A new AI hardware startup, Majestic Labs, founded by three former Google and Meta silicon executives, has raised $100 million to develop next-generation AI servers that promise to dramatically expand memory capacity and reduce data centre costs for hyperscalers.
The company’s patent-pending silicon architecture claims to deliver up to 1,000 times the memory capacity of conventional enterprise servers. Designed to collapse multiple racks of hardware into a single unit, Majestic’s system targets the growing challenge of memory-intensive AI workloads, which often push traditional GPU-based setups to their limits.
Breaking the Compute-Memory Barrier
Co-founder and CEO Ofer Shacham said Majestic Labs isn’t trying to replace GPUs but to solve one of AI’s biggest constraints — the fixed compute-to-memory ratio. “We’re solving for memory-intensive AI workloads where that ratio becomes the bottleneck,” he told CNBC.
The company, founded in late 2023, has been operating in stealth mode with a team of fewer than 50 employees split between Los Altos, California, and Tel Aviv, Israel. Its architecture integrates compute and storage into a single high-capacity system that can replace up to ten standard racks, improving power efficiency, cooling, and space utilisation.
Majestic’s technology, the founders say, could deliver significant gains for sectors like finance, pharmaceuticals, and data analytics, where large-scale model training and retrieval operations demand higher memory bandwidth and density.
Funding and Industry Context
Majestic’s $71 million Series A round, led by Bow Wave Capital with participation from Lux Capital, closed in September 2025. The company expects to ship its first prototypes by 2027 and has already begun discussions with major hyperscalers for pre-orders.
The timing aligns with an ongoing global surge in data centre investments, as companies such as Alphabet, Meta, Microsoft, and Amazon collectively project over $380 billion in capital expenditure for AI infrastructure in 2025 alone.
Co-founders Sha Rabii and Masumi Reynders, who previously led chip development at Google and Meta, said their goal is to complement, not compete with, established players like NVIDIA. “NVIDIA has driven incredible AI innovation,” said Shacham. “Our mission is to extend that innovation by removing memory bottlenecks that limit scalability.”
A Decade-Long Collaboration Reimagined
The trio’s partnership dates back two decades, including their work together on Meta’s Facebook Agile Silicon Team (FAST) before regrouping after 2023 layoffs at Reality Labs. Majestic Labs represents a continuation of their shared vision — to reshape the core architecture of AI data infrastructure.
As AI adoption drives unprecedented compute demand, their approach to high-memory architecture could determine how efficiently the next wave of models scale — from training to inference.
