A Stealth Startup Thinks It Just Hacked the Memory Shortage — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Kepler Computing emerged from seven years of stealth with a 3D-stacking architecture and proprietary ferroelectric material that it claims can increase HBM and SRAM density without relying on expensive extreme ultraviolet lithography.
- Kepler has raised $468 million from GlobalFoundries, Intel Capital, AMD Ventures, Baillie Gifford, and Bill Gates, and in July secured a US Department of Commerce commitment of up to $245 million to build a new class of high-performance AI memory in the US.
- GlobalFoundries converted a fab for Kepler's memory production in eight months versus a typical 24-month build, with Kepler running its "mini fabs" in Singapore and at GlobalFoundries' Burlington, Vermont facility.
- Kepler says its low-voltage composite material, developed after 35 iterations, enables SRAM density comparable to 2- or 3-nanometer chips; cofounder and CTO Sasi Manipatruni said the team solved the underlying physics problem before settling on the material class.
- Kepler plans to ship its first HBM chip samples later this year, ramp production out of Singapore next year, and begin US chip production in 2028 — having run its technology on roughly 2,000 wafers to date.
- GlobalFoundries senior VP Ed Kaste flagged iron contamination as a key manufacturing risk, noting Kepler's material system must run on dedicated equipment or be fully encapsulated so it cannot escape into the production flow.
Why it matters: Kepler's bet to retrofit existing fabs with new materials directly challenges the multibillion-dollar buildout that SK Hynix and Micron are racing to fund for HBM. With only 2,000 wafers of validation, iron contamination flagged by its own manufacturing partner, and US production not slated until 2028, Kepler has years of work before it meaningfully eases the shortage the headlines claim it has already hacked.
Ask SkimNews
