OpenAI Enters Chip Design AI, Claims Low-Cost Models Are Cheaper Than Chinese Open-Source Alternatives
nashnova research
OpenAI CFO Sarah Friar told Goldman Sachs' Communacopia+ conference that its in-house Jalapeno chip went from design to tape-out in 9 months, while its budget model Luna — after an 80% price cut — now costs less to deploy than Chinese open-source rivals. This marks OpenAI's first direct pricing challenge to China's model ecosystem.
AI-designed chips — how fast, exactly?
OpenAI's in-house Jalapeno chip went from design to tape-out — finalizing the design and sending it to a foundry — in just 9 months, using the company's own AI models throughout.
This means → AI has moved beyond writing code; it is now embedded in hardware design, a process that traditionally takes far longer.
In plain terms = OpenAI used itself as the test case — built its own chip with AI, then turned that story into a sales pitch for other industries.
Luna's 80% price cut — really cheaper than Chinese open-source?
OpenAI slashed the price of its budget model Luna by 80%; usage surged roughly 10x afterward.
Friar named names: "If you deploy Luna, it is cheaper than running Zhipu AI's GLM 5.3 in the cloud."
This means → open-source models have long been seen as the budget option. OpenAI is now challenging that assumption — arguing a closed-source model can undercut them on price.
One caveat: this is OpenAI's own cost comparison. No third-party verification exists yet.
How fast is enterprise revenue growing?
From June to July, OpenAI's enterprise revenue grew 32% month-on-month — outpacing the roughly 20% growth rate in overall annualized revenue.
The revenue split between enterprise and consumer hit roughly 50-50 at mid-year, ahead of a year-end target.
This reflects a shift among enterprise clients from "testing AI" to "buying at scale," with willingness to pay accelerating.
What is changing in dev tools and pricing?
OpenAI's coding tool Codex now has 25 million users.
The company is exploring pricing tied to business outcomes rather than usage volume.
In plain terms = enterprise buyers don't want to pay for "how many API calls I made" — they want to pay for "how much money AI saved me." OpenAI is restructuring pricing along that logic.
What does this mean for the market?
OpenAI is competing with Chinese open-source models on two fronts simultaneously: cost (Luna's price cut) and capability (chip design with AI).
This means → the competitive focus is shifting from "whose model is smarter" to "who is cheaper and who delivers results in real-world use cases."
The key test ahead: whether Luna's pricing edge can drive enterprise-scale adoption, and whether cost advantages in specialized domains like chip design prove durable.
市场有风险，内容仅供研究参考，不构成投资建议。