Newman credits Broadcom for OpenAI Jalapeño chip, dismisses Nvidia killer hype
- Futurum CEO Daniel Newman credits Broadcom for OpenAI's Jalapeño chip performance
- Jalapeño showed 1.5-1.9x more AI work per watt vs Nvidia GB200/GB300
- Newman dismisses "Nvidia killer" narrative as exhausting hyperbole
- OpenAI plans Jalapeño deployment later this year with ramp in 2027
- Broadcom shares rose 0.54% in premarket trading after initial dip

*this image is generated using AI for illustrative purposes only.
Futurum Group CEO Daniel Newman credited Broadcom Inc. (NASDAQ: AVGO) for the performance of OpenAI’s new Jalapeño AI chip, while warning against labeling the custom silicon an Nvidia Corp (NASDAQ: NVDA) "killer."
Newman made the comments on X after OpenAI published benchmark results for Jalapeño, its first custom inference chip developed with Broadcom. He stated that Broadcom is "likely the reason this went well," though he reserved judgment on the broader silicon program pending external testing.
Benchmark Performance and Efficiency
OpenAI evaluated Jalapeño using InferenceX, a benchmarking platform designed to assess AI inference performance. The company compared the chip against Nvidia’s GB200 and GB300 superchips, which were the top results available at the time.
According to OpenAI, Jalapeño delivered 1.5 to 1.9 times more AI work per watt across GPT-OSS 120B, DeepSeek R1, and Kimi K2.5 1T models. It also achieved 1.7 to 3.6 times lower end-to-end latency.
| Metric | Improvement Factor | Comparison Basis |
|---|---|---|
| AI work per watt | 1.5x to 1.9x | Nvidia GB200/GB300 |
| End-to-end latency | 1.7x to 3.6x lower | Nvidia GB200/GB300 |
The chip is designed specifically for AI inference rather than model training. OpenAI plans to begin deploying it in its infrastructure later this year, with production expected to ramp in 2027. The company is also developing second- and third-generation versions.
Skepticism on Nvidia Replacement Narrative
Newman pushed back against the growing narrative that custom chips will immediately displace Nvidia hardware. He described the "$NVDA killer hyperbole" as "exhausting" and called OpenAI’s assertion of "inference supremacy" a "massive claim."
He emphasized the need for external benchmarks before accepting the hype. A key limitation noted is that Jalapeño’s published comparison was against Nvidia’s GB200 and GB300 systems, not the newer Vera Rubin generation.
OpenAI has stated it will continue using Nvidia and other accelerators, suggesting Jalapeño aims to diversify computing infrastructure rather than replace Nvidia hardware immediately.
Market Reaction
Broadcom closed at $356.74, down 0.56% on Tuesday. Its shares were up 0.54% to $358.68 in Wednesday’s premarket trading. According to Benzinga Edge Rankings, AVGO ranks in the 94th percentile for Quality, though its price trend remains negative across short, medium, and long terms.
How might Broadcom's successful partnership with OpenAI influence its future contract wins against Nvidia for other major AI labs?
What impact could the 2027 production ramp-up timeline have on Broadcom's near-term revenue projections and stock valuation?
Will the focus on inference-specific chips like Jalapeño accelerate a broader industry shift away from training-centric hardware architectures?

































