📰 Key Takeaways

J.P. Morgan Asset Management data shows hyperscaler capex has overtaken free cash flow, and valuations look rich relative to the S&P 500. Jordi Visser, a former Morgan Stanley managing director and CIO of Weiss Multi-Strategy Advisers, now runs AI Macro Nexus Research at 22V. He argues that AI agents will let users direct software to carry out trades and other financial actions on their behalf, and for crypto that’s like the App Store or the smartphone. Agent adoption should drive activity in tokenized assets and crypto payments, so the investment opportunity isn’t limited to the companies supplying compute. His own holdings have expanded from bitcoin, ether and Solana to a broader basket of tokens.

He’s getting more cautious on the AI trade itself, though. He started trimming Micron, the memory chipmaker, in mid-May because hedge funds and retail investors had crowded in too far. He thinks bitcoin and ether have struggled partly because the market’s attention is all on AI. On the numbers: Microsoft said on its April 29 earnings call that its AI business hit a $37 billion annualized revenue run rate, more than double a year earlier, and Micron’s fiscal second-quarter revenue nearly tripled to $23.86 billion. But the 100-stock AI basket he tracks is up about 40% this year, and its rolling 126-day volatility is now higher than bitcoin’s. Funding pressure is showing up too: Oracle’s fiscal 2026 capex was $55.7 billion against just $32 billion in operating cash flow, and it added $43 billion in new debt. He stresses that “the AI fireworks show is over.” That doesn’t mean AI is over. He still expects agents to drive multi-year demand for compute, but investors should lower their return expectations. If an AI selloff drags down equities, crypto could retest its recent lows, but that wouldn’t end the bull market.

💬 JudyAI Lab Take

The Microsoft and Micron numbers look great, but J.P. Morgan Asset Management’s data also shows hyperscaler capex outrunning free cash flow, with valuations rich relative to the S&P 500. When a former senior fund manager like Jordi Visser says “the AI fireworks show is over,” it points to the market shifting from euphoria to a cooler second look.

For AI builders, this story shows heat splitting between the infrastructure side and the application side. The original piece notes that Microsoft’s AI business hit a $37 billion annualized run rate and Micron’s fiscal second-quarter revenue nearly tripled, so demand is real. But Oracle’s fiscal 2026 capex was $55.7 billion against only $32 billion in operating cash flow, plus $43 billion in new debt, so the funding strain on compute suppliers is already visible. Visser isn’t bearish on AI. He still thinks agents will drive multi-year compute demand, and he expects agent adoption to let users direct software to execute trades and financial actions, boosting tokenized assets and crypto payments. That suggests value could spread from the companies supplying compute to the application layer where agents are actually used. Our take: when there’s a gap between capex and returns, what survives is products with a clear use case, not scale piled up by chasing the compute boom.

If you’re building with AI, check whether your project leans too heavily on one vendor’s compute and pricing, set your return expectations more conservatively, and prioritize proving that agents can create value reliably in real workflows.

📅 Source Info

🔗 Further Reading