AI’s capex supercycle is siphoning risk capital from bitcoin — but profits could rotate back, says Relai’s CEO

Bitcoin’s recent softness isn’t happening in a vacuum. Julian Liniger, CEO of Relai, contends the current bear phase is partly a byproduct of capital migrating into the AI buildout. That framing fits what many allocators are doing right now: crowding into the highest-velocity trade in the market and funding an unprecedented spend on chips, data centers, and power. If that is the binding constraint, the question is not whether bitcoin demand exists, but when the liquidity pendulum swings back.

The single lens worth using here is risk-budget competition. Capital rarely “leaves markets”; it rotates. Today’s marginal dollar often chases AI beta—NVIDIA and the data-center stack—because the feedback loop is immediate: order books, revenues, guidance. Meanwhile, BTC’s reflexivity is muted: ETF inflows have slowed, miner supply is steady post-halving, and macro liquidity isn’t doing bitcoin any favors. In that context, the AI capex supercycle acts like a sink—soaking up equity flows, private credit, and corporate cash that might otherwise feed crypto risk.

How the AI drain shows up in bitcoin demand - Public equity gravity: Momentum allocators frequently treat BTC and AI as substitutable risk exposures. When AI names print outsized gains, incremental flows lean that way, suppressing bitcoin’s bid even if its long-term thesis is intact. - Primary market absorption: The financing of chips, data centers, and power projects taps equity, debt, and retained earnings. That capital doesn’t directly sell BTC, but it reduces the free cash available for new crypto allocations across funds, corporates, and high-net-worth investors. - Attention liquidity: Narrative bandwidth is finite. As AI dominates headlines and P&Ls, discretionary traders often underweight BTC until price leadership changes. Less attention generally means thinner marginal demand.

Liniger also suggests the flow can reverse if AI gains compound. That makes sense. Wealth effects drive allocation decisions. When a trade mints substantial, realized profits, sophisticated investors often recycle a slice into scarce, uncorrelated assets to diversify and protect gains. BTC tends to be a beneficiary of exactly that behavior when its relative momentum turns and its macro hedge value is back in focus.

What could flip the rotation back toward BTC - Profit recycling: As AI winners realize gains via buybacks, dividends, or secondary sales, a portion of those proceeds can move into BTC, especially if bitcoin’s trend inflects and ETF pipes are ready to scale new demand. - Marginal return compression: If AI capex meets power, land, and supply-chain bottlenecks, returns may normalize. When the “certain growth” premium narrows, allocators frequently re-risk into convex assets like BTC. - Rails for AI-native commerce: Should autonomous agents and inference marketplaces need open, instant settlement, bitcoin’s Lightning and other crypto rails can capture transactional demand. It won’t dominate flows overnight, but utility narratives tend to reinforce price when they coincide with capital rotation. - Policy and optics: Energy constraints will keep colliding with AI and mining. If regulators stabilize frameworks for both, it reduces headline risk that currently skews allocator behavior toward AI incumbents and away from BTC.

What to monitor - Relative strength: BTC vs. an AI basket (chips, servers, power). When that ratio turns, flows usually follow. - ETF net flows and basis: A re-acceleration in spot ETF inflows alongside tightening futures basis is a clean tell that demand is returning. - Corporate actions in AI: Equity issuance, convertibles, and capex guidance signal where the next dollar is going—and when profits might be harvested.

My read: Liniger’s liquidity-absorption thesis maps cleanly to how money actually moves. The AI supercycle is taxing bitcoin demand at the margin right now, but it also plants the seeds for a reversal if profits get banked and relative performance shifts. Treat it as a timing problem, not a thesis problem. When the flow turns, it often turns faster than people expect.