Here's a problem worth solving: AI models are getting smarter, but the infrastructure running them is increasingly creaky. Enter Majestic Labs, an AI infrastructure startup that just landed $100 million to reimagine how servers handle memory-intensive AI workloads.

The Series A round was led by Bow Wow Wave Capital, with participation from Lux Capital. But the real story here is what the founding team is building and why it matters.

Rethinking the Memory Problem

Majestic Labs was founded by Ofer Shacham and Masumi Reynders, both veterans of Google and Meta. Their pitch is straightforward: current AI infrastructure has a memory bottleneck, and they're building servers with 1,000 times the memory capacity of standard systems to fix it.

The co-founders told CNBC that Majestic develops its entire system from scratch, both hardware and software. When complete, each server could replace up to 10 traditional racks. That's not just a space saver—it's a complete rethinking of how AI infrastructure operates.

"Majestic is built on a simple and powerful insight: AI's next leap forward will come from access to more powerful AI infrastructure, and more powerful AI infrastructure requires a reimagination of the memory system," Shacham said in the company's funding announcement.

Co-founder Rabii expanded on the efficiency gains: "Majestic allows for a level of scalability and operational efficiency that simply isn't possible with traditional GPU-based systems. Our systems support vastly more users per server and shorten training time, lifting AI workloads to new heights both on-premises and in the cloud. Our customers benefit from tremendous improvements in performance, power consumption, and total cost of ownership."

The Team Behind the Technology

Shacham, Rabii, and Reynders have been quietly developing Majestic's servers since 2023. The three met while working at Google before all decamping to Meta in 2018. The idea for Majestic Labs emerged from brainstorming sessions about AI's biggest obstacles.

"We've been friends and colleagues for a long time, so this notion of working together and doing something exciting has always been in the periphery," Reynders told CNBC.

As they scale up, the co-founders plan to tap their extensive network to build out the team. With 1,500 former colleagues across companies like Meta and Google, they have a deep talent pool to draw from.

"There's that trust they already have with us," Rabii explained.

What's Next for Majestic

Beyond team building, the $100 million will fund further development of Majestic's software stack and launch a pilot program. The co-founders told CNBC they expect prototypes to reach customers as early as 2027.

The timing is significant. Major tech companies have collectively raised their capital expenditure guidance to $380 billion this year, and industry experts expect spending to balloon even more in 2026. The AI infrastructure buildout is happening at breakneck speed, and inefficiencies are becoming impossible to ignore.

"AI infrastructure is scaling at unprecedented speed, but the industry has not solved key fundamental architectural inefficiencies," Reynders said in the company's statement. "Majestic addresses this by delivering immediate operational gains on today's workloads while maintaining full programmability and flexibility to adapt as AI evolves beyond transformer-based models."

Lux Capital Partner Shahin Farshchi sees the bigger picture: "Majestic has engineered a new system for AI from the ground up, encompassing silicon, IO, packaging, and software, that is specifically tailored for the most advanced AI workloads. The team is making the most powerful AI accessible at an unprecedented scale, providing an opportunity to truly reshape how AI is delivered globally."

Whether Majestic can deliver on its ambitious vision remains to be seen, but with $100 million in backing and a team that's already built infrastructure for some of the world's largest tech companies, they're betting they can solve one of AI's thorniest problems: giving these increasingly powerful models the memory they need to run efficiently.