Kepler computing aims to revolutionize memory technology amid global shortage

Kepler Computing, a San Jose-based startup, has unveiled a new memory chip design to address the global shortage of high-bandwidth memory (HBM). Founded in 2018 by physicists and computer scientists, the company uses 3D stacking and proprietary materials to boost memory density without EUV lithography. Kepler has raised $468 million, including support from GlobalFoundries, Intel Capital, and Bill Gates, with the U.S. Department of Commerce pledging $245 million for U.S. development. The startup plans to ship HBM samples by 2026, scale production in Singapore by 2027, and start U.S. manufacturing in 2028. Kepler’s CEO, Debo Olaosebikan, noted AI growth has driven HBM demand, prompting simultaneous SRAM and HBM development. Srini Ananth of Intel Capital said market forces will determine Kepler’s technology’s role. GlobalFoundries’ Ed Kaste highlighted challenges in integrating iron-based materials, requiring dedicated equipment. Cofounder Sasi Manipatruni revealed the team tested 35 material composites before finding a viable solution. Despite progress, scaling remains a hurdle, with the company currently using around 2,000 wafers.