Editor’s Note
The period began with strong cloud growth and plans for more AI capacity. It ended with a sharp tech selloff as investors focused on possible limits to data center construction. The business numbers still show demand, but they do not settle how quickly companies can build or whether the spending will earn enough.
AI Demand Is Growing, But Building Capacity Is in Question
Cloud growth and backlogs gave investors evidence of demand. One commentator put major cloud companies’ signed revenue backlog at $2.48 trillion. Google Cloud revenue grew 103% to $30.7 billion, Azure grew 48%, and AWS grew 39%. Amazon CEO Andy Jassy said the company was installing an AI accelerator about every 14 seconds, with customers using new capacity almost as fast as it arrives.
The scale of planned infrastructure is striking. An account citing Elon Musk said buildings in Memphis would hold around 1.2 million Nvidia Blackwell GPUs in about two months, with estimated electricity needs of around 2 gigawatts. Another account estimated that 2027 AI systems could need ~200kW to ~1,000kW per rack, compared with ~10kW for traditional CPU racks.
Then the political risk became more concrete. After Democrats won control of both houses of Congress, Goldman Sachs warned that state and local Democratic victories could affect 45% of planned US data center capacity, while expecting any construction pauses to be short-lived. Some lawmakers called for pauses or stronger rules on energy use, water, and grid impacts. Texas governor-elect Gina Hinojosa said data centers had moved into mostly rural communities without rules to protect Texans. An account of her interview said she appeared to support continuing projects that had started and received all permits, while pausing new permits until protections were in place.
That distinction matters: the argument is over the pace and location of new construction, while cloud companies are reporting strong demand. The selloff did not establish which projects will face tighter limits.
Strong Results Could Not Protect Chip Stocks
Company results remained strong across parts of the AI supply chain, but stock reactions turned sharply negative. AMD reported $13.4 billion in Q3 revenue, up 45%, and data-center revenue of $8.55 billion, up 97%. It guided for about $17 billion in Q4 revenue. Free cash flow, though, was $554 million, below the $1.05 billion expected.
Other reported results were also large: Super Micro Computer revenue quadrupled to $19.9 billion, Astera Labs revenue grew 154%, Nvidia revenue doubled to $96 billion, and TSMC revenue grew 41%. Those results sat beside a rout in technology shares. On Thursday, the S&P 500 fell 4.4% and the Nasdaq dropped 5.5%; the semiconductor index lost about 12%. Nvidia fell roughly 9%, AMD and Micron around 17%, and SanDisk more than 20%. The Nasdaq 100 closed 1.4% higher Friday after falling 6.4% in the previous session.
The week exposed a gap between business performance and share prices. Strong reported growth did not prevent investors from selling chip stocks when the construction outlook became less certain. One market account also said analysts expected AI capital spending growth to peak in 2027, and cited a 17% yield for CoreWeave. That is a concern about the pace and financing of expansion, not evidence that AI spending has stopped.
Big AI Plans Meet Questions About Cash
The spending question was already visible in the cash figures. One comparison said Microsoft spent 68% of operating cash on capex last quarter, versus 106% for Meta, 121% for Google, and 179% for Amazon. Amazon reported $75.9 billion in property and equipment spending over three months, against $42.4 billion from operations. AWS grew 39%, with a reported 38.7% margin and $578 billion backlog, but Amazon’s trailing free cash flow was reported at -$42 billion.
Anthropic’s filing showed another side of the buildout. It reported an $80.7 billion net loss for the first six months of the year, with more than $72 billion attributed to changes in the value of financing deals. An account of the filing put operating loss at just under $700 million, down from $2 billion in the same period last year. It also reported almost $4 billion in positive operating cash flow in the second quarter, while computing spending fell from 251% of sales early last year to 48% in that quarter.
The future commitments are large, too. Anthropic entered computing agreements with minimum obligations of $417 billion since the end of 2025; including other deals, an account put commitments above $800 billion. These figures do not answer whether future revenue will cover those commitments.
Meta’s personal AI agent, Muse, is one attempt to put a price on AI services. Meta introduced a free tier and monthly plans of $20 or $100. The company says Muse can book appointments, fill out forms, and monitor home security feeds. Users must opt out if they do not want interactions used to train Meta’s models. Meta is exploring a cut of shopping transactions through agents, but has no concrete plan.
Rates Add Another Unsettled Question
The October jobs report added a separate tension. The US added 165,000 jobs, against 59,000 expected, and unemployment fell to 4.1%, below the 4.2% consensus. At the same time, people working part-time for economic reasons rose by 339,000 to 4.8 million. The 10-year Treasury yield rose to just below 4.86% after a $6 billion Treasury bond buyback came in below Wall Street estimates of $8 billion and $10 billion.
President Donald Trump renewed his threat to cut off trade with countries that have surpluses with the US unless the Fed lowers rates. The stronger jobs figure and higher long-term yield sit alongside that call for lower rates, leaving the next rate decision unsettled in the supplied material.
Counter-Thesis and Risk Watch
One commentator argues that AI revenue may not grow fast enough to cover Anthropic’s commitments. Using assumptions about limited prospectus disclosures, the commentator estimated roughly $80 billion in data center commitments in 2027 and $100 billion in 2028, and said Anthropic would need about $120 billion in gross revenue to make $100 billion in net revenue. Those are the commentator’s estimates and view.
Other commentators argue the selloff could reflect forced selling and a policy shock, while AI demand remains strong. A semiconductor skeptic calls buying chip stocks now gambling, saying weaker AI demand could bring lower sales and margins for Nvidia, while acknowledging that a fast recovery could happen. The competing explanations remain unresolved.
Looking Ahead
Can companies keep adding data center capacity as fast as customers take it up, and can the resulting revenue cover the spending?