Silicon Separatism: OpenAI Releases Jalapeño Chip, While IBM Breaks the 1nm Barrier

The monopoly on graphics processing units has cracked. On June 25, 2026, OpenAI, in collaboration with Broadcom, officially unveiled `Jalapeño`—a custom chip optimized exclusively for the inference of large language models. This is a direct strike against NVIDIA’s position in the B2B market: the creators of ChatGPT no longer want to overpay for universal GPUs, shifting to custom ASIC solutions that radically reduce the cost of token generation.

Synchronously, IBM Research announced it had overcome a critical physical barrier in lithography, demonstrating a sub-nanometer (0.7 nm) process node for AI computing. This is a technological breakthrough that will allow corporations to pack significantly more compute cores onto a single die with reduced power consumption. The industry is attacking on two fronts simultaneously: OpenAI is optimizing chip architecture for specific LLM math, while IBM is physically shrinking transistors, pushing the death of Moore’s Law back another decade.

Source: OpenAI / Broadcom | IBM Research / Reuters