Every major Nvidia customer is now a competitor as Microsoft reveals new Maia chip - Axios

Microsoft's Maia launch marks the irreversible onset of vertical integration in AI compute. For five years, Nvidia sustained 75%+ gross margins through CUDA lock-in and generational hardware leads. Now, Microsoft, Amazon's Trainium, and Google's TPU collectively push the market toward an oligopoly-for-training, fragmentation-for-inference equilibrium. Technically, Maia's ARM-based architecture forces the software stack to migrate from a single CUDA backend to multi-backend compilation. The real moat shifts from silicon to compilers and operator libraries. Upstream, TSMC's CoWoS advanced-packaging capacity in Taiwan, China transitions from single-customer concentration to multi-source allocation, reshaping pricing power. On compliance, tightening export controls reduce cloud providers' tail-risk of single-supplier disruption, yet foundry dependence on Taiwan, China remains a structural geopolitical vulnerability. Strategic response: Nvidia will likely deepen DGX system-level bundling and leverage NVLink interconnect to raise switching costs. However, within 12-24 months, inference workloads will see material de-Nvidiaization, and data-center gross margins will likely compress from 75% toward the 60-65% band. Training remains defensible near-term, but ecosystem openness is the existential long-term threat.