OpenAI confirmed its custom Broadcom-designed accelerator is delivering strong performance, raising questions about long-term Nvidia dominance in inference workloads.
In today’s Homestretch update from the CNBC Investing Club with Jim Cramer, equity markets are slightly higher on Tuesday, with the S&P 500 on track to recover nearly all of Monday’s dip. The upward move is being driven by declining oil prices and bond yields, alongside broadly improved sentiment around artificial intelligence equities. At the center of this momentum is OpenAI’s Tuesday announcement that testing results for its new custom inference chip “show a significant performance advance.”
The processor, designated Jalapeño, was introduced in June through a joint venture with Broadcom and represents a pivotal step in OpenAI’s strategy to engineer the complete technology stack supporting its models and products. While Google stands as the premier example of a fully integrated AI company, and numerous competitors assert architectural superiority, Jalapeño has earned independent, third-party validation from SemiAnalysis. OpenAI noted that the chip “can process more AI workloads per unit of power while also delivering faster responses,” directly addressing a persistent industry constraint where legacy hardware systems “often require a trade-off between the two.”
OpenAI intends to begin deploying Jalapeño across its compute infrastructure by the end of the year. This milestone should reinforce investor confidence in Broadcom’s fiscal year 2027 AI revenue projection, which surpasses $100 billion. Nevertheless, the development does not shift our view on the regulatory and political pressures complicating data center construction, nor does it invalidate our decision to halve our Broadcom position on Monday. Market participants should expect questions regarding the sales trajectory during Broadcom’s September 2 earnings call, with particular focus directed at CEO Hock Tan.
Despite the advancement in custom silicon, OpenAI’s announcement should not be read as the commencement of a split from Nvidia. Just last week, Nvidia pledged up to $105 billion in credit backing for OpenAI’s expansive data center initiative in Ohio, a facility that will exclusively run on Nvidia compute architecture. Additionally, OpenAI confirmed it anticipates to “widely deploy accelerators from Nvidia and other partners for both training and inference workloads.” Custom chip rivalry remains a staple discussion point on Nvidia’s quarterly briefings, and we anticipate CEO Jensen Huang will field related inquiries once again on Wednesday evening.
Corporate reporting continues to dominate the near-term calendar. Intuit and Zoom will post results after the closing bell on Tuesday. Prior to Wednesday’s market open, earnings will be released by Kohl’s, Abercrombie & Fitch, Williams-Sonoma, and JM Smucker. Macroeconomic releases will feature the Federal Reserve’s preferred inflation metric, the PCE price index, alongside the second estimate for second-quarter GDP and durable goods orders.
As a participant in the CNBC Investing Club with Jim Cramer, subscribers are notified via trade alert prior to any portfolio action. Jim Cramer observes a 45-minute waiting period after issuing an alert before executing trades for his charitable trust, extending that interval to 72 hours if the security has been featured on CNBC television. All investing club materials are governed by applicable terms, conditions, privacy guidelines, and disclaimers. No fiduciary duty is established through receipt of this information, and no specific financial outcome or profit is assured.