Intel says AI 'gold rush' is over, pushes prompt routing across open hardware

Intel's OCP Summit positioning is, in substance, a direct strike at NVIDIA's CUDA moat. Elevating prompt routing onto open hardware standards signals that AI inference is pivoting from compute monopoly to scheduling democratization — whoever owns the routing protocol owns the TCP/IP of the inference era. Technical cascade: HBM and CoWoS packaging premiums will erode as inference chip design shifts from peak FLOPS to tokens-per-watt efficiency. Data center topology degrades from monolithic GPU clusters into a hybrid mesh of routing nodes and edge inference units. Market dynamics: NVIDIA will likely fortify software lock-in via NIM microservices and DGX Cloud. AMD is positioned to exploit OCP alignment to accelerate MI300 ecosystem penetration. Hyperscalers face the most awkward position — open hardware directly undermines their vertically integrated custom-silicon-plus-private-cloud narrative. 12–24 month outlook: inference will undergo an Android-ization — hardware commoditizes, routing protocols become the new battleground. Intel's real play isn't selling silicon; it's selling rule-making authority. The gold-rush framing is a deliberate narrative reset to shift buyer psychology from buy-the-biggest-GPU to build-the-smartest-routing-layer.