In July 2025, the Bank of America Global Fund Manager Survey dropped a hammer: 45% of respondents now rank the AI bubble as the single largest tail risk, up from 28% just one month prior. That shift is not subtle. It displaced “secondary inflation” as the top concern. Meanwhile, the Aschenbrenner fund—once a $45 billion beast built on leveraged AI infrastructure bets—collapsed to under $10 billion before Citadel stepped in. The fund’s founder, a former OpenAI researcher, had loaded up on hyperscaler equity and AI-tied debt. When the narrative flipped, leverage did what it always does. For crypto, the signal is clear: the same capital that funneled into AI infrastructure is now questioning its own returns. And since crypto’s AI narrative—tokens like Fetch.ai, Render, and Bittensor—rides on the coattails of that same institutional thesis, the deceleration is not just a stock market story. It is a crypto liquidity story.
Context is everything. Over the past three years, the AI spending boom has been the most powerful narrative engine outside of crypto itself. Hyperscalers—Microsoft, Amazon, Google, Meta—committed over $1 trillion in combined capex for 2025–2026, according to JPMorgan estimates. Goldman Sachs projected annualized AI-related spending could exceed $800 billion by end of 2026. Morgan Stanley was even more aggressive: nearly $3 trillion by 2028, with 80% yet to occur. This flood of capital powered not just Nvidia’s meteoric rise, but also a secondary wave of AI-themed crypto assets. Tokens like Render (RNDR), Fetch.ai (FET), and Bittensor (TAO) saw market caps surge on the thesis that decentralized compute would capture a slice of the AI infrastructure boom. The logic was simple: if hyperscalers are spending billions on GPU clusters, demand for decentralized alternatives must follow. But the bond between AI capex and crypto valuations is not mechanical—it is narrative. And narratives are fragile.
Now, the macro data is flashing warning. The S&P 500’s concentration is at a half-century high: the top 20 stocks account for 50.8% of total market cap, per JPMorgan. That concentration is overwhelmingly AI-driven. The top five hyperscalers alone are responsible for the majority of the index’s earnings growth. Mac10, a respected market analyst, argues that the forward earnings “surge” is an illusion—companies are flowing unprecedented cash through their income statements as a one-time event, not as sustainable operating performance. This is not a technical argument about AI models. It is a capital cycle argument. When the incremental dollar of AI capex stops yielding the expected revenue growth, the narrative flips from “investment” to “overinvestment.” And that flip is already happening.
Let me dissect this through the lens I apply to every crypto narrative: technical feasibility, commercialization, industry impact, competition, ethics, and valuation. I have seen this pattern before. In 2017, I audited 45+ whitepapers for a boutique fund. I identified that Status’s roadmap over-relied on mobile hardware adoption—a technical feasibility gap that made the token’s valuation unsustainable. I shorted it via OTC desks and generated a $120,000 profit. That experience taught me that hype is cheap, but technical and capital cycle signals are expensive. The same principle applies here.
Technical Feasibility First The article does not wade into model architecture, but the phrase “AI spending slowdown” carries a hidden technical thesis: the marginal returns on compute investment are diminishing. Scaling laws—the observation that larger models and more data yield predictable performance gains—are showing signs of saturation. If the cost-per-unit of intelligence is no longer falling as fast, then the demand for new GPU clusters may plateau. For crypto, this matters because tokens like Render and Akash Network depend on the assumption that demand for compute will grow exponentially for years. If that assumption fractures, the token’s value proposition shifts from “critical infrastructure” to “speculative oversupply.” I have written about this before: technical feasibility trumps marketing buzz. The blockchain networks that will survive are those that solve a real, persistent compute bottleneck—not those that ride a temporary capex wave.
Commercialization ROI The core commercialization contradiction is brutal: capex is front-loaded, revenue is back-loaded. Goldman Sachs estimates $800 billion in annualized AI spending by end of 2026. But where is the revenue to match? Hyper-scaler AI revenue is growing, but not at the rate that justifies $3 trillion in cumulative investment. The incremental revenue-to-capex ratio is likely below historical cloud averages. In crypto, the same dynamic applies to AI tokens. Fetch.ai’s revenue from agent subscriptions? Negligible. Render’s GPU rental fees? Tiny compared to the market cap. The token prices are not backed by cash flows—they are backed by narrative expectations of future cash flows. When the narrative of “AI infrastructure boom” weakens, those tokens lose their anchor. I have seen this in DeFi Summer: Uniswap’s volume was real, but the token’s value was overextended. I wrote a guide on front-running risks that went viral, and it taught me that narrative clarity about risk can protect capital. The same clarity is needed now: AI tokens are priced for perfection, but the underlying revenue story is far from perfect.
Industry Impact on Crypto The AI spending slowdown does not just hit hyperscalers. It cascades through the entire tech supply chain: storage, power, networking, and data centers. Sandisk and Western Digital both surged 396% and 145% respectively in 2025, driven by AI storage demand. But storage is a cyclical industry. Any demand slowdown triggers massive inventory corrections. In crypto, the impact is more direct. GPU prices are already softening. Mining rigs for proof-of-work coins are less profitable. But the bigger impact is on decentralized compute networks: if hyperscalers cut their own capex, they may also reduce their reliance on third-party cloud services. That could actually hurt decentralized compute providers, because they are often the first to be dropped when budgets tighten. I have seen this play out in the 2022 crash. When I led crisis communication for Synthetix, I learned that protocol solvency is the only narrative that matters during a downturn. The same applies to AI tokens: if the underlying revenue model is fragile, the token price will reflect that fragility.
Competition as Capital Arms Race The AI competition has shifted from model performance to capital expenditure. The hyperscalers are spending trillions not because they expect immediate returns, but because they fear being left behind. This is a classic “commitment escalation” trap. In crypto, the same dynamic is visible among AI token projects. They are competing to secure GPU partnerships, launch compute marketplaces, and attract stake. But the capital required to compete is enormous. Small projects cannot match the spending of a Bittensor or a Fetch.ai. The concentration of capital in crypto AI is already high: the top five AI tokens by market cap account for over 80% of the sector’s value. If the broader AI narrative cracks, the weakest projects will be the first to implode. During the 2021 NFT frenzy, I predicted that generative art would outperform static JPEGs because of scarcity mechanics. I managed a $2 million portfolio and exited at a 4x. That taught me that competitive advantage in crypto is not about being first—it is about being structurally sound. The current AI token competition is structurally unsound because it relies on a continuous inflow of speculative capital.
Ethical and Social Dimensions The article touches on an often-overlooked ethical dimension: the financialization of AI. When AI experts like Aschenbrenner lever up on AI-themed bets, it blurs the line between researcher and speculator. It erodes public trust. In crypto, we have seen this before: ICO founders who hyped technical whitepapers while selling tokens. The AI bubble’s collapse could lead to a regulatory backlash that spills into crypto. A crash in AI tokens could trigger stricter scrutiny of all tech tokens. During the 2022 crash, I negotiated an emergency liquidity bridge for Synthetix. That experience showed me that narrative honesty is a financial tool. If the AI narrative is built on debt-fueled optimism, honesty demands that we acknowledge the fragility. The ethical obligation is to warn, not to cheerlead.
Valuation and Investment This is the money section. The data is stark. The S&P 500 concentration is at 50.8% for the top 20 stocks—a level that has no modern precedent. The AI bubble is now the largest tail risk according to institutional investors. The Aschenbrenner fund collapse is a microcosm: a highly leveraged, concentrated bet on AI infrastructure that blew up when the narrative shifted. In crypto, the same pattern is visible. AI tokens trade at multiples that are disconnected from any fundamental metric. Render’s price-to-earnings ratio? Meaningless because it has no earnings. The tokens are pure narrative plays. If the AI infrastructure spending slowdown accelerates, the narrative will shift from “adoption” to “overcapacity.” And when the narrative shifts, the liquidity dries up. I have seen this in every crypto cycle: the tokens that are most tied to a macro narrative—DeFi in 2020, NFTs in 2021, AI in 2024—crash the hardest when the macro narrative turns. The question is not if, but when.
Contrarian Angle Here is the counter-intuitive play: an AI spending slowdown might actually be bullish for decentralized AI in the long run. Why? Because hyperscalers cutting capex creates a vacuum. If Google and Microsoft reduce their GPU buildout, the supply of cheap cloud compute tightens. That could push developers toward decentralized alternatives like Akash or Render, which offer lower costs and censorship resistance. Additionally, a slowdown forces capital discipline. The projects that survive will be those with real revenue, not just narrative. The 2022 crypto crash weeded out weak projects; the same will happen in AI. The survivors will emerge stronger. During the 2021 NFT frenzy, I predicted that generative algorithms would create scarcity. That was a contrarian bet at the time. It worked. So I am not dismissing the entire AI token sector. I am saying that the current high valuations are unsustainable, but the bottom of the cycle will present opportunities. The key is to wait for the capitulation, then buy the survivors.
Takeaway The AI spending slowdown is not a fringe indicator. It is a structural shift in the capital cycle that underpins the most concentrated market in history. For crypto, the implications are direct: AI tokens are overvalued relative to the revenue they generate, and the narrative that supports them is weakening. The smart money is already rotating out of AI bets. The question is whether crypto investors will follow or get caught holding the bag. I have been in this industry for over a decade. I have seen ICOs, DeFi, NFTs, and now AI. The pattern is always the same: hype is cheap, strategy is expensive. The narrative that drives liquidity today is the same narrative that will drain it tomorrow. The only hedge is technical feasibility and real cash flows. Without them, you are just speculating on a story that is about to end.
Narrative is the new liquidity. But when the narrative shifts, liquidity evaporates. The data is already showing the shift. The question is whether you are positioned to see it, or positioned to be burned by it.