Sovereign AI spending is about to supercharge demand for NVIDIA’s rack-scale systems, with a new JPMorgan analysis of supplier Foxconn pointing to a surge that could reshape the AI infrastructure market over the next few years.
According to the report, national AI initiatives are rapidly becoming a dominant buyer category for hyperscale-class compute. Foxconn, also known as Hon Hai, expects sovereign AI investments to top one trillion dollars within five years as governments bankroll large, strategic AI projects. Efforts such as the United States’ Stargate program and Saudi Arabia’s Humain AI are cited as examples of this momentum, bringing both capital and logistical support that can sustain multi-year buildouts.
One project looms especially large. Foxconn is reportedly more optimistic about the US-backed Stargate initiative, which targets a 10-gigawatt mega-facility. The computing footprint envisioned here is massive—potentially equal to around 70,000 NVIDIA GB200 AI servers. That single deployment could absorb years of production, highlighting just how deep the order book for high-end AI infrastructure may run.
Against this backdrop, NVIDIA’s current and next-gen platforms appear set for continued strength. Foxconn expects demand to remain robust for GB200 systems while production of GB300 Blackwell Ultra racks ramps. JPMorgan estimates Foxconn’s AI revenue could climb by as much as 63% into next year, supported by yearly rack-scale demand of roughly 50,000 to 60,000 units across configurations such as NVL72 and NVL36.
The implications extend beyond server supply. As AI clusters scale, operators will need to secure unprecedented amounts of power and cooling capacity, making energy procurement and data center design as critical as silicon availability. Even so, the report makes clear that NVIDIA and its partners are positioned to ride the next wave of AI buildouts, with sovereign AI emerging as a powerful new engine of demand.
Bottom line: Government-backed AI is accelerating from ambition to infrastructure, and that shift could keep NVIDIA’s rack-scale pipeline full well into the future.
