Editor’s Note
AI demand stayed strong while the cost of money moved higher. The Fed raised rates, oil remained above $100, and the 10-year Treasury yield returned to about 5%. The bigger story was the growing gap between infrastructure demand and the revenue and cash flow needed to pay for it.
AI Demand Did Not Turn Into an AI Spending Pause
Calls for slower frontier AI development shook infrastructure stocks, but they did not produce a confirmed spending retreat.
The market split sharply:
- AI infrastructure stocks fell 3% to 7% in one session.
- Palantir, Atlassian, CrowdStrike, ServiceNow, Adobe, and Intuit rose 2% to 5%.
- HPE and Nokia fell 10.5% and 10%.
Company forecasts remained aggressive. Broadcom said the safety debate had not changed its AI semiconductor revenue forecasts of $115 billion in FY27 and $230 billion in FY28. Nvidia’s Jensen Huang said the company expects to sell twice as many chips next year as this year.
The demand figures were large:
- Global semiconductor revenue reached $425 billion in Q2 2026, up 31.4% from the previous quarter.
- Q3 revenue is expected to exceed $500 billion.
- Nebius raised GPU prices about 20%, after a roughly 30% increase four months earlier.
The buildout is also spreading beyond chips. Generac signed an agreement worth up to $8 billion to supply backup generators for Amazon data centers. Anthropic agreed to pay SpaceX roughly $1.25 billion per month through May 2029 for the full compute capacity of the Colossus 1 data center, including an initial 3.5 gigawatts of custom TPU capacity.
That is both a demand story and a funding story. OpenAI expects revenue to rise from $36 billion this year to $350 billion in 2030. It also projects $278 billion in negative free cash flow through 2030, alongside $856 billion in compute and infrastructure spending.
Rates and Oil Made the Buildout Harder to Finance
The Fed raised the federal funds target by 25 basis points to 3.75% to 4.00%, its first increase since July 2023. Its projections included a year-end fed funds target of 4.1% and PCE inflation of 3.7%.
Oil added pressure. Crude stayed above $100 a barrel, gasoline was up 45%, and diesel was up 67% and at a record high. The 10-year Treasury yield was reported at just below 4.86%, while another market account said it closed above 5%.
The Treasury’s $6 billion debt buyback also disappointed investors because estimates of $8 billion to $10 billion had circulated. Higher long-term yields raise borrowing costs for households and companies, which makes expensive AI projects harder to fund.
Housing showed the pressure clearly. The average 30-year mortgage rate reached 6.95%, while homebuilder confidence fell to 32, its weakest reading since late 2022.
The period therefore produced an uncomfortable combination: AI companies continued to plan for more chips, data centers, and power, while inflation and bond yields gave investors less patience for spending that has not yet produced matching cash flow.
Meta Put the Revenue Question in Front of Consumers
Meta’s Muse made the AI monetization test more concrete. The personal agent can book appointments, complete electronic forms, and monitor home security feeds.
It has a free tier and paid plans costing $20 or $100 per month, depending on usage. Muse also reached number one on Apple’s App Store, according to the supplied reports.
Meta is considering taking a share of shopping transactions completed through AI agents, but no concrete plan has been established. Users must opt out if they do not want eligible interactions used to train AI models. Meta says Muse operates in an isolated environment, does not see passwords or payment details, and asks permission before sensitive actions.
That trust question matters because Meta recently agreed to pay nearly $17 billion to settle claims brought by state attorneys general over alleged misrepresentation of harm on Facebook and Instagram.
Meta is also trying to reduce the cost of serving AI. It committed to more than 1 gigawatt of capacity using custom chips over a 12-month period, and early testing of its first Arke chips showed performance within 2% to 3% of simulations.
The stock market’s message was mixed. Infrastructure shares fell when investors questioned the pace of frontier development, but the reported demand forecasts did not change. The unresolved question is whether the selloff represented weaker demand or simply lower prices for future growth.
Counter-Thesis and Risk Watch
Michael Burry called OpenAI’s and Anthropic’s safety warnings “self-serving” IPO hype. The Wall Street Millennial speaker argued that slowing frontier AI could protect incumbents and reduce spending before possible IPOs.
That speaker cited reported commitments of $750 billion in data center spending through 2030 against annualized revenue below $40 billion as of August 13. These were the speaker’s claims and interpretations.
ClearValue Tax also questioned the inflation picture by comparing reported inflation of 3.4% with M2 money supply growth of 3.59% through July. Dan Ives named political opposition to data centers as one of the largest risks to the AI buildout.
Looking Ahead
Can AI revenue grow fast enough to support the spending now being planned? And can that spending continue if oil stays above $100 and long-term borrowing costs remain near 5%?