The numbers are staggering. Goldman Sachs projects AI-related annualized spending could top $800 billion by the end of 2026. Morgan Stanley’s forecast is even more aggressive: nearly $3 trillion in AI infrastructure investment by 2028, with over 80% yet to be deployed. But here’s the thing that keeps me up at night—those numbers are starting to feel like a phantom limb. The market still feels the pain of a missing limb, but the limb itself is starting to ache.
We’ve been here before. Not in crypto, but in the early 2000s dot-com bubble. The difference? This time, the capital expenditure is front-loaded, and the revenue is back-loaded, like a bad trade where you’re paying for the premium but the option is still out of the money. And the market is starting to smell it.
Context: The Index’s AI Addiction
The S&P 500 has become a one-trick pony. JPMorgan reports that the top 20 stocks now account for roughly 50.8% of the index’s total market cap—a concentration level with “no modern precedent.” This isn’t just a statistic; it’s a structural vulnerability. The index’s fate is now intertwined with the AI trade. If AI spending slows, the index doesn’t just dip—it cracks.
BofA’s July fund manager survey confirms this shift. 45% of respondents now identify “AI Bubble” as the biggest tail risk, up from 28% the previous month. For context, that’s higher than the risk of a second inflation spike. The market is officially scared.
But here’s the paradox: Goldman Sachs notes that roughly 64% of S&P 500 companies are beating earnings by at least one standard deviation. The “present” is strong. The “forward” is the problem.
Core Analysis: The Signal in the Noise
Let’s break down the math. The $800 billion to $3 trillion in spending is projected over the next 2-4 years. But the question isn’t whether Big Tech can afford it—BlackRock argues they can, pointing to real profits and strong balance sheets. The question is whether the return on that capital will ever justify the outlay.
I’ve seen this pattern before in my own trading. In 2020, during DeFi Summer, I identified an arbitrage opportunity across three DEXs that generated a 400% return in six weeks. But the volatility nearly liquidated the fund twice. The lesson: high yield equals high fragility. The same principle applies here. The yield on AI capex is uncertain, and the fragility is systemic.
Mac10’s insight is the most damning: companies are pouring unprecedented cash into AI as a “one-time event” flowing through the income statement. This inflates forward earnings growth, making it look sustainable when it’s actually a one-off. It’s like a trader booking a massive gain on a single trade and then spending the next quarter convincing themselves they can replicate it.
The Contrarian Angle: The Smart Money Trap
The conventional narrative is that “institutions are smart, they know what they’re doing.” But the Aschenbrenner fund case tells a different story. The fund, run by a former OpenAI researcher, grew to $45 billion before collapsing to roughly $10 billion after a sharp drawdown in AI infrastructure stocks. Citadel eventually took over.
This is the classic “smart money” trap. The insider, with the best information advantage, still got crushed by leverage and concentration. The fund’s post-collapse move—investing $400 million in an unnamed private company while retaining private equity stakes—smells like a desperate attempt to salvage a narrative.
And here’s my contrarian take: the real danger isn’t that AI spending is slowing. It’s that the slowing is being misinterpreted as a temporary dip rather than a structural shift. If the marginal return on every additional GPU starts to decline, the capex cycle will turn faster than anyone expects. The first sign will be a single hyperscaler cutting its guidance.
Takeaway: The Yield Was Real, the Trust Was Phantom
I’ll leave you with this. The AI infrastructure boom is not a fraud. It’s a real, multi-trillion-dollar buildout. But the market has priced in a linear continuation of exponential growth, and that’s where the disconnect lies. The question isn’t whether AI will change the world—it will. The question is whether the current capital expenditure cycle will generate returns that justify the current valuations.
If the answer is “no,” we’re looking at a correction that makes the 2022 crypto winter look like a warm breeze.
We traded sleep for alpha, and alpha for scars. The algorithm doesn’t lie, but the narrative does. And right now, the narrative is telling us that the yield was real, but the trust was phantom.
I didn’t write this to scare you. I wrote it because I’ve seen this movie before.