Majestic Labs, an AI chip startup based in Los Altos, California, has successfully closed a $100 million Series A funding round to develop a groundbreaking memory-pooled server design. This new architecture is set to offer up to 100 terabytes of DRAM per accelerator, a significant leap beyond current high-bandwidth memory (HBM) capabilities. The company's co-founder and president, Sha Rabii, announced this ambitious project, which aims to consolidate the memory capacity and bandwidth of 10 racks of state-of-the-art GPUs into a single server.
What Happened
Founded in 2023 by industry veterans Masumi Reynders, Ofer Shacham, and Sha Rabii, Majestic Labs is poised to address the growing gap between compute capabilities and memory bandwidth. The company's innovative approach disaggregates memory from compute, allowing for independent scaling of memory. This architecture is particularly suited for AI inference tasks that are often limited by memory bandwidth. The startup's goal is to create a server that can handle the expanding size and complexity of AI models, providing the necessary memory bandwidth for efficient processing.
Majestic Labs is developing two key components: a memory interface chiplet and a many-core AI acceleration chip. The memory pool design will utilize over 100 terabytes of standard LPDDR in a server configuration alongside up to 12 of Majestic's AI accelerator chips. This setup allows each AI compute chip to access a single contiguous flat memory space, simplifying programming and ensuring consistent bandwidth and latency across the system.
Why It Matters for the AECM Industry
The advancement in AI server technology has significant implications for the Architecture, Engineering, Construction, and Manufacturing (AECM) industry. As AI models become more complex and data-intensive, the demand for high-performance computing solutions grows. Majestic Labs’ memory-pooled server design could drastically reduce the physical footprint and energy consumption of AI infrastructure, leading to lower operational costs and enhanced efficiency for AECM firms.
Furthermore, the potential to process large datasets more efficiently could accelerate the development of AI-driven design and analysis tools, improving project timelines and outcomes. By addressing the memory bandwidth bottleneck, Majestic Labs is setting the stage for more robust and scalable AI applications in the industry.
What's Next
Majestic Labs will continue to r
Is your firm ready for what’s next?
VisioneerIT helps AECM and government contractors modernize operations, achieve compliance, and implement AI.
Explore VisioneerIT Solutions →
