China-Linked AI Proxy Payments Spike 55% After US Launches

Payment activity from suspected China-linked crypto wallets ran 55% above normal in the first three days after major US AI model releases, according to a chart in a Chainalysis report on the “proxy services” that resell access to banned American models. The figure spread on X on Saturday after Cointelegraph posted it.

A proxy service sits between the user and a model like Claude or ChatGPT, hiding where the request comes from. Chainalysis surveyed 100 of them and found 40% accept crypto, mostly stablecoins, digital tokens pegged to the dollar. China is among the places these labs bar from direct access.

Business Pill · Business Engineer

An evergreen Business Engineer explainer. It teaches a general idea only and does not discuss this report.

The key insight: Chainalysis found China-linked crypto payments to AI proxy services run 55% above normal in the three days after a US frontier model release, and dip 21% below around day 13, when a Chinese open-weight model typically ships.

What the Chart Shows

Chainalysis measured how many days were unusually busy for payments from China-linked senders, compared with normal, in three-day blocks after each frontier model release. The first block, days 0 to 2, was 55% above normal. Days 3 to 5 were 9% above, and days 6 to 8 were 8% above.

Then the line turns. Days 12 to 14 were 21% below normal, the only deep dip on the chart. Chainalysis marks that window as when a Chinese open-weight model is typically released, around day 13. Days 18 to 20 were back to 10% above.

By our arithmetic, the swing from the first block to the dip is 76 percentage points, and the first three days ran 5.5 times hotter than any later block.

Bar chart: of 100 AI proxy services surveyed by Chainalysis, 40 accept crypto, mostly stablecoins, and 60 do not
Chainalysis surveyed 100 AI proxy services; 40 accept crypto, mostly stablecoins.
China-linked payment activity to AI proxy services runs 55% above normal in the first three days after a US fr
China-linked payment activity to AI proxy services runs 55% above normal in the first three days after a US frontier model release and dips 21% below around day 13 (Chainalysis).

Why It Matters for Export Controls

Chainalysis links the pattern to distillation, where a lab trains its own model on a stronger model’s answers. It cites Anthropic’s accusation that DeepSeek, Moonshot and MiniMax extracted capabilities from Claude through 24,000 fraudulent accounts run through proxy services.

The report adds to a week of evidence about AI capability crossing borders. The new US chip rule, announced by the Commerce Department on 16 November, now requires licences for data centers abroad that rent large amounts of computing to companies based in China, not only for shipping chips.

Epoch AI found China recorded $3.75 billion of server imports from Malaysia between April 2024 and June 2025, while Malaysia recorded $0.6 billion of exports to China. And OpenAI said on 20 November that attackers’ success at pulling reasoning out of its models fell from 54% to 0.2%.

Most proxy services do not take crypto directly, Chainalysis found. They route it through custodial payment processors, which Chainalysis says adds distance between users and providers, and means funds linked to proxies can reach regulated crypto platforms with no clear direct ties back to them.

The Structural Read

As we read it, the timing is the story: the busiest days come right after a US launch, which fits buyers rushing to new capabilities, though Chainalysis says the cause is not proven.

As we read it, the payment trail is a new pressure point for export controls, because blocking chips does not block someone renting answers through a proxy.

Chainalysis, in its 10 Nov 2026 report

“The aggregate data cannot, however, confirm what is driving the observed activity.”

Three Implications

FOR AI LABS proxy services resell access to restricted models, often at a discount (Chainalysis).

FOR CRYPTO FIRMS proxy payments often pass through custodial processors before reaching regulated platforms (Chainalysis).

FOR READERS the report does not name the model releases, wallet counts or dollar amounts (Chainalysis).

What Is Not Established

Chainalysis says plainly that “the aggregate data cannot, however, confirm what is driving the observed activity.” The report does not name the model releases it studied, the number of wallets, or any dollar amounts. The chart measures busy payment days, not money spent. The report is dated 10 November; it drew wide attention on 21 November.

This Week In The AI Supercycle

A World Without an Order, and an AI Central Bank Without a Successor

This week’s issue: the week AI memory stayed largely sold out while the people holding the shares paid: why the AI Supercycle is not a dot-com replay, read through geopolitics, macro and technology, with the gap between building and financing the widest we’ve recorded.

Read the full issue →

Business Engineer Framework

The AI Supercycle

The Business Engineer map of who builds, finances and captures value in AI.

Read the Map of AI →

The Bottom Line

Chainalysis found China-linked crypto payments to AI proxy services run 55% above normal in the three days after a US frontier model release, and dip 21% below around the time a Chinese open-weight model typically ships. Chainalysis says the data cannot prove distillation is the cause.

97,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. Every figure is from Chainalysis’s own report, which we read in full, including its chart of payment activity by three-day period. The 76-point swing and the 5.5 times comparison are our arithmetic on that chart. Cointelegraph’s post is how the figure spread, not its source.

Sources: Chainalysis, ‘The Shadow Market Behind China’s AI Model Distillation Boom’, 10 Nov 2026 · Cointelegraph on X, 21 Nov 2026

Scroll to Top
FourWeekMBA