TL;DR
Thorsten Meyer AI’s first Control Series article argues that AI access in 2026 is being shaped by six chokepoints: power, compute, data, model access, distribution and capital. The source cites reported examples including a global model shutoff, Ukraine’s combat-data licensing, and rival AI labs renting supercomputers from xAI. Several contract terms, model-access details and source documents remain outside the provided material.
Thorsten Meyer AI has published the first article in its ‘Control Series,’ arguing that AI access in 2026 is being shaped by six chokepoints – power, compute, data, model access, distribution and capital – rather than behaving like a neutral utility. The report matters because it frames recent events across government controls, defense data licensing and supercomputer leasing as evidence that AI infrastructure can be throttled, gated or revoked by a small set of holders.
Confirmed: the published source frames the story as Part 1 of an occasional series on where power sits in the AI stack. It identifies six control points: power, compute, data, model access, distribution and capital.
Reported or claimed: the source says a frontier model was switched off worldwide on about 90 minutes’ notice, Ukraine’s Avengers Labs is licensing combat footage for training while keeping improved models, and xAI’s Colossus cluster contains about 555,000 GPUs. It also says Anthropic agreed to pay about $1.25 billion a month and Google about $920 million a month for Colossus output, or roughly $26 billion a year combined.
The article also points to a Memphis power buildout of roughly two gigawatts, a $60 billion interface bet tied to Cursor, and intra-industry financing as examples of control moving to fewer owners. The supplied material cites Anthropic statements, Axios, The Wall Street Journal, Reuters, CBS, TechCrunch, Semafor, Ukraine’s Ministry of Defense, Perplexity Research, Challenger Gray and SpaceX SEC filings, but it does not include those documents in full.
The Six Chokepoints
For a decade AI was sold as a utility — abundant, neutral, always on. In 2026 it became a lever: scarce, controlled, revocable. Here are the six places power actually sits — and who started to squeeze.
Every layer is concentrating into fewer hands, and 2026 is the year the holders stopped treating their leverage as theoretical. A kill switch wasn’t discussed — it was pulled. The utility you’re allowed to forget about; the lever, you have to watch who’s holding. Optionality just became architecture.
Chokepoints Shift AI Power
The report’s main claim is that AI buyers, developers and investors can no longer treat access as a plain utility purchase. If compute, data, power and model access are concentrated, then pricing, uptime, access rules and permitted uses can change quickly.
That has consequences for startups building on frontier models, companies moving internal work to AI systems, and governments trying to regulate the stack. A lab may have the model but not the power. A developer may have the app but not the distribution channel. A state may not build models directly, yet may still control access through permits, export rules or shutdown authority.
Two Channel SXM2 Expansion Board Builts for Data Center GPUs Featuring Advanced 300G Cooling Solution Servers GPU Accelerators Board
Engineered for, the SXM2 two GPU expansion baseboard 300G supports two SXM2 GPUs ( V100) with integrated NVLink…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Utility Pitch Met Its Limits
For about a decade, AI companies often described artificial intelligence as infrastructure: available on demand, broadly neutral and priced by usage. Thorsten Meyer AI says that metaphor helped justify capital spending and reassured customers that AI would be available like the power grid.
The 2026 examples cited in the report challenge that framing. The power example shows how large AI clusters can outgrow local grid capacity. The compute example shows frontier labs renting major capacity from rivals. The data example shows unique war footage being treated as a sovereign asset, not a commodity.
“AI does not flow freely like a utility.”
— Thorsten Meyer AI
ENTERPRISE AI INFRASTRUCTURE: Modern MLOps, Vector Databases, GPU Clusters, and Scalable Data Architecture for LLMs (The Enterprise AI Architect’s Handbook)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Contract Details Remain Open
The supplied material does not identify the government or frontier model involved in the reported global shutoff. It also does not provide the full text of the compute contracts, the exact reclaim clauses, or independent documentation for every dollar figure in the excerpt.
It is also unclear whether the six chokepoints will keep tightening, or whether new competitors, regulation, open models, lower-cost chips or alternative energy projects will reduce the concentration described by the report.
Understanding Open Source and Free Software Licensing
Used Book in Good Condition
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Installments Track Each Chokepoint
Thorsten Meyer AI says later installments will examine each chokepoint separately. The next developments to watch are power permits, cloud and compute leasing agreements, defense-data licensing terms, model-access rules, distribution deals and financing arrangements among AI firms, cloud providers and sovereign investors.
Models, Methods and Tools for Product Service Design: The Manutelligence Project (SpringerBriefs in Applied Sciences and Technology)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
Is this a breaking-news story?
No. This is an analysis piece based on a newly framed Control Series article and the 2026 developments it cites.
What are the six AI chokepoints named in the report?
The report identifies power, compute, data, model access, distribution and capital as the six places where control can be applied.
Are the figures independently verified here?
They are attributed to the supplied source material, which cites named outlets, company statements, Ukraine’s Ministry of Defense and filings. The excerpt does not include the underlying documents.
Why should developers or customers care?
If the report’s thesis is right, access to AI systems can depend on contracts, permits, platform rules and funding sources, not only on technical performance or price.
What remains disputed or unknown?
The model-shutoff details, full contract terms, the durability of compute concentration and the likely policy response remain unsettled based on the material provided.
Source: Thorsten Meyer AI
