Meta raises the lower end of its capital spending guidance as general-purpose servers gain demand from AI agents

Why remain optimistic about the supply chain at year-end when third-quarter spending missed expectations?

Meta’s third-quarter cash capital spending was approximately 34 billion dollars, up 13% quarter on quarter and 80% year on year, but still below market consensus. J.P. Morgan attributes the gap primarily to the timing of production and shipments.
Meanwhile, the company raised its 2026 capital spending guidance from 130 billion–145 billion dollars to 135 billion–145 billion dollars, increasing only the lower end; the new midpoint implies year-on-year growth of approximately 94%. By the bank’s calculations, this implies a quarter-on-quarter increase of approximately 70% in fourth-quarter spending, with support for supply-chain revenue concentrated more heavily at year-end and in the following several quarters.
The case for 2027 rests on management’s reaffirmation of an aggressive investment in AI; there is no quantified capital spending guidance yet. On that basis, J.P. Morgan is optimistic about the following year’s investment environment, but has not given a specific spending figure.

Meta’s fourth-quarter revenue guidance is 73 billion–76 billion dollars, with the midpoint up 24% year on year and slightly above market expectations. J.P. Morgan believes engagement with core apps, advertising revenue and incremental demand from agent businesses such as Muse provide support.
AI investment has already generated returns through the advertising business: improvements to recommendations drove Facebook and Instagram engagement to new highs, while the use of large language models for ad retrieval and ranking and wider adoption of advertiser tools also improved monetization efficiency. The bank therefore concludes that revenue growth makes high capital spending more sustainable.
Another path for Muse is to expand its fee-based model as commercial transactions grow and reach more small and medium-sized businesses. Its workflows are compute-intensive, and greater usage will require additional computing capacity, thereby linking commercialization to server demand.

J.P. Morgan believes Muse’s AI agent workflows increase demand for CPU computing and could further boost general-purpose server shipments by cloud service providers. The bank also observes that enterprises are upgrading infrastructure to support agent applications.
In a scenario where Muse reaches 100 million monthly active users, the bank assumes 2 virtual CPU cores per user, agent workloads accounting for 80%, and computing utilization of 30% per user. The utilization assumption includes shared cores during idle periods and capacity provision for peak demand.
For the equipment, the model further assumes a virtual-to-physical core ratio of 2∶1, 64 physical cores per CPU, two CPUs per server, and a price of 30 thousand dollars per server. The model estimates an additional general-purpose server market opportunity at Meta of approximately 8.789 billion dollars.
If Wiwynn captures 50% of that opportunity, potential incremental revenue would be approximately 4.395 billion dollars, equivalent to 38% of the bank’s estimate for Wiwynn’s 2026 general-purpose server revenue from Meta and 10% of its total full-year revenue. This market opportunity scenario depends jointly on user scale, usage intensity, configuration and share, and has no specified year of realization.

For contract manufacturers, the direct link is Meta’s procurement: Wiwynn and Quanta are its main server suppliers. For switches, J.P. Morgan favors Accton based on its rising share of Meta’s 800G switches; ASPEED is also identified as a beneficiary of AI agent demand.
The case for branded servers also involves pricing: J.P. Morgan believes ongoing supply constraints support pricing and margins in the near to medium term, and therefore favors Lenovo. This earnings assessment depends on both demand growth and tight supply.