TL;DR
Thorsten Meyer AI’s new Control Series frames 2026 as the year AI stopped looking like a neutral utility and started acting like a controlled lever. The report identifies six chokepoints: power, compute, data, model access, distribution and capital.
Thorsten Meyer AI has published the first part of its Control Series, arguing that a series of 2026 developments showed artificial intelligence is no longer behaving like a neutral utility but as a controlled system shaped by a small number of chokepoints.
The article identifies six layers where control over AI is concentrating: power, compute, data, model access, distribution and capital. Its central claim is that recent events showed those layers can be throttled, repriced, leased, withheld or shut off by governments, infrastructure owners, platform operators and major financiers.
The report cites several examples: a frontier model that was switched off worldwide on roughly 90 minutes’ notice; Ukraine’s defense ministry licensing wartime data while retaining improved models; large AI labs renting compute from direct competitors; and major spending on distribution layers such as AI coding interfaces.
The piece attributes its synthesis to public reporting and statements from sources including Anthropic, Axios, The Wall Street Journal, Reuters, CBS, TechCrunch, Semafor, Ukraine’s Ministry of Defense, Perplexity Research, Challenger Gray and SpaceX securities filings from March through June 2026.
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.
Control Moves Into Infrastructure
The report matters because it shifts attention from model capability to the systems that decide who can use AI, under what terms and for how long. If AI access depends on scarce electricity, leased GPU clusters, unique datasets, model permissions, app distribution and concentrated financing, then users and businesses face risks beyond model quality.
For companies building on AI tools, the practical concern is dependency. A model may be available today but limited tomorrow by a government order, vendor policy, contract clause, power shortage or platform decision. The report frames that dependency as an architecture problem rather than a temporary market issue.
For policymakers, the analysis points to a governance challenge. Oversight of AI may increasingly involve energy permits, cloud contracts, defense data arrangements, competition rules and sovereign investment flows, not only model safety evaluations.
high performance AI compute GPU
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Six Layers Under Pressure
The Control Series begins at the physical layer. The report says frontier AI is increasingly constrained by gigawatt-scale power access, citing SpaceX’s Memphis complex and its reported move toward roughly two gigawatts of on-site gas generation.
At the compute layer, Thorsten Meyer AI points to xAI’s Colossus cluster, described as holding about 555,000 GPUs, and says rivals including Anthropic and Google have agreed to rent output from it under large monthly contracts. The report presents this as evidence that leading AI labs often do not fully own the infrastructure they depend on.
The data layer is framed through Ukraine’s Avengers Labs, which the report says licenses annotated combat footage to companies while keeping the improved model. At the model-access layer, the cited example is the rapid worldwide shutdown of a frontier model. Distribution and capital round out the list, with the report pointing to high valuations for AI interfaces and roughly $26 billion a year in intra-industry financing.
“For a decade AI was sold as a utility: abundant, neutral, always on.”
— Thorsten Meyer AI
enterprise AI data storage solutions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Claims Still Need Verification
The report is an analytical synthesis, not a single new filing, court record or company announcement. Some figures and contract terms cited in the piece, including monthly compute spending, cluster scale and clawback provisions, depend on the underlying sources named by Thorsten Meyer AI.
It is not yet clear how durable these chokepoints will be. More energy projects, new chip supply, open-source models, regulation or alternative distribution channels could weaken some points of control. The report argues that 2026 showed leverage being used, but the long-term market structure remains unsettled.
AI model access management tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Installments Track Each Chokepoint
Thorsten Meyer AI says each of the six chokepoints will receive its own installment later in the Control Series. The next test for the thesis will be whether future reporting shows these examples as isolated pressure points or part of a lasting pattern in AI infrastructure.
Readers should watch for new disclosures on energy permitting, GPU leasing, defense and proprietary data deals, model-access restrictions, platform distribution agreements and financing arrangements among AI companies and sovereign investors.
AI distribution platform software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the actual news development?
Thorsten Meyer AI published the first part of its Control Series, setting out a six-part framework for where control over AI infrastructure now sits.
Is this breaking news?
No. This is an analysis based on recent developments and cited reporting from March through June 2026.
What are the six AI chokepoints named in the report?
The six are power, compute, data, model access, distribution and capital.
What is confirmed versus claimed?
The publication of the report and its framework are confirmed by the supplied source material. Specific figures and contract details are claims attributed to the report and its cited sources.
Why does this matter for AI users?
If AI access is controlled through a small number of infrastructure layers, companies and users may face outages, pricing changes, access limits or policy restrictions beyond their control.
Source: Thorsten Meyer AI
