Former Google and Meta engineers have launched Majestic Labs AI, a startup built to break the “memory wall” that throttles ever‑larger AI models. The trio—Ofer Shacham, Masumi Reynders and Sha Rabii—draw on years of designing data‑center and mobile chips at Alphabet before spearheading Meta Reality Labs’ custom silicon team. Their new server promises to keep high‑end processors fully utilized in the market.

Majestic Labs AI disclosed in November that it closed a $100 million round backed by Bow Wave Capital, Lux Capital, Grove and other investors. The capital infusion will fund production of the server architecture designed to alleviate memory bottlenecks that force data centers to idle expensive GPUs. If successful, the hardware could reshape cost structures for AI workloads across cloud providers.

Industry analysts see the venture as a direct response to the growing gap between model size and server memory capacity, a mismatch that has slowed deployment of next‑gen generative AI. By integrating more memory per node, Majestic’s design aims to boost throughput without adding new chips. The company plans to ship its first units to select hyperscalers later this year.