TL;DR
AI Dispatch’s first Control Series installment says recent 2026 events show frontier AI is governed through six chokepoints rather than delivered like a neutral utility. Confirmed developments include Anthropic disabling Fable 5 and Mythos 5 after a US directive, major SpaceX compute leases and a reported $60 billion Cursor deal; some data and contract details remain incomplete.
AI Dispatch published the first installment of its Control Series, mapping six chokepoints – power, compute, data, model access, distribution and capital – that it says now determine who can build, deploy and keep using frontier AI.
The report’s central claim is that AI stopped behaving like a broadly available utility in 2026 and began looking like a set of revocable dependencies. Its evidence starts with model access: Anthropic said in a June 12 statement that a US government export-control directive required it to disable Fable 5 and Mythos 5 for all customers. Axios reported Anthropic was given about 90 minutes after Amazon raised concerns over a possible jailbreak.
Compute is the next test. Business Insider reported that Anthropic agreed to pay SpaceX $1.25 billion a month for Colossus capacity through May 2029, with termination rights on 90 days’ notice. Barron’s reported Alphabet had a separate $920 million-a-month SpaceX compute deal. Together, the deals support the Control Series argument that access to large GPU clusters is now a point of market power.
Confirmed facts in the record include Anthropic’s model disablement, the reported 90-minute window, the SpaceX compute payments and the reported $60 billion Cursor deal. The broader conclusion – that these events amount to a new AI control structure – is the series’ analysis. Details of some data-licensing arrangements and private contract clauses remain incomplete.
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 Shifts Toward Gatekeepers
The practical risk is concentration. A company building products on AI may have model contracts, GPU access, electrical service, data rights, app distribution and financing all controlled by different parties. The 2026 examples show several of those parties can change terms fast, whether through a legal order, an energy permit, a lease clause or a platform acquisition.
For customers, service reliability is no longer only a software uptime question. For policymakers, the focus moves from model behavior alone to ownership of upstream infrastructure. For workers and developers, the tool in the interface may be less powerful than the company that controls the interface, compute or data behind it.
Large-Scale AI Engineering: Design, Train, and Optimize Foundation Models on NVIDIA GPU Clusters
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Utility Pitch Meets 2026 Limits
The utility framing helped sell AI as broadly available, stable infrastructure. The Control Series argues that framing no longer fits the market after a run of events in spring 2026: a sudden US order affecting Anthropic models, billion-dollar monthly compute deals involving SpaceX facilities, fights over gas turbines around Colossus sites, a large Cursor acquisition, and reports of defense data being treated as a sovereign asset.
The series identifies six leverage points. Power sets the physical ceiling; compute determines who can train or serve large models; rare datasets create bargaining power; model access can be limited by labs or governments; distribution gives app and platform owners control over users; and capital decides who can finance the stack long enough to stay in the race.
“We are complying with the government’s legal directive”
— Anthropic, June 12 statement
CASB Essentials: Protect SaaS With CASB | CASB For Cloud Control | SaaS Risk Management | Monitor Access & Use | Simplify Compliance | Ethical AI practices for responsible tech adoption
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Gaps In Contracts And Oversight
It is not yet clear how much of the Anthropic action becomes precedent, whether US officials will disclose the full technical basis for intervention, or how quickly a formal process will replace case-by-case decisions. The full text of SpaceX’s compute contracts and any owner right to reclaim capacity are not public in the article record. Details around Ukraine’s Avengers Labs licensing, including the scope of foreign access and ownership of resulting models, also need primary documentation.
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.
Formal Rules Face Pressure
The next milestones are follow-on Control Series installments, any new filings tied to the Anthropic directive, added SpaceX compute disclosures, closing steps for the Cursor deal, power-permitting decisions in Tennessee and Mississippi, and any Ukraine Ministry of Defense documents that clarify defense-data licensing terms.
Platform Engineering for Artificial Intelligence: Designing scalable infrastructure, data pipelines, and model lifecycle management for generative AI and agentic protocols (English Edition)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the actual development here?
AI Dispatch released an analysis tying recent 2026 AI events into one thesis: control over AI is concentrating at six points in the stack.
What are the six chokepoints?
- Power
- Compute
- Data
- Model access
- Distribution
- Capital
Was the Anthropic model shutdown confirmed?
Yes. Anthropic said on June 12 that a US directive required it to disable Fable 5 and Mythos 5 for all customers. Axios separately reported the company was given about 90 minutes.
Does this prove AI is no longer a utility?
That is the Control Series’ interpretation. The confirmed events show that access to models, compute and distribution can be restricted or repriced quickly; whether that fully replaces the utility model is still a live debate.
Who holds the most leverage?
No single actor holds all of it. Governments and labs control model access, cluster owners and Nvidia shape compute, data owners control rare inputs, platform owners control users, and large balance sheets decide who can keep paying.
Source: Thorsten Meyer AI
