AINEWS
Alphabet's record US$119.8b quarter couldn't offset US$44.9b in capital expenditure, pushing its free cash flow negative for the first time since 2004.
Alphabet made a record US$119.8 billion quarter and the stock fell about 7% the next day. For the first time since the 2004 initial public offering, Alphabet's free cash flow went negative, dragged down by US$44.9 billion of capital expenditure in just three months. The sharemarket is no longer rewarding every AI dollar a hyperscaler spends; it is starting to price on the cash those dollars leave behind.
Alphabet's Q2 2026 result was a top-line blowout. Revenue climbed 24% year-on-year to US$119.8 billion, the highest quarterly print in the company's history. Capital expenditure roughly doubled, and full-year guidance was lifted to about US$200 billion, with chief financial officers already flagging a further rise in 2027. The 24% growth did not save the share price. Charlie Aitken, portfolio manager at Regal Partners, said a quarter like that "would have sent the stock up a year ago." This one shaved about 7% off the market value in a session.
The reason is unit of account, not narrative. The market is repricing on sustainable free cash flow, and for the first time the biggest spenders in artificial intelligence are running their capex ahead of the cash the underlying business throws off. Alphabet's free cash flow of negative US$5.9 billion in the quarter is the first negative print since 2004. That is the trade flipping, and it is no longer just an equity trade.
The bond market joined it. Hyperscaler bond yields widened, and Amazon had to offer extra yield to clear a US$25 billion bond sale, an unusual concession for a name with that balance sheet. A company the credit market treats as a sovereign can only place paper by paying up; that is the credit side of the same discipline trade. Aitken's read on the bond tape is that it is "kind of saying, slow it down a bit, lads." Equity investors are pricing the same move on the equity side, and it is the first time in this cycle both sides of the capital structure have agreed.
Apple, by spending far less on AI infrastructure, retook the title of the world's most valuable company. Tesla fell about 14% on its own result. The pattern is consistent: the highest spenders got marked down, the lowest spender got marked up, and the in-between names moved on their own AI exposure rather than business momentum.
There is a second beat to this story, and it is in Australia. The largest names on the Australian Securities Exchange (ASX) software tape have been sold off as if AI is about to wipe out their businesses. Xero is down about 64% over the past year, WiseTech Global has more than halved in 2026, and Seek is off about 30% year-to-date. Revenue and cash flow for those three names have largely held. The cut is to the price-to-sales multiple, not to the earnings power of the businesses. Aitken's description is that the selloff has been "indiscriminate. It's a global event... There's definitely clear value in those SaaS names. They're mostly good businesses with high barriers to entry, and not just can be replicated by AI."
This is the portable mental model: separate the AI capex discipline, which is real and is being applied to the hyperscalers, from the indiscriminate SaaS haircut, which has overcorrected in places like Xero, WiseTech and Seek. The first is a verdict on cash flow math; the second is a verdict on fear. The cross-asset confirmation is the discipline signal. When the software bounce comes, it will be a multiple rerating, not a business recovery, because the businesses have not deteriorated.
None of this is a buy recommendation. The legitimate criticism of the spenders still stands: the capex may not pay off, and some software fears are well-founded for names whose products are genuinely substitutable. Goldman Sachs has flagged US$500 billion-plus in AI investment for 2026 alone, and the same caution is showing up in research from Morgan Stanley, J.P. Morgan Asset Management and HSBC, with HSBC explicitly highlighting AI-overspending concerns as a driver of market sentiment.
The next data point is mechanical. Watch the next hyperscaler result for the same test Alphabet just failed: revenue growth running ahead of, or behind, capital expenditure growth in the same quarter. Watch the hyperscaler bond yield tape the day any new deal is priced, because Amazon's yield concession on a US$25 billion deal is the precedent the rest of the complex will be marked against. The trade has flipped, and the bond market has confirmed it. The next print is the falsifier.