AI infrastructure demand now shows up from chips to fuel cells in company filings

Spending on artificial-intelligence infrastructure has become visible in the reports of companies that sit at different points in the supply chain. Third-quarter filings from chip makers, memory and storage producers, optical and networking suppliers, electronics manufacturers and power-equipment companies, published in the week of October 19, all point the same way: demand tied to AI data centers. At the same time, a market report on October 22 linked a stock sell-off to concern about that spending. This analysis lines up the figures, from the buyers to the suppliers, and notes what the filings do not say.

The demand signal in industry data

The Semiconductor Industry Association said global chip sales reached $159.7 billion in August, according to its monthly release, up 8% from $147.9 billion in July and 144.3% more than the $65.4 billion of August 2025. Sales into the Americas rose 180.9% from a year earlier, the fastest of the regions. Annual sales passed $1 trillion through August for the first time, and the association projects more than $1.6 trillion for 2026. The data are a three-month moving average compiled by the World Semiconductor Trade Statistics organization, so they smooth out single months.

The buyers: very large capital budgets

The firms spending on AI capacity report the biggest numbers. Meta spent $34.91 billion on capital expenditures in the third quarter, leaving free cash flow negative at $2.12 billion even though cash flow from operations was $32.79 billion. The company narrowed its 2026 capital expenditure outlook to $135 billion to $145 billion from $130 billion to $145 billion. SpaceX reported that its nameplate computing capacity reached 1.7 gigawatts on September 30, and that purchases of property and equipment were $48,069 million in the first nine months of 2026, against $13,846 million in the same period of 2025. Its cash and equivalents were $51,294 million.

The suppliers: chips, memory and networking

The suppliers are reporting growth on a similar scale. Super Micro Computer estimated revenue for the quarter ended September 30 of $18.5 billion to $19.5 billion, against its guidance of $14.5 billion to $15.5 billion, and gross margins of 12.0% to 12.5% against guidance of 10.4% to 10.8%, which it attributed primarily to customer and product mix. The numbers are preliminary and unaudited, and the company said actual results may differ materially.

Western Digital reported revenue of $4.38 billion, up 55% from a year earlier and 17% from the previous quarter, and expects second-quarter revenue to rise 66% to 72% from a year earlier. Texas Instruments, whose products are used in many industries, said revenue rose 28% to $6.06 billion, with "continued broad growth in industrial, automotive and data center". Its fourth-quarter outlook is for revenue of $5.80 billion to $6.30 billion, so the midpoint is about the same as the third quarter.

Nokia said net sales to AI and cloud customers rose 123% in the quarter and that it secured EUR 1.5 billion in new orders from these customers. It also said it expects its combined optical and IP network sales to grow 25% to 30% in 2027. MaxLinear's revenue rose 72% to $218.1 million, driven by an infrastructure business that grew 210%, which the company said reflected "accelerated adoption of optical AI data center products".

Parts and power

TTM Technologies reported net sales of $1.2 billion, up 60%, with its data center and networking end market at 41% of net sales and growing 115%, according to its chief executive. Camtek, which is based in Israel, guided to fourth-quarter revenue up 40% from a year earlier and an outlook for 2027 growth of more than 35%, with orders above $800 million in 2026, about 80% for advanced packaging.

Bloom Energy, which makes fuel cells that supply electric power on site, reported record revenue of $1,235.8 million, up 138.1%, and raised its 2026 revenue guidance to $4.4 billion to $4.6 billion. Its chief executive said that "what began with hyperscalers is rapidly moving across market segments because every business needs reliable power to grow."

Not every number is a boom

IBM reported total revenue up 5% and software revenue up 8%, slower than most of the growth rates above.

On Thursday, October 22, the Nasdaq Composite fell 2.6%. A market report attributed the fall to concerns about AI infrastructure spending, which it said outweighed concerns about inflation. This is the report's explanation, not a measured cause, and the filings above do not contain an assessment of whether the spending will continue.

Policy risks

Two government actions in October concern companies at the edges of this market. The Department of Homeland Security published an order on October 20 excluding information technology products and services developed by Hangzhou DeepSeek Artificial Intelligence from federal civilian agency procurements and removing them from federal systems. And on October 9 the International Trade Commission instituted an investigation under section 337 concerning vertical power delivery systems and components used in data center servers to power AI accelerators; the respondents include Foxconn, Quanta and Flex. Neither action contains findings about the companies' business results.

What to watch

The filings show demand for AI infrastructure running well ahead of the rest of the technology sector, and they show that the buyers are financing it with free cash flow that is already negative at Meta. What they do not show is how long the spending will last: Meta said only that it is still developing its 2027 plans. The next quarterly reports will show whether the guidance ranges above hold.