: A Return to Vendor Lock-In:
Because 3GPP’s architectural specs lacked concrete realization guidelines and strict inter-vendor edge-case definitions, a major interoperability gap emerged:
- Custom Customization: An operator cannot realistically buy an AMF from Vendor A, an SMF from Vendor B, and a UPF from Vendor C, plug them together over standard 3GPP interfaces, and expect a stable network. The specs are too loose.
- The “Joint Specification” Reality: In practice, every major 5G SA core deployed requires extensive, customized, and often proprietary software engineering ironed out directly between the wireless operator and a single primary core vendor (such as Huawei, Ericsson, or Nokia).
- Network Slicing Disarray: This gap severely crippled Network Slicing. While 3GPP authored elegant logical diagrams for end-to-end network slicing, the absolute lack of unified, cross-domain transport realization standard meant that slicing remained confined to single-vendor testbeds and highly customized, non-scalable deployments for years.
Source: Siarhei Yurchanka/Alamy Stock Photo
The Warning Signs for IMT-2030:
As we look ahead to 6G and IMT 2030 recommendations, the ITU-T remains largely sidelined from mobile core network architecture. 3GPP operates as the de facto absolute authority on both the radio and the cloud core stack. If left uncorrected, the structural loop will repeat. 3GPP will deliver an idealized, highly complex, AI-driven 6G core architecture on paper, leaving actual functional implementations to be sorted out via proprietary vendor middleware.
Light Reading reports, that there are currently three competing 6G core architectures within 3GPP, which threatens to splinter their specs for 6G core networks. Most network operators prioritize affordable evolution, but debates over who controls device intelligence and revolutionary design risk causing divergence. Two fundamental points have emerged: how to handle or embed AI functionality and what to do with non-access stratum (NAS) signaling. NAS is the secure signaling pipeline between the device and core network, handling authentication, encryption, mobility management and session setup. Mobile architectures have relied on it as a foundational pillar since 2G (GSM), embedding it in device baseband silicon and SIM security frameworks.
Reworking NAS requires modem redesign, core security procedure overhaul and device backward-compatibility planning. Inside 3GPP SA2, the study phase has identified three candidate directions for the core network (based on AI functionalities):
- Direction 1 – Separate AI domain: A dedicated AI domain for intent handling independent of the packet-switched network, with the AMF routing NAS signaling while AI fulfillment occurs via SBI; backed by Nokia, T-Mobile USA, Verizon, NVIDIA, Deutsche Telekom, Apple, Qualcomm, Ericsson and NEC, though implementation requires complex UE AI domain client integration.
- Direction 2 – AI functionality in 6G NFs approach: Integrates AI through dedicated 6G network functions following standardized 3GPP procedures with flexible NAS routing (standalone or combined with core functions like AMF); supported by NTT DOCOMO, Samsung, LG Electronics, NEC, IIT Bombay, Vodafone, Apple and China Telecom (partly).
- Direction 3 – AI/agent-handled connectivity approach: Deeply integrates AI/agents into network procedures with dynamic coordination of capabilities and tool invocation, utilizing a signaling routing function (SRF) independent of AMF with user/control plane integration; aligned with Huawei, HiSilicon, China Mobile, ZTE, vivo, CATT, ETRI, Ewha Womans University and OPPO (partly), though it introduces high procedural disruption.
Of these three paths, the first (Direction 1, separate AI domain) has gained the most traction among Western operators and vendors (T-Mobile USA, Verizon, Deutsche Telekom, Nokia, Ericsson, NEC, NVIDIA and, to some extent, Apple). Anchoring the core to the 5G service-based architecture (SBA) interface preserves 5G core investments while letting intent handling mature alongside existing services. However, choosing a direction without standardizing the underlying AI Model Interchange Formats or Agent-to-Agent telemetry will create a brand new flavor of proprietary vendor lock-in.
Conclusions & Engineering Mandates for the 6G Core:
To prevent an amplification of the 5G SA core interoperability deficit, the global engineering community must pivot away from abstract logical modeling. If 3GPP selects an architectural path for IMT-2030 without standardizing the deep execution and semantic layers, we will simply replace legacy infrastructure silos with an unmanageable layer of proprietary AI middleware.
We must advocate for four critical shifts in our approach to 6G core network standardization:
- Mandate Behavioral, State, and Data Realization: Future core specifications must move beyond simple API syntax definitions and logical block diagrams. 3GPP must explicitly standardize end-to-end state-machine behaviors, cross-vendor error-handling conditions, and database synchronization baselines—specifically defining the realization mechanics for stateless network functions and the Unstructured Data Storage Function (UDSF) layer across multi-vendor boundaries.
- Standardize AI Semantic Frameworks and Agent Telemetry: Choosing a 6G core direction without anchoring it to open, deterministic data layers will trigger an unprecedented form of algorithmic vendor lock-in. Standards must rigidly define AI Model Interchange Formats, intent-handling taxonomies, and Agent-to-Agent telemetry protocols. If an autonomous agent invokes a tool or alters network routing dynamically, its procedural boundary conditions must be universally verifiable across competing vendor stacks.
- Bridge the Transport and Cloud-Native Infrastructure Void: 3GPP must abandon its isolationist posture and actively co-author implementation blueprints with open-source infrastructure bodies (such as the Linux Foundation’s telecom initiatives and the Cloud Native Computing Foundation). Specifications must establish standardized, open baselines for underlying container orchestration, state-data persistence, and low-level user-plane hardware acceleration using eBPF (Extended Berkeley Packet Filter) or DPDK (Data Plane Development Kit) architectures.
- Prioritize Cross-Domain Transport Mapping from Day One: Inter-operator roaming and cross-domain networking cannot be treated as downstream implementation details or left to bilateral vendor agreements. 3GPP must natively bake the mapping mechanics between mobile network slices (S-NSSAI) and IETF-defined transport network layers (such as Segment Routing over IPv6 – SRv6) directly into the initial 6G core architecture.
If 6G is to debut as a truly transformative, global platform rather than an incremental upgrade to radio spectral efficiency, we must bridge the chasm between paper standards and software reality. The architecture of the 6G core must be built from its inception for actual cloud realization, absolute multi-vendor interoperability, and rigorous operational clarity. We cannot afford another ghost in the machine.
If 6G is to be a truly transformative global platform rather than an incremental upgrade to radio efficiency, we must bridge the gap between paper standards and software reality. The architecture of the 6G core must be built for actual realization, true multi-vendor interoperability, and absolute operational clarity. We cannot afford another ghost in the machine like we had with the 5G SA core network specs.
References:
https://www.lightreading.com/6g/the-6g-core-divide
https://techblog.comsoc.org/2024/12/18/5g-network-slicing-progress-report-with-a-look-ahead-to-2025/
Building and Operating a Cloud Native 5G SA Core Network
Evaluating Gaps and Solutions to build Open 5G Core/SA networks
Omdia’s 2025 Mobile Core Network Leaders: Huawei #1 in market share; Nokia #1 for portfolio competitiveness
Telco investments in mobile core networks surge 83% in 2025-Q4, but what about ROI?
Téral Research: 5G SA core network deployments accelerate after a very slow start
Dell’Oro: RAN market stable, Mobile Core Network market +14% Y/Y with 72 5G SA core networks deployed
