How Nvidia Uses Its Soft Power to Influence Customers and Shape Policy - The Information

Nvidia's "soft power" is fundamentally an institutional lock-in mechanism. The CUDA ecosystem, co-design partnerships, and developer community form three layers of switching cost that push customer migration from a technical problem to an organizational inertia problem. More critically, the definitional boundary of "high-performance computing chips" in export control frameworks is effectively anchored by Nvidia's own product specifications. This isn't lobbying—it's writing your own specs into the rulebook. Chain reaction: When policy benchmarks to Nvidia, AMD's ROCm, Huawei's CANN, and hyperscaler custom silicon (TPU, Trainium) are all relegated to "non-standard" status. Ecosystem fragmentation accelerates irreversibly. On compliance, multi-architecture procurement now stacks export licensing, data sovereignty audits, and supply-chain traceability—costs compound exponentially. Competitive response: AMD will deepen sovereign-capital partnerships. Huawei leverages domestic-substitution mandates to build a closed stack. Hyperscalers elevate custom silicon from backup to strategic hedge. 12-24 month call: Global compute infrastructure will fracture along geopolitical axes into two or three mutually incompatible stacks. The more successful Nvidia's soft-power play, the faster the bifurcation becomes irreversible.