Analysis: Nvidia ramped up AI investments to US$45.3 billion in 2026, reshaping AI supply chain

Nvidia's $45.3B AI investment in 2026 is less about expansion and more about vertical consolidation of compute sovereignty. Technically, its focus on optical interconnects and custom inference ASICs pressures TSMC’s CoWoS capacity toward HBM-integrated photonics, crowding out general-purpose GPU alternatives. Compliance-wise, tightening U.S. export controls on advanced semiconductor tools force Nvidia to reroute supply chains through Mexico or Vietnam—adding 15–20% to operational costs. In response, AMD tightens MI300X integration with Azure, while Google’s TPU v6 pivots to in-house silicon photonics. Over the next 18 months, a ‘compute arms race’ tailwind will cement Nvidia’s full-stack dominance: smaller model firms must adopt its proprietary stack, turning ‘AI democratization’ into a myth. The real beneficiaries? Niche players holding advanced packaging and optical I/O IP.