Majestic Labs

Majestic Labs builds power-efficient AI servers designed to remove the memory constraints that limit GPU-based AI systems. Its flagship architecture combines custom accelerator chips with internally developed memory interface modules to deliver up to 128 terabytes of RAM in a single server - packing the memory capacity of roughly 10 racks into one machine. The systems support both training and inference for large language models and are intended to serve vastly more users per server than traditional equipment, in some cases eliminating entire racks of it.

The company designs its own silicon. Its custom accelerator chips were developed by former Google and Meta silicon engineering executives, and its memory interface modules are built in-house to enable the unusually high memory capacity per server. This represents a fundamentally different approach to AI infrastructure, one that competes directly with incumbent GPU-centric designs rather than extending them.

Majestic Labs was founded by former Google and Meta executives with deep silicon engineering expertise and operates as a team spanning Israel and California. It has raised an initial $100 million in funding. Target verticals include hyperscale data centres, financial services and pharmaceutical firms, alongside AI/ML infrastructure more broadly. The company's stated aims include democratising access to advanced AI capabilities and reducing the environmental impact of AI computing; its first systems are expected to reach customers starting in 2027.