OpenAI disclosed benchmark data showing its custom Jalapeño chip outperforming Nvidia's GB300 processors on power efficiency and latency, adding a new competitive threat just as Bank of America warns semiconductor stocks could fall further. Nvidia reports earnings after the market closes Wednesday, with Bank of America still flagging the stock as a buy opportunity despite risks to its capital returns and sales growth.
OpenAI has disclosed benchmark results for its custom-designed inference chip, codenamed Jalapeño, showing it outperforming Nvidia's GB300 processors in internal and public tests run on the InferenceX platform. The disclosure lands as Bank of America warns semiconductor stocks could fall further, adding fresh pressure on Nvidia ahead of its earnings report Wednesday.
Jalapeño narrows Nvidia's inference edge
Jalapeño delivered 1.5 to 1.9 times more AI work per watt than Nvidia's GB300, and cut end-to-end latency by 1.7 to 3.6 times across models including GPT-OSS 120B and DeepSeek R1. The chip is an ASIC built with Broadcom and manufactured by TSMC, purpose-built for AI inference rather than the general-purpose training and inference Nvidia's GPUs handle.
Broadcom CEO Hock Tan said Jalapeño matches the performance of Nvidia's Blackwell architecture and Google's TPU while offering roughly a 50% cost advantage on a per-token and per-kilowatt basis. OpenAI reportedly went from schematic to tape-out in about nine months, with the company's own AI models assisting the chip's design.
OpenAI still needs Nvidia for training
OpenAI plans a limited deployment of Jalapeño by the end of 2026, with a full-scale rollout expected in 2027, and a second-generation chip already in development. The company will keep buying hardware from Nvidia, AMD and other suppliers for its computing needs, particularly for model training, where Nvidia still dominates.
Bank of America sees more downside for chip stocks
Bank of America analyst Vivek Arya sees a risk that the Philadelphia semiconductor index could fall as much as 10%, which would extend two months of declines and pull the index's valuation relative to the S&P 500 back to levels unseen since ChatGPT's 2022 launch. According to Arya, rising rates, data-center backlash and circular financing could cap chip stocks in the near term: "Headwinds could cap chip stocks near-term".
The index's price-to-earnings multiple trailed the S&P 500's by about 9% before ChatGPT's launch, swelled to a 15% premium during the AI boom, and has since equalized at about 20 times earnings for both. With chip companies growing forward earnings per share at roughly 70% a year, Arya called that valuation cheap despite the recent pullback.
Bank of America named Nvidia among its enhanced buy opportunities in chips, alongside Marvell Technology, Micron, Lam Research, AMD, Intel, Analog Devices and ON Semiconductor. For Nvidia specifically, Arya flagged potential for decelerating capital returns and lumpy, unpredictable sales in new enterprise markets as risks heading into Wednesday's earnings report.
Sources: Crypto Briefing, CNBC
Trading involves risk.
Risk Disclaimer
