Foxconn racing to supply AI racks — margins under pressure
Foxconn says AI server shipments are surging — management forecasts record 2026 revenue with rack volumes ramping sharply and claims a 40% share of AI server builds, while the company still posted a small profit miss for the quarter reported reported. That combination of scale and squeezed consumer margins is driving urgency for cost-efficient rack designs.
Foxconn said it already has the capability to manufacture roughly 1,000 AI server racks per week and signalled plans to raise that output further into 2026. (manufacturing.economictimes.indiatimes.com) The company formalised a co‑design partnership with OpenAI to build next‑generation data‑center hardware in U.S. factories, with the initial agreement carrying no purchase commitments but giving OpenAI early access to evaluate systems. (honhai.com) Hon Hai has been moving manufacturing closer to major cloud customers, announcing a $569 million expansion at its Wisconsin site to boost U.S. rack output and additional capacity increases around Jalisco, Mexico. (wedc.org) Quarterly results showed a narrower profit than analysts expected—net income of NT$45.21 billion versus a consensus near NT$60.88 billion—while full‑year sales still hit a record NT$8.1 trillion and net profit of about NT$189.3 billion. (msn.com) Management highlighted margin compression from the consumer‑electronics side and a higher tax rate as offsets to booming server revenues, and said server mix is predominantly GPU‑based (around 80%) with the remainder in ASIC/APU builds. (investing.com) To cut per‑rack cost, Foxconn and partners are standardising rack subsystems—cabling, networking and power modules—through the U.S. co‑design work announced with OpenAI, aiming to compress assembly and integration labour costs per rack. (marketbeat.com)
