OpenAI Says Its In-House Chip Handles 1.9 Times More Work Than NVIDIA's Chip [Global AI Briefing]

[Financial News] OpenAI officially announced that the first performance test of its self-designed artificial intelligence inference chip, "Jalapeño," showed far superior performance to NVIDIA Corporation's chips.
OpenAI plans to begin pilot deployment of Jalapeño in real-world AI infrastructure later this year and move into full-scale production next year, expanding its ecosystem from AI model development to in-house chip production.
On the 25th local time, OpenAI said that at "Hot Chips 2026" held at Leland Stanford Junior University (Stanford University), it tested its first in-house AI chip, "Jalapeño," co-developed with Broadcom Inc., using SemiAnalysis's AI inference chip benchmark platform, InferenceX. The results showed that compared with NVIDIA Corporation's GB200 and GB300 systems, the chip processed about 1.5 to 1.9 times more AI work per 1 kW of power and reduced total response latency by about 1.7 to 3.6 times.
This means that existing AI infrastructure could handle more AI requests with the same amount of power while also improving response speed.
This is the first time OpenAI has presented specific performance figures since unveiling Jalapeño with Broadcom Inc. in June.
OpenAI emphasized that Jalapeño is part of its full-stack strategy.
The company said it aims to deliver more "useful intelligence" with the same computing resources by optimizing every layer together, including AI models, chips, computing, software, and data centers.
OpenAI also stated, "Different AI tasks require different hardware," adding, "Rather than solving every computation with a single chip, we will choose the most efficient computing resources for each task."
Because Jalapeño is an inference chip rather than a training chip, it is designed to handle the computations needed when already trained AI models answer user questions or perform agent tasks.
This reaffirms OpenAI's strategy of not replacing NVIDIA Corporation's xPUs with Jalapeño all at once, but instead adding in-house chips optimized for specific tasks to reduce inference costs and power consumption as ChatGPT and AI agent usage surges.
Meanwhile, Anthropic has also officially announced that it has created an internal team to design custom chips for running Claude and will begin full-scale in-house chip development.
As global AI companies rush to develop in-house chips that can improve AI model performance and cut inference costs, market watchers say competition in the global AI industry is expanding into a race to build full-stack systems that connect AI models, agents, and hardware.
[email protected] Lee Gu-soon Reporter