Kepler Computing emerged from stealth to address the critical AI hardware memory bottleneck. The startup uses proprietary ferroelectric material and 3D stacking to increase memory density. This technology scales High-Bandwidth Memory (HBM) and SRAM without relying on complex EUV lithography. Kepler aims to deliver memory with SRAM-level performance and capacity exceeding current HBM.
Surging AI demand has created a severe HBM shortage with supply sold out for years. AMD Ventures is an investor in the company. Intel is a collaborator. Kepler is partnering with GlobalFoundries for high-volume manufacturing.
The company plans to ship initial samples by late 2026. Full production will ramp up in 2027 to potentially reshape the AI chip supply chain.