In September 2026, Kepler Computing, a seven-year-old U.S. memory chip startup, emerged from stealth mode, announcing a cumulative funding of $468 million. The U.S. Department of Commerce plans to provide up to $245 million in funding support under the CHIPS Act, with participation from renowned investors such as GlobalFoundries and Intel Capital. Kepler Computing focuses on developing new high-bandwidth memory chips for AI applications using ferroelectric materials and 3D stacking architectures. Its founding team, largely composed of alumni from Intel Labs, brings extensive experience in semiconductor materials, architectural design, and volume production. Originally, the company's technology focused on memory alternatives for the post-CMOS era, but after the release of ChatGPT, it added a parallel HBM (High Bandwidth Memory) development path. Kepler employs a hafnium oxide-based low-voltage ferroelectric composite material, enabling production on existing mature process lines with modifications, improving the memory bandwidth-to-power ratio and capacity while reducing power consumption. Currently, the company has established a mini-fab at GlobalFoundries' facility and successfully tape-out approximately 2,000 wafers. However, its ferroelectric memory technology still faces challenges in endurance, production line contamination isolation, interface ecosystem compatibility, and large-scale volume production. Kepler plans to ship its first HBM samples by the end of 2026, initiate volume production at GlobalFoundries' Singapore facility in 2027, and expand to U.S. production by 2028.