A new challenger has entered the high-stakes arena of artificial intelligence infrastructure, armed with substantial funding and a plan to address a critical bottleneck holding back the industry. Majestic Labs, a startup founded by former senior chip architects from Google and Meta, has emerged from stealth operations with a combined $190 million in seed and Series A funding.
This significant financial backing, led by Lux Capital and Bow Wave Capital with participation from a global consortium of investors, signals strong belief in the company’s ambitious mission to redefine the hardware underpinning advanced AI model training.
The core issue Majestic Labs aims to solve is the severe memory limitation inherent in current AI systems. As large language models grow to encompass trillions of parameters, the industry is hitting a wall.
Today’s most powerful graphics processors, including those from market leader Nvidia, are constrained by a maximum of 192 gigabytes of memory per chip. This limitation forces companies to purchase vast arrays of expensive processors simply to have enough active memory to run complex computations, a phenomenon that some industry observers liken to a modern-day “Altman’s Law,” where AI processing costs escalate exponentially with user growth.
This creates not only a financial strain but also a physical one, as data centers grapple with immense heat output that further throttles performance.
Rather than designing another graphics processing unit to compete directly, Majestic Labs is pursuing a more radical architectural approach. The company has engineered a completely new server design that replaces the standard Nvidia AI server. This system features a proprietary dedicated processor and a novel structure that connects each GPU to vast pools of memory.
According to the company, this innovation increases memory accessibility by a factor of up to one thousand times compared to existing solutions. The practical result is a claimed 50-fold improvement in computing performance, potentially allowing a single Majestic Labs server to replace clusters of ten to twenty traditional Nvidia servers, offering unprecedented efficiency and cost savings.
The vision is being driven by a deeply experienced founding team with unparalleled industry credentials. CEO Ofer Schacham, previously the head of AI processors for Google’s edge devices and the leader Meta hired to establish its chip lab for the metaverse, is now steering his own venture. He co-founded Majestic Labs with former Meta and Google colleagues Sha Rabii and Masumi Reynders.
Their combined expertise, backed by numerous doctorates and hundreds of patents, provides a formidable foundation. This deep industry knowledge allows them to engage with potential customers on a strategic level, developing server solutions today that are projected to meet the anticipated demands of the AI market in 2027 and beyond. In an industry racing to overcome its most fundamental constraints, Majestic Labs is positioning its memory-centric architecture as a pivotal leap forward.
