Volantis raises $97M to enhance AI memory with photonics
Summary
Volantis has announced significant advancements in AI memory technology, aiming to achieve ultra-fast inference speeds of up to 10,000 tokens per second per user on models exceeding 10 trillion parameters. The company is addressing the AI hardware bottleneck, where traditional AI chips face a trade-off between high-speed SRAM and higher-capacity but slower HBM memory. By utilizing photonics to connect over 220 memory chips, Volantis is positioned to enhance both memory capacity and bandwidth simultaneously, paving the way for faster, more efficient AI applications. With $97 million in funding from notable investors like Sam Altman and Jeff Dean, Volantis is poised to revolutionize AI inference and potentially reshape the industry.
Analysis
Volantis: Volantis is a semiconductor startup developing an optical interconnect technology called the optical fabric to overcome memory bottlenecks in AI accelerators. The company integrates custom micro-VCSELs and waveguides directly into chip packages to pool hundreds of memory chips with high bandwidth and low latency. Its approach targets ultra-fast inference for models exceeding 10 trillion parameters, with initial applications in rapid coding agents. Jeff Dean: Jeff Dean is a senior fellow at Google and leads research efforts in machine learning systems and infrastructure. He joined the investor group for Volantis's Series A, backing the effort to scale memory performance for frontier AI models. John Doerr: John Doerr is a partner at Kleiner Perkins and a longtime investor in technology and clean energy companies. He contributed to Volantis's $97 million raised to date, focusing on the startup's potential to advance AI hardware capabilities. Sam Altman: Sam Altman is the CEO of OpenAI and a prominent investor in AI infrastructure companies. He participated in Volantis's recent funding round as one of several high-profile backers supporting the company's photonic memory solutions. Dylan Patel: Dylan Patel is a semiconductor industry analyst and founder of SemiAnalysis. He participated in Volantis's funding round, aligning with his expertise in chip technologies and AI hardware trends. Lachy Groom: Lachy Groom is an investor who led Volantis's $88 million Series A alongside Abstract Ventures. His involvement highlights support for innovative approaches to AI memory architecture using photonics. AI Hardware Bottleneck: Current AI chips must trade off between high-speed but limited SRAM and higher-capacity but slower HBM, constraining performance on very large models. Photonics Scaling Potential: Optical interconnects offer a path to increase both memory capacity and bandwidth simultaneously by orders of magnitude across future chip generations.
Categories
techaimachine_learningai_agents