OpenAI is expanding its advanced models into chip design, life sciences, and finance. The company argues that a model with a higher upfront cost can ultimately be more economical by performing tasks with fewer tries.
According to reports from Reuters, CFO Sarah Friar presented this argument at Goldman Sachs’ Communacopia + Technology Conference in San Francisco. Instead of focusing solely on the lowest token price, OpenAI encourages businesses to evaluate which model offers the cheapest completed task.
Chip Design and the Jalapeño In-House Proof Point
OpenAI is not attempting to replace electronic design automation systems currently used by engineers. Instead, its solutions assist in enhancing those processes by helping engineers analyze problems, evaluate approaches, and accelerate specific tasks.
The company's custom inference chip, Jalapeño, serves as an in-house proof point. OpenAI stated that its AI technology facilitated the chip's transition from conception to tapeout within nine months. Designed by Broadcom strictly for OpenAI's internal use and not for open market sale, the Jalapeño chip demonstrated higher AI work-per-watt performance in company figures compared to certain Nvidia equivalents. SemiAnalysis observed the InferenceX runs but did not independently reproduce the full test suite.
Market Context and Competitors
OpenAI enters a field where AI-assisted chip design is already established. Companies like Synopsys and Cadence offer generative-AI tools that reduce time-to-solution and shorten validation cycles, while Google’s AlphaChip has also been utilized in advanced semiconductors.
Independent testing by Artificial Analysis compares pricing across different models, noting variations between price per standard task and performance intelligence scores. Meanwhile, broader industry forecasts from PwC estimate that global AI-infrastructure capital expenditure could reach $31.6 trillion through 2050, driven largely by recurring chip and server upgrades.
