Nvidia crashes Intel’s party: $5T giant surges as AI market pivots to CPUs

The key second-order read is that AI capex is broadening from GPU-heavy training into a CPU-rich inference stack, which shifts bargaining power toward vendors that can bundle the whole rack. That is incrementally bullish for NVDA because it can monetize the control plane around inference workloads, but it is also the cleanest reopening of the semis landscape for INTC and, to a lesser extent, AMD/ARM as CPU content per deployment rises. The most important implication is not unit growth alone; it is that each AI server may require a higher-value mix of CPUs, networking, memory, and interconnect, making the ecosystem less winner-take-all than the market assumed.

The market is still underpricing how fast “good enough” CPU performance can matter when inference scales across enterprise and consumer endpoints. If inference stays the dominant near-term workload, the constraint becomes latency, power efficiency, and total system cost rather than raw training throughput, which could compress NVDA’s marginal share gains while improving pricing power for CPU incumbents. That said, INTC’s move is still a sentiment-driven rerating until customers validate socket wins in actual deployments; the next 1-2 quarters of design-win commentary matter more than the earnings beat itself.

Contrarian view: the move may be overextended on the idea that one earnings report resets a multiyear competitive gap. A lot of the upside is already reflecting a regime shift in expectations, so the stock/sector reaction likely outruns fundamentals over the next few weeks even if the thesis is directionally right. The cleaner setup is to fade the most crowded expression of the narrative after momentum cools, while keeping exposure to the names with the best operating leverage to inference adoption and the least balance-sheet or execution risk.