Alphabet shed roughly 10% on Monday in one of its sharpest single-session losses in months, as investors ran out of patience with a free cash flow trough that the company’s own finance chief has signaled will deepen further before it recovers. The selloff hit the broadest cohort of AI infrastructure names \u2014 Palantir fell nearly 9%, Amazon and Meta each lost around 4%, and SpaceX dropped 16% after filing to launch its first-ever bond offering \u2014 but the market’s response was not uniform. While hyperscalers sank, memory chipmakers surged. SK Hynix closed up 5.6% in Seoul, overtaking Samsung Electronics to become South Korea’s most valuable listed company for the first time since 2000. In the US, Micron and other memory names rose ahead of Micron’s quarterly earnings report due Wednesday, June 24. The divergence captures something important about where AI investment risk is actually concentrated in mid-2026.<\/p>\n
Capex Numbers That Moved Markets<\/p>\n
The scale of what Alphabet is committing is nearly without precedent in the history of corporate capital spending. In 2022 the company spent roughly $31 billion building data centers, servers, and network equipment. In 2026 it expects to spend $180 billion to $190 billion \u2014 six times that figure and nearly double the $91 billion deployed in 2025. First-quarter capital expenditure alone reached $35.7 billion<\/a>, with roughly 60% directed at servers and 40% at data centers and networking. Free cash flow in Q1 fell 47% year over year to $10.1 billion, as spending consumed a growing share of operating cash. Consensus estimates project full-year free cash flow of roughly $20.5 billion in 2026 \u2014 down approximately 72% from the $73.3 billion Alphabet generated in 2025.<\/p>\n The situation across the hyperscaler tier is consistent in direction if not identical in magnitude. Combined 2026 capital expenditure<\/a> from Alphabet, Amazon, Microsoft, Meta, and Oracle now tops $452 billion. Amazon’s trailing free cash flow has collapsed 95% to $1.2 billion, even as its AWS cloud division grew 28% in Q1 to $37.59 billion \u2014 its fastest quarterly growth pace in 15 periods \u2014 against Q1 capital expenditure of $44.2 billion, up 77% year over year.<\/p>\n What made Monday’s session turn is not that these numbers were new \u2014 investors have known them since Q1 earnings in late April. What changed is the absence of a counter-signal. Alphabet CFO Anat Ashkenazi told analysts on the Q1 earnings call that 2027 capital expenditure would “significantly increase” compared to 2026’s already-elevated range. There is no guided year when the compression reverses. That is the sentence investors are sitting with.<\/p>\n Talent Exits Sharpen the Concern<\/p>\n The capex pressure became harder to dismiss on Monday alongside a personnel story that had been building since June 18. Noam Shazeer \u2014 vice president of engineering at Google and co-lead of its Gemini AI models<\/a> \u2014 announced his departure to OpenAI that day. Shazeer is among the eight co-authors of the 2017 “Attention Is All You Need”<\/a> paper, the research that introduced the transformer architecture underpinning virtually every modern large language model. Google reportedly spent approximately $2.7 billion acquiring Character.AI assets in 2024 specifically to bring Shazeer and a team of researchers back from the startup he had co-founded after previously leaving Google. He left for a direct competitor less than two years later. John Jumper, a Nobel Prize-winning researcher who led the development of AlphaFold2 at Google DeepMind, separately departed for Anthropic around the same time.<\/p>\n Together the departures flipped Alphabet’s sentiment on Reddit’s r\/stocks community from bullish to bearish within days \u2014 one widely shared thread titled “Google loses two top AI researchers to OpenAI and Anthropic” drew nearly 500 upvotes and more than 200 comments by Monday morning. Whether the exits signal a durable internal culture problem at Google DeepMind or represent a routine \u2014 if expensive \u2014 personnel shuffle in a market where elite AI researchers can command extraordinary leverage is the question the market is now pricing without a clear answer.<\/p>\n Why the Bull Case Has Not Collapsed<\/p>\n The selloff did not reflect weak revenue. Alphabet delivered Q1 2026 revenue of $109.9 billion, up 22% year over year, ahead of analyst expectations of $107.2 billion. Google Cloud revenue rose 63% to $20 billion \u2014 a record pace \u2014 and the company’s contracted cloud backlog nearly doubled sequentially to more than $460 billion, with roughly half expected to convert to recognized revenue within the next 24 months. Berkshire Hathaway committed $10 billion as the anchor investor in Alphabet’s $84.75 billion equity raise in early June, a conviction signal from one of the most studied long-term capital allocators in the market.<\/p>\n Consensus estimates project free cash flow recovering to roughly $35.5 billion in 2027 and roughly $68.1 billion in 2028, as the new data centers Alphabet is currently building come online and begin generating depreciation cycles that normalize the capital cost on the income statement. That recovery path \u2014 if it materializes \u2014 would make the current trough look like a temporary compression rather than a structural deterioration. The stock’s next concrete data point is Alphabet’s Q2 2026 earnings report, scheduled for July 28.<\/p>\n How SpaceX Bond Filing Compounded the Pressure<\/p>\n SpaceX, which went public on Nasdaq on June 12 at $135 per share and briefly traded above $225 by June 16, fell 16% on Monday after announcing its first-ever bond offering<\/a> of senior unsecured notes targeting at least $20 billion. The proceeds are earmarked to repay bridge financing arranged when SpaceX acquired Elon Musk’s AI startup xAI in February 2026. The announcement arrived less than two weeks after the largest initial public offering in Nasdaq history, and the market read it as evidence that the company’s AI infrastructure ambitions require more capital than its record-setting IPO alone could supply. SpaceX shares closed at $154.60, roughly 31% below their all-time high and just 14% above the IPO price.<\/p>\n Why Memory Chips Went the Other Direction<\/p>\n