AI Labs Build Their Own Chips

Updated: 2025.10.13 8D ago 4 sources
OpenAI’s chief product officer says the company is developing in‑house chips and using AI to optimize chip design and layout. Vertical integration would reduce reliance on Nvidia and tight supply chains while tightening the link between model design and custom silicon. — Control of hardware becomes a strategic lever in AI competition, reshaping antitrust, export‑control, and industrial‑policy debates.

Sources

OpenAI, Broadcom Forge Multibillion-Dollar Chip-Development Deal
msmash 2025.10.13 90% relevant
The article reports OpenAI will design its own GPUs and co‑develop custom AI chips and systems with Broadcom, directly matching the trend of labs moving away from off‑the‑shelf Nvidia toward vertically integrated, in‑house silicon.
Microsoft's CTO Hopes to Swap Most AMD and NVIDIA GPUs for In-House Chips
EditorDavid 2025.10.04 70% relevant
Like OpenAI’s move toward in‑house silicon, Microsoft’s Kevin Scott says Redmond aims to run the majority of its AI workloads on its own Maia accelerators, reducing dependence on Nvidia/AMD for core infrastructure.
Links for 2025-09-06
Alexander Kruel 2025.09.06 86% relevant
The roundup notes OpenAI will launch in-house chips co-designed with Broadcom for internal use, directly aligning with the idea that leading labs are vertically integrating hardware to reduce reliance on Nvidia.
Links for 2025-08-24
Alexander Kruel 2025.08.24 100% relevant
Kevin Weil: “We’re working on our own chips and using AI to improve chip design and layout. We’d be crazy not to.” (video interview timestamped in the post)
← Back to All Ideas