OpenAI is shifting its enterprise pitch from raw token prices to cost per completed task, with CFO Sarah Friar detailing the strategy at Goldman Sachs’ Communacopia + Technology Conference in San Francisco. The company is targeting specialized sectors including chip design, life sciences, and financial services, arguing that a higher-priced model can ultimately be cheaper if it solves a problem in fewer attempts.
Friar said OpenAI recently cut the price of its lower-cost Luna model by 80%, helping drive roughly tenfold growth in usage. She also said the company’s coding assistant Codex has reached 25 million users. According to Friar, enterprise revenue grew 32% between June and July, outpacing overall annualized revenue growth of 20%, and OpenAI reached an even split between enterprise and consumer revenue by mid-year—ahead of its end-of-2026 target.
The company is using its custom inference chip, Jalapeño, as proof of what its models can do. OpenAI said the Broadcom-designed chip moved from conception to tapeout in nine months, and claimed it delivers 1.5 to 1.9 times higher AI work per watt than Nvidia’s GB200 and GB300 equivalent devices on workloads involving GPT-OSS 120B, DeepSeek R1, and Kimi K2.5. The chip is intended only for OpenAI’s own purposes and is not being sold commercially.
Independent comparisons provide a more mixed picture. Artificial Analysis priced GPT-5.6 Luna at roughly $0.18 for a standard task versus $2.01 for Z.ai’s GLM-5.3, but GLM-5.3 scored higher on the firm’s Intelligence Index—45 versus Luna’s 38. Friar argued directly that ‘if you’re deploying Luna and compare that to GLM 5.3, for example, on a cloud layer, we are cheaper.’
The strategic stakes are large. PwC estimated global AI-infrastructure capital expenditure could reach $31.6 trillion through 2050, with annual spending rising from about $800 billion in 2026 to $1.8 trillion in 2050. At the same time, the OECD said quality-adjusted AI model prices fell nearly 80% between January 2024 and April 2026, even as compute and chip inputs remain concentrated. OpenAI is betting that tighter integration across models, software, and silicon will turn that concentration into an efficiency advantage.
