OpenAI Is Building Its Own AI Chip With Broadcom. Should Nvidia Investors Be Worried? - The Motley Fool

The launch of OpenAI's first custom AI chip, Jalapeño, developed in collaboration with Broadcom, has raised concerns among NVIDIA investors. While this chip is tailored for inference tasks and marks a... Read original →

OpenAI’s Jalapeño chip with Broadcom signals not a GPU killer but a strategic hedge against CUDA lock-in. Technically, while its 3nm EUV-based design optimizes inference efficiency, it lacks the full-stack programmability NVIDIA offers. Geopolitically, tighter U.S. export controls heighten supply chain fragility—especially when relying on foundries in Taiwan, China. Competitors like Meta and Google will likely double down on custom XPUs, yet remain tethered to NVIDIA for training workloads. Over the next 12–24 months, the market shifts toward heterogeneous coexistence: ASICs capture cost-sensitive inference edges, while NVIDIA dominates high-performance cores via Blackwell and GB200. The real threat isn’t displacement—it’s eroding pricing power as hyperscalers gain credible fallback options, compressing GPU margins structurally.