TL;DR
Thorsten Meyer AI published the first part of The Control Series, arguing that AI is shifting from a neutral utility model toward a controlled infrastructure stack. The piece identifies six chokepoints: power, compute, data, model access, distribution and capital.
Thorsten Meyer AI has published the first installment of The Control Series, arguing that artificial intelligence is no longer behaving like a broadly available utility but like a controlled infrastructure system shaped by six chokepoints: power, compute, data, model access, distribution and capital.
The analysis says recent 2026 developments show AI access can be throttled, licensed, repriced or shut off by a small set of actors. It points to a frontier model reportedly switched off worldwide on about 90 minutes’ notice, defense-linked combat data licensed with conditions, and large compute contracts in which AI labs rent capacity from rivals.
The piece identifies power as the base constraint, citing a Memphis complex described as running toward roughly two gigawatts through on-site generation rather than waiting for standard grid interconnection. It also cites xAI’s Colossus cluster at about 555,000 GPUs and says Anthropic and Google entered large rental agreements tied to that capacity.
Other chokepoints named in the analysis are hard-to-collect data, model access, distribution through AI interfaces, and capital. The source frames those areas as places where ownership or regulatory authority can decide who gets access, on what terms, and for how long.
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.
AI Access Becomes Controllable
The analysis matters because it challenges a common assumption behind enterprise AI adoption: that model access, compute and related services will be broadly available so long as customers can pay. If the source’s framing is correct, AI buyers may face a market where infrastructure access depends on power permits, data rights, platform placement, capital relationships and government decisions.
For companies building on external models, the risk is operational as much as strategic. A service can be available one week and limited the next if a supplier, regulator, compute owner or platform changes terms. The article argues that optionality has become part of architecture, meaning firms may need fallback models, varied compute suppliers and clearer data rights before committing core workflows to one provider.
ASUS Dual AMD EPYC 9004 Series 4U NVMe 8X Dual Slot PCIe Gen 5.0 GPU Server (ESC8000A-E12P), 8X Trays, 2X H200 NVL Tensor Core 141GB HBM3e PCIe 5 Accelerator, Rails (Renewed)
No Processor Installed; Supports 2x AMD EPYC 9004 Series Processors
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Six Pressure Points Named
The piece is presented as Part 1 of an occasional series on where power sits in the AI stack. It says the utility metaphor dominated AI sales for much of the last decade because it made AI sound abundant, neutral and always available.
The six pressure points listed are power, compute, data, model access, distribution and capital. The examples include large-scale power generation for AI facilities, concentrated GPU clusters, sovereign control over wartime data, revocable access to frontier models, the value of user-facing AI applications, and large intra-industry financing flows.
The source says its synthesis draws on reporting and statements from outlets and organizations including Anthropic, Axios, The Wall Street Journal, Reuters, CBS, TechCrunch, Semafor, Ukraine’s Ministry of Defense, Perplexity Research, Challenger Gray and SpaceX filings. The excerpt does not provide full underlying documents for each claim.
“AI does not flow freely like a utility.”
— Thorsten Meyer AI dispatch
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.
Claims Still Need Detail
Several claims in the excerpt are attributed to the Thorsten Meyer AI analysis rather than independently documented inside the provided material. The excerpt does not identify the government or model involved in the reported global switch-off, nor does it provide contract language for the compute rental clauses it describes.
It is also unclear how durable each chokepoint will be. Power, GPU supply and capital are current constraints, but technology, regulation and market entry could change the balance. The analysis presents a thesis based on recent examples, not a settled industry consensus.
A Friendly Guide to Data Science: Everything You Should Know About the Hottest Field in Tech (Friendly Guides to Technology)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Series Turns To Each Lever
The next step is the rest of The Control Series, which the source says will examine each chokepoint separately. Readers should watch for sourcing on the model-access incident, details of compute leasing terms, and further evidence on whether AI distribution platforms are gaining more leverage than model developers.
Access Control Systems: Security, Identity Management and Trust Models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the actual news development?
The development is the publication of Thorsten Meyer AI’s first Control Series installment, which organizes recent AI infrastructure events into a six-chokepoint framework.
What are the six chokepoints?
The six named chokepoints are power, compute, data, model access, distribution and capital.
Is this breaking news?
No. This is best classified as analysis. It draws on reported events and claims from March to June 2026 rather than announcing a single new transaction or policy decision.
What is confirmed from the provided material?
It is confirmed that the source argues AI control is concentrating across six layers. Specific examples, such as rental terms, model shutdown timing and data licensing structures, are presented as claims from the source material and should be read with that attribution.
Why should readers care?
If AI access depends on a small number of infrastructure owners, companies and users may face higher prices, service limits, changed terms or sudden loss of access. That affects planning for products, data strategy and vendor risk.
Source: Thorsten Meyer AI
