Arthur Hayes Explains Why Bitcoin Is Sold First When Markets Turn Risk-Off
Key Takeaways
- •Arthur Hayes said Bitcoin and crypto are sold first during risk-off events because their 24/7 trading makes them the easiest assets to liquidate for meeting margin calls or repaying loans.
- •Hayes identified the Anthropic IPO as a key market test, with a strong debut potentially restoring AI sector confidence and a weak one raising doubts about valuations, compute commitments, and sustainable profits.
- •He warned that a reversal in AI capital expenditure could trigger a severe global downturn, which he expects governments and central banks to counter with bailouts and money creation that could ultimately benefit scarce assets like crypto.
- •Hayes described FLOP, a proposed decentralized compute network where providers earn tokens for performing inference, arguing that compute-linked tokens may suit AI agents better than stablecoins, which can be frozen.
- •Near term, Hayes advised preserving cash and avoiding excessive leverage, while favoring established, liquid assets with proven product-market fit over speculative micro-cap tokens.

Arthur Hayes, Chief Investment Officer at Maelstrom, said Bitcoin and the broader cryptocurrency market continue to behave as risk assets that investors liquidate first whenever global markets turn risk-off — phases in which investors pull back from riskier holdings. He made the remarks on the New Era Finance podcast on November 17, 2026.
Because crypto trades around the clock, Hayes argued, investors facing losses in other positions frequently sell it to meet margin calls — broker or lender demands for cash or additional collateral when positions fall in value — or repay bank loans. The dynamic leaves Bitcoin exposed to heavy selling, particularly when stress is concentrated in technology stocks.
Why Bitcoin Is Sold First During Risk-Off Events
Hayes explained that Bitcoin's continuous liquidity makes it the easiest asset to sell. Because crypto markets never close while traditional exchanges operate on set hours, traders who need cash quickly tend to turn to crypto before touching other holdings, while positions in less liquid markets cannot be closed nearly as fast.
Even when investors expect bailouts to lift prices later, they still need cash in the present. Selling crypto helps them avoid liquidation elsewhere, and as a result Bitcoin tends to track the Nasdaq closely.
WuBlockchain noted that Hayes has a mixed track record and often revises his views. Readers should therefore treat his analytical framework as the focus, rather than specific price targets or timelines.
The discussion also covered the risks surrounding the AI investment boom. In his framing, the AI cycle is not a separate storyline but the force he expects to shape the liquidity conditions crypto ultimately trades on. Hayes described the Anthropic IPO as a key market test: a strong debut could restore confidence in AI and encourage renewed risk-taking, while a weak one could raise doubts about sector valuations, large compute commitments, and whether AI firms can generate sustainable profits.
Hayes warned that a reversal in AI capital expenditure — the large-scale spending on computing infrastructure behind the boom — could trigger a severe global downturn. He expects governments and central banks to respond with bailouts and money creation rather than austerity.
Compute, Tokens, and Crypto Strategy for AI Agents
Crypto, as a scarce asset, could then benefit from the renewed liquidity, according to Hayes. Until conditions improve, he advises preserving cash, avoiding excessive leverage, and preparing for volatility. He added that DeFi and other crypto sectors may still fall during a broad sell-off.
Hayes also argued that AI compute may ultimately become a commodity. In his view, the largest long-term winners could be companies built on top of the infrastructure, and he expects AI agents and machines to require open, transparent, and censorship-resistant financial systems.
To pursue that idea, Hayes described FLOP, a proposed decentralized compute network. Under the model, providers contribute computing power and earn tokens for performing inference — running trained AI models to produce outputs — which agents can then spend to obtain compute. He argued that a compute-linked token may suit agents better than stablecoins, which can be frozen and remain tied to traditional financial systems.
FLOP's proposed growth strategy combines testnet incentives — rewards for exercising the network during its pre-launch testing phase — mining rewards, low compute prices, and rising demand from AI agents.
On token design, Hayes criticized projects that give users little value while concentrating ownership among insiders, citing Hyperliquid's user-aligned model as a better example. For investing, he said he favors established, liquid assets and projects with proven product-market fit over speculative micro-cap tokens. Taken together, the remarks cast crypto as a liquidity-sensitive risk asset in the short run and, in Hayes's view, potential financial infrastructure for AI agents over the long run.