DigiTimes复盘Meta财报：2026年capex上调至1350–1450亿美元，纬颖/广达为Meta服务器主板两大供应商，但无新增订单

Meta has become the first of the four major US hyperscale cloud providers to report its latest earnings, with supply chain players closely monitoring changes in the company's capital expenditure plans and investment priorities.

Meta CEO Mark Zuckerberg said the company is continuing to expand its computing capacity to serve more than one billion users in the coming years, signaling sustained investment in AI computing infrastructure.
Notably, Zuckerberg emphasized the economics of computing, explaining that AI training involves substantial costs. With leading models already in place, Meta intends to meet as much high-margin inference demand as possible.
This strategy aligns with growing industry expectations that the shift in AI workloads from training toward inference will provide fresh momentum for CPU server shipments.
Meanwhile, the scale of AI computing investment has raised financial concerns. Trading in credit default swaps (CDS) tied to debt financing for AI infrastructure has recently increased, adding to worries over the sustainability of AI-related financing.
US Treasury Secretary Scott Bessent has weighed in, expressing confidence in AI's potential benefits for the US economy. He also emphasized his ability to recognize signs of systemic financial stress, saying he currently sees no systemic problems in AI development.
Supply chain sources, for their part, maintain that customer orders remain strong and the outlook positive, with no indication of weakening demand.
The only significant challenge involves component shortages and supply imbalances, although their severity varies depending on the customer and server model. Given the substantial purchasing power of the four major US cloud service providers (CSPs), these companies are relatively less exposed to component shortages.
Meta expands computing capacity to support billions of AI users
During Meta's third-quarter 2026 earnings conference call, Zuckerberg highlighted the recent launch of Muse, the company's personal AI agent, describing it as one of Meta's most successful product introductions.
Within just six weeks of its launch, more than 10 million users across the United States and Canada had configured Muse.
Zuckerberg outlined three priorities for Meta: scaling its products, driving revenue growth, and optimizing computing costs.
With Meta aiming to expand Muse to billions of users, the company is aggressively increasing computing capacity to support those ambitions.
Capital expenditure figures provide the clearest indication of its investment trajectory.
Meta reported third-quarter capital expenditures of US$34.9 billion, including principal payments on finance leases, primarily directed toward servers, data centers, and networking infrastructure.
Meta CFO Susan Li said the company now expects total capital expenditures of US$135 billion to US$145 billion in 2026, raising the lower end of its previous guidance from US$130 billion.
However, when asked by analysts about the company's capital expenditure outlook for 2027, Li said Meta remains in the very early stages of developing its budget and computing capacity plans for next year.
She indicated that the company expects to provide more specific guidance during its next earnings conference call.
Wiwynn and Quanta closely monitor Meta's server procurement
According to DIGITIMES Intelligence, Taiwanese manufacturers' server motherboard shipments by customer in the third quarter of 2026 indicate that Wiwynn and Quanta are Meta's two principal server suppliers.
Wiwynn accounted for 72.7% of shipments, while Quanta was primarily responsible for the remaining portion, which the data identifies as 33.3%.