Analysis: OpenAI's "Jalapeño" AI Chip Expansion May Be Affected by Samsung's HBM4 Supply Constraint

BlockBeats News, August 26th, Citrini analyst Jukan published an analysis stating that OpenAI's "Jalapeño" AI chip may not be able to reach the level of NVIDIA's Rubin platform in terms of scale-out. He believes that if OpenAI's HBM4 high-bandwidth memory is fully dependent on Samsung's supply, the production capacity ramp-up will face a certain limit.

Jukan further pointed out that if OpenAI hopes to further expand the deployment scale of "Jalapeño", it may need to reduce the requirements for HBM4 transmission speed and other specifications, in order to simultaneously adopt HBM4 products from other suppliers to alleviate supply chain constraints.

BlockBeats reported earlier today that OpenAI has released the test results of its self-developed AI inference chip Jalapeño, stating that the chip outperformed NVIDIA's GB200 and GB300 super chips in the InferenceX benchmark test. In models such as GPT-OSS 120B, DeepSeek R1, and Kimi K2.5 1T, the AI workload per watt reached 1.5 to 1.9 times that of competitors, and the end-to-end latency was reduced by 1.7 to 3.6 times.