TL;DR
Thorsten Meyer AI’s Control Series frames 2026 as the year AI stopped looking like a neutral utility and began operating through six control points. The report cites recent examples involving power, compute, combat data, model access, distribution and capital, while some figures remain based on reported sourcing rather than public contracts.
Thorsten Meyer AI published the first part of its Control Series, arguing that recent 2026 developments show frontier artificial intelligence is no longer best understood as a neutral utility but as infrastructure controlled through six chokepoints: power, compute, data, model access, distribution and capital.
The piece says the shift became visible during a short period in 2026, citing three examples: a government-backed shutdown of a frontier model on roughly 90 minutes’ notice, Ukraine’s use of battlefield data as a licensed AI asset, and major AI companies renting large-scale compute from direct rivals under restrictive terms.
Thorsten Meyer AI identifies power as the base constraint, saying frontier systems increasingly depend on gigawatt-scale energy access. The report points to SpaceX’s Memphis complex as an example of a builder using on-site gas generation to move faster than traditional utility interconnection timelines.
The article also describes compute as highly concentrated. It cites xAI’s Colossus cluster, estimated at about 555,000 GPUs, and says Anthropic and Google have agreed to large monthly payments for Colossus output, while OpenAI relies on Stargate and Oracle. Those figures are presented as sourced claims rather than publicly verified contract terms.
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.
Who Controls AI Access
The argument matters because businesses, developers and governments increasingly treat AI systems as dependable infrastructure. If access can be throttled, repriced, redirected or shut down, customers face a different risk profile than they would with a neutral service available on stable terms.
The report’s central claim is that control is moving to owners of scarce inputs rather than to model builders alone. Power permits, GPU clusters, unique datasets, app distribution and financing capacity can shape who builds competitive models and who can use them.
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.
Six Layers Of Control
The six chokepoints named in the report are power, compute, data, model access, distribution and capital. Each layer is described as a point where a small group of actors can limit, price or redirect AI capability.
On data, the report cites Ukraine’s Avengers Labs, which turns annotated combat footage into a training resource for domestic and foreign companies while requiring that Ukraine retain the improved model. On distribution, it points to the value placed on AI interfaces such as Cursor, saying the interface can matter as much as the underlying model. On capital, it describes large intra-industry financing flows and sovereign funding as a source of leverage.
“AI does not flow freely like a utility.”
— Thorsten Meyer AI
8/5 SO Electrical Cable (8/5 SO, 50)
Heavy Duty – Bulk 8/5 electrical cable with an SO jacket
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unverified Contract Details
Several cited figures, including monthly compute payments and cluster size estimates, depend on the report’s referenced sourcing and have not all been confirmed through public contract documents. The exact terms under which cluster owners may reclaim compute capacity are also not fully public.
It is also not yet clear whether these chokepoints will remain as concentrated if new chip supply, energy projects, open models or regulatory action change the economics of frontier AI.
AIPI-Lite AI Robot, Smart Customizable Character with Voice Cloning, GPT Model Access, Memory Storage, and Real-Time Interactive Reactions
Fully Customizable AI Companion: Define unique roles and personalities with simple text—no coding needed. Shape your AI’s behavior,…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Series Turns To Each Layer
Thorsten Meyer AI says later installments will examine each chokepoint separately. The next test for readers will be whether fresh deals, shutdowns, licensing terms or financing rounds support the report’s claim that AI’s main power centers sit outside the model layer alone.
Hewlett Packard Enterprise ProLiant Compute DL360 Gen12 w/one Intel Xeon 6530P Processor, 1P 2x32GB-R 8SFF NS204i-u v2 MR408i-o 2x1000W PS (HPE Smart Choice P89997-005)
HPE SMART CHOICE MODEL – P89997‑005 – ENTERPRISE 1U RACK SERVER Preconfigured and factory‑tested, this Smart Choice DL360…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the main claim of the Control Series?
The series argues that frontier AI is controlled through six chokepoints rather than operating like a neutral utility available on equal terms.
What are the six chokepoints?
The report names power, compute, data, model access, distribution and capital as the six layers where control can be exercised.
What is confirmed in the report?
The article confirms the publication of the Control Series and its thesis. It also reports cited examples from 2026, while some contract values and terms remain based on sourced reporting rather than public filings.
Why does this matter for AI users?
If access to models depends on scarce infrastructure and revocable agreements, companies using AI may face service, cost and dependency risks that are not visible when AI is treated as a simple utility.
What happens next in the series?
Thorsten Meyer AI says future installments will examine each chokepoint in more detail, starting from the index laid out in Part 1.
Source: Thorsten Meyer AI
