随笔

TSMC, Google, and Tesla All Fell After Earnings — AI Capex Anxiety Spreads

2026-07-23

I. Three Reports, Three Different Stories

In mid-to-late July, TSMC (TSM), Google (GOOGL), and Tesla (TSLA) all released Q2 2026 earnings, and all three stocks fell — but look closer and each "fall" has its own cause.

  • TSMC (reported July 16): Revenue of $40.3B (NT$1.27 trillion), up 36% year-over-year, beating expectations; gross margin of 67.7%, also beating the company's own guidance; TSMC even raised its full-year 2026 revenue growth guidance to "above 40%." Even so, the stock fell about 1.55% after-hours. This looks more like a "sell the news" reaction — the good news was already largely priced in, and the company's own raised capex guidance (roughly $18.75–20B) added one more layer of investor unease.
  • Google (reported July 22): Revenue of $119.8B, beating the expected $116.9B; Cloud revenue surged 82% to $24.8B. The results themselves were genuinely strong, but the stock sank during the earnings call — because the company sharply raised its 2026 capex guidance to as much as $205B.
  • Tesla (also reported July 22): Revenue of $28.24B, beating the roughly $26.4B expected and a record high; but non-GAAP EPS came in at just $0.33, well short of the $0.53 expected, down 18% year-over-year; gross margin slid to 16.8%, below both last year's 17.2% and the expected 19.4%; operating income collapsed 57%, with operating margin compressed from 4.1% to 1.4%; capex surged 142% year-over-year to $5.79B, pushing free cash flow negative. The stock fell about 3% after-hours. This is the only one of the three that was a genuine, fundamentals-level miss — Tesla was already down about 17% for the year going into this report.

II. The Common Thread: AI Capex Anxiety

Different as they are, all three reports share one underlying story: the market is questioning whether Big Tech's unprecedented AI capital spending will actually pay off in kind.

This isn't a single-company issue. Through July, the semiconductor sector went through a sharp valuation reset, wiping out more than a trillion dollars in market value. Alphabet, Amazon, Microsoft, and Meta alone are projected to spend a combined ~$710B on capex in 2026. Investors are increasingly asking: will the compute this money buys actually get used, and will it ever earn its cost back?

On July 17, China's Moonshot AI released its new Kimi K3 model, which sharpened this anxiety further — if an open-weight, lower-cost model can get close to frontier-level performance, does that undercut the case for hundreds of billions in Western AI infrastructure spend? Some commentary has already called this a "DeepSeek 2.0 moment."

III. What This Selloff Actually Means

Most analysts view this as a valuation reset and profit-taking event, not a collapse in underlying AI demand. TSMC's and Google's actual businesses both had strong quarters — what's really being sold off is the premium the market is willing to pay for future capex, not a vote of no confidence in the current business.

Tesla is the exception here: its decline reflects a genuine deterioration in margins and operating profit — a fundamentals problem, not just sentiment.

Summary

Taken together, these three reports prove the same point as yesterday's entry: an earnings report is never itself a reason to buy or sell — what actually moves the stock is usually the line about how much the company plans to spend next. For a long-term DCA holding like TSMC, the question that matters was never "how much did the stock fall this quarter," but whether this capex — being questioned right now — will look, five or ten years from now, like it built a deeper moat, or like money that shouldn't have been spent. No one can answer that on an earnings call. Only time can.

References