AI Capex Slowdown: The Decoupling Test Crypto Markets Have Been Waiting For

The Aschenbrenner fund imploded from $45 billion to $10 billion in a matter of months. Citadel stepped in, the market yawned. But the 450 → 100 trajectory is not just a hedge fund failure—it’s a microcosm of the entire AI capex narrative. When the smartest insider money in AI leverage-crashes on AI infrastructure longs, you have to ask: what is the market pricing in, and what is it ignoring? Context: The AI capex bubble has become the biggest tail risk in global equity markets, according to 45% of fund managers surveyed by Bank of America in July 2025—up from 28% just a month earlier. JPMorgan’s data shows the top 20 stocks in the S&P 500 now represent 50.8% of total market cap, a concentration without modern precedent. Goldman Sachs projects AI-related annualized spending could exceed $800 billion by end of 2026; Morgan Stanley sees nearly $3 trillion in AI infrastructure investment by 2028, with 80% yet to be deployed. The driver? Five hyperscalers are set to deploy over $1 trillion in 2025-2026 alone. BIS has warned this spending spree could turn into a long-term investment crash. The bulls—BlackRock, for instance—argue AI leaders generate real profits, have strong balance sheets, and are funding most of the capex from internal cash flows. But the market is now pricing in a slowdown, and the debate is whether that slowdown is a buying opportunity or the beginning of a systemic unwind. Core: The AI capex thesis is structurally identical to the DeFi liquidity mining boom of 2020-2022—except the numbers are two orders of magnitude larger. In 2022, I analyzed the recursive yield farming models that caused the collapse of Terra and the cascade through lending protocols. The AI hyperscalers are running a similar recursion: they invest billions in GPUs, rent that compute to AI startups via cloud APIs, those startups burn cash on inference and training, and the revenue circularly flows back to the hyperscalers as “AI revenue.” The problem is that the final end-user revenue—actual paying customers for AI applications—has not materialized at the scale needed to support the $800 billion annualized capex. Think of it as a constant product curve where the supply of compute is being added at a steep slope, but the demand curve is flat. The result is a liquidity funnel that leaks value. My 2020 Python simulation on AMM liquidity fragmentation showed that when supply grows faster than organic demand, the pool becomes a death spiral. The same applies here: hyperscalers are the liquidity providers, and the AI startups are the traders. When the traders stop providing returns, the LPs are left holding the bags. Aschenbrenner’s fund is the canary in the data center. He was a former OpenAI researcher, deeply embedded in the AI narrative. His fund went from $45 billion to $10 billion by betting on AI infrastructure stocks—NVIDIA, hyperscaler suppliers, storage companies like Sandisk (up 396%) and Western Digital (up 145%). The fund’s collapse was triggered by a combination of leverage and a repricing of AI infrastructure names. But the market hasn’t fully repriced yet. The 64% of S&P 500 companies beating earnings by one standard deviation, as Goldman notes, is a lagging indicator of past performance, not a validation of future capex returns. Mac10’s point is crucial: the record-breaking forward earnings growth is inflated by a one-time event—the massive cash being spent on AI, which flows through the income statement as “investment” but generates no recurring revenue. This is exactly the same dynamic I saw in 2022 when DeFi protocols booked token emissions as revenue. The market eventually corrected that accounting arbitrage. Contrarian: The conventional wisdom says that if AI spending slows, the entire tech ecosystem—including crypto—will suffer. The logic is simple: crypto is risk-on, tech is risk-on, and AI is the biggest risk-on theme. If the S&P 500 corrects due to AI capex disappointment, Bitcoin and altcoins will crash alongside. But this view ignores the structural decoupling that has already begun. Crypto markets have been repricing since the 2022 bear market. The current cycle is not driven by AI hype; it’s driven by monetary policy normalization, ETF flows, and the maturation of DeFi. In fact, AI’s capex slowdown could be a net positive for crypto for two reasons. First, capital that was flowing into AI infrastructure stocks will seek alternative high-growth narratives. Crypto offers a playbook that has already been tested through a boom-bust cycle. Second, the AI capex crash validates the thesis of decentralized compute networks. When hyperscalers overbuild and then cut spending, the excess capacity creates a price war. But that price war occurs inside centralized clouds. In contrast, decentralized networks like Akash or Render use token-based incentives to dynamically adjust supply. They are more capital-efficient and less prone to the “commitment escalation” trap that hyperscalers face. I’ve been modeling this since 2024 when I analyzed the latency arbitrage between traditional settlement layers and on-chain liquidity. The lesson is that rigidity creates inefficiency, and inefficiency creates arbitrage opportunities. The AI capex slowdown is an arbitrage event for decentralized compute. Moreover, the Aschenbrenner fund blowup is a classic example of “smart money” being wrong because they were too close to the narrative. In 2022, I wrote an internal memo arguing that the FTX collapse was not a leverage problem but a recursive yield model failure. I was dismissed by senior analysts who wanted a simple story. Today, the AI narrative is the same: everyone wants to believe that the capex will pay off because the technology is transformative. But the technology is not the market. The market is a liquidity pool that prices expectations, not fundamentals. The AI capex slowdown is expected to be a gradual deceleration, not a sudden stop. But as the BIS warned, the spending spree could turn into a “long-term investment crash.” The key question is whether the market will front-run that crash or wait for the data. Given the concentration risk—50.8% of the S&P 500 in 20 stocks—any downward revision in AI capex guidance could trigger a rapid unwind. The crypto market, with its decentralized and fragmented nature, may actually absorb this shock better than the centralized equity market. The liquidity pool is a mirror, not a vault—it reflects the crowd’s fear, but it doesn’t hold the truth. Takeaway: The AI capex slowdown is not a catastrophe for crypto; it’s a decoupling test. The market is asking: can crypto sustain its own narrative when the dominant macro narrative (AI) falters? My answer is yes, but only if the crypto ecosystem focuses on what it does best—providing verifiable, autonomous trust substrates. The AI capex cycle is a reminder that capital allocation is a constant game of musical chairs. When the music stops—and it will—the chairs that remain are those built on code, not on promises. The algorithm optimizes for survival, not for you. The question is not whether AI spending slows, but whether you are positioned in the liquidity pool that actually reflects the underlying economic reality, not the one that mirrors inflated expectations.