Majestic Labs just pulled off the kind of Series A that makes the AI infra world sit up straighter. 71M led by Bow Wave Capital, bringing total verified funding to 81M, and the ripple it creates is bigger than the number itself. When training clusters double every 5 months and datasets grow every 8, you eventually hit a wall. Not a metaphorical wall, a real one built from the limits of GPU memory. Majestic Labs looked at that wall, shrugged, and started carving a new doorway through it with 128 TB memory first servers that turn rack rooms into relics.
What makes this move hit harder is who is building it. Ofer Shacham spent years in Silicon Valley building Pixel Visual Core at Google, then ran Mark Zuckerberg’s silicon division out of Israel, all while teaching VLSI at Tel Aviv University. Shahriar Rabii stitched together silicon engines at Google and Meta after founding Arda Technologies and selling it to Google, then ran the Argos video chip team that helped drive YouTube into a new era. Masumi Reynders brought the product, legal, and partnership backbone that kept AR, VR, and custom silicon programs moving at Meta and Google, navigating everything from supply chain to advanced optics. This trio did not meet in a coffee shop with a slide deck. They earned their reputations by shipping hundreds of millions of devices and knowing exactly where the system breaks.
Majestic Labs set up base in a modest Hod Hasharon office, added Los Altos and Tel Aviv into the mix, and kept hiring until they hit just under 50 people split between Israel and the US. No noise, no chest beating, just engineering discipline mixed with a quiet confidence that feels almost unfair. While everyone else is trying to stretch GPUs past their limits, Majestic Labs is consolidating memory from 10+ racks into a single server and giving organizations the room to breathe again. That is why hyperscalers like AWS, Google Cloud, and Azure, along with finance and pharma giants, are already lining up for pilots.
Investors like Lux Capital, SBI, Upfront Ventures, Grove Ventures, Hetz Ventures, QP Ventures, Aidenlair Global, and TAL Ventures did not show up for theory. They backed a system level approach that complements Nvidia instead of fighting it, a design that treats memory bandwidth like the bloodstream of modern AI. Prototypes are expected in 2027, and pre-order conversations are already moving because the market finally sees a way to escape its own gravity.
Majestic Labs is scaling its team, refining its software stack, expanding its manufacturing pipeline, and working with early customers who want to build AI systems without tripping over memory ceilings. If the AI boom is a marathon disguised as a sprint, Majestic Labs is the runner who shows up without theatrics, fully aware that the real flex is finishing strong when everyone else burns out early.
Startups Startup Funding Venture Capital Series A AI Data Data Driven GPU Semiconductors Deep Tech Infrastructure Technology Innovation Tech Ecosystem Startup Ecosystem Hiring Tech Hiring
