OpenAI CFO Sarah Friar explains how advances across chips, compute, models, and products compound to deliver more useful intelligence at greater scale and lower cost. Progress in AI compounds fastest when the entire system improves together. OpenAI views its compute strategy as one integrated system spanning data centers and chips, frontier models, developer platform, consumer and enterprise products, and AI-native devices, with each layer strengthening the next. Better software makes hardware more productive, while hardware designed for OpenAI workloads improves speed and efficiency. More capable models unlock better products, generating more demand, usage, and learning that flows back through the system to improve it again.

Today, OpenAI shared the first measured performance results from Jalapeño, its first custom inference chip. On Inference X, a public benchmark using GPT-OSS 120B, Jalapeño delivered more peak throughput per kilowatt and lower token latency than the commercial systems in the comparison. It also performed strongly on Deep Seek R1 and Kimi K2, showing gains extend across model families. Jalapeño widens the lead at previous-best token throughput per dollar, giving OpenAI greater control over how models run and the economics of serving them. By developing the model, serving software, chip, memory, and network together, OpenAI can improve throughput, latency, energy efficiency, and cost as one system.

OpenAI's portfolio includes Microsoft compute and NVIDIA chips, alongside AWS, AMD, Broadcom, Cerebras, Core Weave, Oracle, SB Energy, and Soft Bank. Each brings different strengths across cloud infrastructure and accelerated computing. OpenAI actively manages this portfolio for both capability and economics, using premium systems where capability matters most and optimizing for efficiency at scale. Project Camellia in Georgia shows how OpenAI can design facilities around customer workloads while creating jobs and supporting local businesses. Turning efficiency into economic value, better models reach the right answer with fewer attempts, and smarter routing reduces wasted work.