AI Stocks Down 60% in a Month, Baker: Fundamentals Still Accelerating
N.R. Finch
AI-linked stocks have pulled back 60% in a month, yet tech investor Gavin Baker argues core metrics — inference demand, GPU supply-demand, token usage — are all accelerating. The gap between price and fundamentals sets up next earnings season as the decisive test.
Down 60% — what actually happened?
AI-related stocks dropped 60% over the past month as sentiment turned sharply bearish.
Baker's read: this is a price-vs.-fundamentals divergence, not a fundamental breakdown.
This means → the market is front-running bad news that operating data has not confirmed.
What evidence does he cite for "acceleration"?
Inference demand keeps climbing across OpenAI, Anthropic, and open-source models — downstream users are genuinely adopting AI, not just experimenting.
GPU supply-demand is tightening and DRAM prices are rising — hardware strain typically signals demand outrunning supply.
Token usage growth is speeding up — the most direct volume-and-price indicator: more users, deeper usage.
In plain terms = every meter along the chain — chips, models, end users — is showing higher readings.
Is the market misreading capex?
Baker argues the market misreads Big Tech spending: large capex does not equal cash burn — much of it maps to compute-lease contracts and long-term revenue streams.
Meanwhile, private AI companies' real profitability is underestimated — the market sees listed-company filings but not private players' cash flows.
This means → treat compute assets as infrastructure, not one-off expense, and the return math looks far better.
What comes next?
The core tension is clear: the market prices AI on a short-term drawdown; Baker prices it on accelerating fundamentals.
As long-term contracts enter repricing cycles (renewing at updated rates) and next-gen GPUs come online, compute-asset cash flows should improve.
This reflects a simple question — whether this pullback is froth removal or a mispricing — and next earnings season will provide the answer.
Content is for reference only, not financial advice.