Foxconn-Nvidia $1.4 billion Taiwan supercomputing cluster to be ready by H1 2026, Foxconn says

Foxconn said it is building a $1.4 billion, 27-megawatt supercomputing centre with Nvidia that will be ready in H1 2026 and, when complete, be Taiwan's largest advanced GPU cluster and Asia's first data centre using Nvidia's Blackwell GB300 chips under a new AI unit, Visonbay.ai. The move cements Foxconn's transition into AI infrastructure — it is now Nvidia's primary maker of AI racks, can produce about 1,000 AI racks per week, plans to invest $2–3 billion annually in AI, and expects AI to drive 2026 growth — positioning the company to benefit from large cloud and enterprise AI capex. Nvidia executives at the event emphasized the economics of renting compute, signaling demand for third-party AI compute providers, while Foxconn also highlighted progress on EV outsourcing as it scales manufacturing capacity.

Analysis

Foxconn (Hon Hai Precision, 2317.TW) said it is building a $1.4 billion, 27‑megawatt supercomputing centre with Nvidia that will be ready in the first half of 2026 and, according to the company, will be Taiwan's largest advanced GPU cluster and Asia's first data centre using Nvidia's Blackwell GB300 chips under a new AI unit, Visonbay.ai. Nvidia personnel at Foxconn's tech day noted that renting compute may offer better ROI and flexibility as GPU technology accelerates, underlining potential market demand for third‑party AI compute.

Foxconn highlighted that it is now Nvidia's main maker of AI racks, can manufacture about 1,000 AI racks per week with an expected increase next year, and that Chairman Young Liu plans to invest $2–3 billion annually in AI with AI as a major driver of 2026 growth. These operational statements position Foxconn to capture cloud and enterprise AI capex, reinforcing its role in the AI infrastructure supply chain while deepening its manufacturing relationship with Nvidia.

Key implications include potential revenue and margin upside from AI‑rack manufacturing and colocation services, optionality toward EV outsourcing as Foxconn’s EV volumes reach levels amenable to third‑party production, and strategic alignment with Nvidia on factory AI adoption. Material risks are execution and timing to meet the H1 2026 target, the capital intensity of the $1.4 billion project plus the planned $2–3 billion annual AI spend, and concentration risk tied to GB300 adoption and Nvidia’s ecosystem.

Key Decisions for Investors

• Consider increasing selective exposure to Foxconn (2317.TW) to participate in AI infrastructure upside given the $1.4bn GB300 data centre project, 1,000‑rack weekly production capability and $2–3bn annual AI investment, while sizing positions for execution and timeline risk
• Maintain or add exposure to Nvidia (NVDA) as a direct beneficiary of GB300 adoption and growing demand for AI racks, and monitor early deployment signals and partner announcements for revenue cadence
• Track three near‑term milestones — H1 2026 data‑centre completion, AI‑rack production ramp, and the annual AI capex run‑rate — and hedge or trim positions if milestones slip or GB300 uptake lags
• Monitor market demand for rented compute and third‑party AI infrastructure as a structural revenue channel; consider selective opportunities in colocation or AI‑rack suppliers if commercial offerings and contracts expand