TeRAM today emerged from stealth with $37 million in equity financing from early-stage investors to tackle the memory wall facing AI deployment. The company is developing custom 3D SRAM for frontier AI applications, initially in data centers. As AI models scale beyond two trillion parameters, new memory technology is needed to increase capacity and throughput while reducing power consumption. TeRAM is validating its technology and targeting initial customer production in 2029.
“Memory is the constraint on AI progress, and incremental improvements will not be enough. Charlie saw this early. He is a rare founder who connects AI architecture, memory physics, and semiconductor manufacturing, and has assembled an extraordinary team to realize his vision. TeRAM is building the memory platform the next generation of AI requires,” said Brian Schechter, a partner at Primary Venture Partners and the lead investor in the seed round.
Under the AI scaling law, frontier models become more capable as parameter counts and context windows grow. Together with agentic workflows, these advances have accelerated AI adoption. Since mid-2025, agentic inference has overtaken model training as the main driver of AI compute and token generation. This has shifted the memory wall from HBM capacity during training to memory bandwidth during inference. Generating each output token requires a pass through the entire model and its trillions of parameters, making memory bandwidth a bottleneck for AI compute, whether the model resides in on-chip SRAM or off-chip HBM modules.
TeRAM is developing memory and packaging technologies to integrate 3D SRAM directly onto AI compute chips. Designed for demanding thermal conditions, these technologies aim to deliver unprecedented memory bandwidth at the lowest possible power consumption.
“I am grateful for Brian’s conviction in our mission and for all the investors supporting this groundbreaking roadmap,” said Charlie Cheng, CEO of TeRAM. “Developing a new memory technology is an ambitious undertaking. We are fortunate to have investors who back our vision with both capital and hands-on support.”
Primary, B Capital, Hyperion, and SemiAnalysis Capital co-led the round, joined by Alumni Ventures and Lightscape Partners, which closed on September 10, 2026. The consortium brings together deep insights and go-to-market reach in the AI ecosystem, as well as broad reaches into all corners of the capital market. The funding will enable TeRAM to validate its technology and scale the team to build the customized 3D SRAM memory chips for the initial datacenter customers.
TeRAM’s founders bring more than 80 years of combined experience developing semiconductor memory technology. The team has also collaborated closely on AI compute system-on-chip (SoC) projects with several leading hyperscalers and AI infrastructure companies. This combination of memory and compute expertise helps TeRAM understand the challenges facing future AI platforms and develop custom 3D SRAM to address each customer’s capacity, bandwidth, and power needs.
ABOUT TeRAM
TeRAM develops cutting-edge custom 3D SRAM for frontier AI deployments, from data centers to the smallest edge devices. The company tailors capacity, bandwidth, form factor, and packaging to minimize power consumption for each customer’s unique computing solution. TeRAM is headquartered in Los Altos, California. For more information, visit teram.com.
View source version on businesswire.com: https://www.businesswire.com/news/home/20260915533953/en/
Media gallery
