AI Capital Flows Are Draining Liquidity from Crypto Markets, Analyst Says

TLDR:

• Hyperscalers are projected to spend $650–700B on AI infrastructure in 2025, outpacing crypto inflows.
• BTC is down 20% YTD while Micron, Nvidia, and AMD posted triple-digit gains since January 2024.
• Stablecoin supply sits at ATH but remains locked in T-bill wrappers, bypassing DeFi yield markets.
• Only AI crypto tokens with real on-chain usage are surviving; most AI microcaps are roundtripping fast.

AI investment is pulling capital away from crypto markets as hyperscaler spending accelerates in 2025. Risk capital appears to be finding a cleaner narrative in AI infrastructure, where real revenue and demand are measurable.

Meanwhile, crypto assets like ETH and BTC continue to underperform. Stablecoin supply sits at all-time highs, yet that liquidity remains parked in tokenized T-bills rather than flowing into DeFi protocols, compressing yields to near Treasury-bill levels.

AI Infrastructure Spending Outpaces Crypto Returns

Hyperscalers are on pace to deploy roughly $650–700 billion in AI capital expenditure this year. AI-related spending now accounts for approximately 2.4% of U.S. GDP, reflecting the scale of institutional commitment.

Nvidia recently reported $81.6 billion in quarterly revenue, reinforcing the demand cycle. SK Hynix is running a 72% operating margin driven entirely by high-bandwidth memory demand from the Nvidia supply chain.

The performance gap between AI equities and crypto has widened sharply since January 2024. Micron returned 987%, Dell 289%, AMD 209%, and Nvidia 246% over that period.

The SMH and SOXX semiconductor ETFs returned 223% and 192%, respectively. SK Hynix delivered a 17x return as HBM became a critical bottleneck.

By contrast, BTC is down around 20% year-to-date while the S&P 500 is up roughly 10%, largely carried by AI-related names.

Crypto analyst Fabius DeFi noted on X that this divergence reflects a structural shift in where risk capital is being deployed. “Has risk capital found a cleaner story with real capex, real demand, real revenue acceleration, and a much bigger TAM than DeFi gov coins?” he wrote.

The question reflects a broader concern that crypto’s value narratives have struggled to compete with measurable AI revenue growth. Stablecoin liquidity at ATH is not translating into crypto market support.

A small number of crypto tokens with genuine AI utility did find buyers this cycle. VVV returned 1,100% on the back of actual private inference usage.

NEAR rose 4x on its intent settlement narrative. TAO reached a local top near $370 before community-related turbulence pulled it back. These are exceptions, not the rule.

Separating Real AI Utility From Token Speculation

The market is beginning to separate AI infrastructure tokens with real demand from those mimicking the Nvidia narrative without revenue.

Compute markets, decentralized GPU capacity, inference rails, agent payment layers, and data provenance protocols represent genuine use cases.

Each requires a token or settlement layer to function, giving them a structural reason to exist. That distinction matters when evaluating which projects survive.

Tokens without a clear revenue path or usage forcing mechanism are cycling through the same pattern as earlier speculative waves. WLD, RENDER, and FET are currently underwater and struggling to establish a next leg.

Most AI microcaps on Base and Solana tend to roundtrip quickly, offering little beyond speculative momentum. Fabius DeFi cautioned that rotating from weak crypto assets into random AI tokens is not a sound strategy at this stage.

The broader risk scenario involves hyperscaler capex cuts if AI return on investment disappoints. A slowdown in enterprise AI adoption could reprice the entire AI investment food chain rapidly.

That repricing would flow through to crypto AI tokens as well, particularly those without on-chain revenue or usage metrics. Macro sensitivity remains a key variable for both TradFi AI equities and crypto AI assets.

The conclusion Fabius DeFi draws is straightforward: TradFi AI is real because its revenue is real. Crypto AI is only real where token or rail usage is forced by the product itself. Everything else is speculation wearing a timely narrative.

The post AI Capital Flows Are Draining Liquidity from Crypto Markets, Analyst Says appeared first on Blockonomi.