Foxconn has announced an ambitious goal to generate roughly $100 billion in U.S. revenue by the end of 2026, signaling a major strategic pivot towards American-based manufacturing.
This target isn't just a number; it translates into a massive operational scale-up. To hit this goal, Foxconn would need to assemble between 1,700 to 2,000 AI server racks per week in the U.S. The math checks out: at 1,700 racks per week with an average selling price of $1.1 million, the annualized revenue comes to about $97 billion. This demonstrates that the target, while aggressive, is grounded in a feasible production plan.
So, what's driving this monumental push? There are three primary factors. First is the explosive demand. Hyperscalers like Amazon, Microsoft, Google, and Meta are projected to spend a staggering $600-$700 billion on AI infrastructure in 2026 alone. This creates an enormous market for the server racks that are the backbone of AI data centers, and Foxconn is positioning itself to capture a significant slice of that pie.
Second, favorable policy is a key enabler. A recent U.S.-Taiwan trade agreement lowers tariffs on Taiwanese goods to around 15% in exchange for significant investment commitments in the U.S. This de-risks and sweetens the deal for onshore manufacturing, making it economically attractive for Foxconn to build its products closer to its biggest customers.
Third, Foxconn has the proven execution capability. The company already reached a capacity of 1,000 racks per week in late 2025 and has publicly stated its intention to increase this rate. This plan is further solidified by a strategic partnership with AI leader OpenAI to co-develop and manufacture AI hardware directly in the U.S., anchoring its expansion with a flagship customer.
However, the path is not without obstacles. The primary risk lies in U.S. power grid constraints. Key data center hubs in Texas, the Mid-Atlantic, and Silicon Valley are facing challenges in supplying enough power for the energy-intensive AI boom. If grid upgrades or interconnects lag, it could delay data center commissioning, which in turn would slow down Foxconn's rack deployments and push revenue recognition into later periods.
- Hyperscaler: A company that operates massive, globally distributed data centers, such as Amazon (AWS), Microsoft (Azure), and Google (GCP).
- Capex (Capital Expenditure): Funds used by a company to acquire, upgrade, and maintain physical assets such as property, plants, buildings, technology, or equipment.
- Rack-level assembly: The process of integrating multiple servers, storage, and networking components into a single standardized frame (a rack) before shipping it to a data center.