OpenAI's president breaks down the AMD deal: 'We need as much computing power as we can possibly get' (www.businessinsider.com)

OpenAI announced a landmark agreement with AMD to deploy 6 gigawatts of GPU capacity — the first gigawatt, built on AMD’s Instinct MI450 accelerators, is slated to begin rolling out in the second half of 2026. The deal includes a warrant enabling AMD to issue up to 160 million shares to OpenAI if benchmarks are met; AMD says every gigawatt of deployed compute represents “double-digit billions” in revenue. The announcement follows OpenAI’s recent expanded relationship with Nvidia (including Nvidia’s reported up-to-$100 billion investment), and sent AMD shares sharply higher on Monday. For the AI/ML community this signals a serious industrial response to looming “compute scarcity.” OpenAI leaders framed the pact as necessary to match explosive demand for training and inference — Greg Brockman said OpenAI needs “as much computing power as we can possibly get,” and has ambitions like provisioning GPUs per user. Technically, the deal diversifies OpenAI’s hardware stack (Nvidia for many workloads, AMD MI450 for new capacity) and accelerates large-scale datacenter deployments that will affect model training throughput, latency for inference, and supply-chain dynamics. The timeline and scale underline that future progress in generative AI will be as constrained by power, cooling and chip supply as by models and algorithms.