TeRAM has emerged from stealth with $37 million in equity financing to address the memory wall that increasingly limits artificial intelligence deployment. The company is developing custom three-dimensional SRAM for frontier AI applications, beginning with data center infrastructure. The funding will support technology validation and team expansion as TeRAM targets initial customer production in 2029.
The Memory Wall Challenge
As AI models scale beyond two trillion parameters, existing memory technology struggles to deliver sufficient capacity and throughput while controlling power consumption. The memory wall has become a central obstacle to scaling next generation AI systems. TeRAM is positioning its custom memory platform as a direct response to these constraints.
TeRAM's Technical Approach
TeRAM is developing memory and packaging technologies that integrate 3D SRAM directly onto AI compute chips. These technologies are engineered for demanding thermal conditions and aim to deliver high memory bandwidth at the lowest possible power consumption. TeRAM also tailors capacity, bandwidth, form factor, and packaging to meet the specific requirements of each customer's computing solution.
Validation and Production Timeline
TeRAM is currently validating its custom 3D SRAM technology with initial data center customers in mind. The company expects to scale its engineering and product teams to support these efforts over the coming years. Initial customer production is targeted for 2029, reflecting the complexity of introducing a new memory platform into AI infrastructure.
Shift Toward Agentic Inference
Since mid-2025, agentic inference has overtaken model training as the main driver of AI compute and token generation. This shift has moved the memory wall from high bandwidth memory capacity during training to memory bandwidth during inference. Because generating each output token requires a pass through the entire model, memory bandwidth has become a critical bottleneck for AI compute.
Funding and Investor Base
Primary, B Capital, Hyperion, and SemiAnalysis Capital co-led the round, with participation from Alumni Ventures and Lightscape Partners. The financing closed on September 10, 2026, and will enable TeRAM to validate its technology and scale its team. The investor group brings deep insights and go-to-market reach across the AI ecosystem.
Leadership Perspectives
Brian Schechter, a partner at Primary Venture Partners, described memory as the constraint on AI progress and said incremental improvements will not be enough. He highlighted TeRAM founder Charlie Cheng's ability to connect AI architecture, memory physics, and semiconductor manufacturing. Cheng expressed gratitude for the investor support and emphasized the ambitious nature of developing new memory technology.
Founding Expertise
TeRAM's founders bring more than 80 years of combined experience in semiconductor memory technology. The team has collaborated closely on AI compute system-on-chip projects with leading hyperscalers and AI infrastructure companies. This combination of memory and compute expertise helps TeRAM understand future AI platform challenges and develop customized 3D SRAM solutions.
TeRAM's emergence highlights the growing importance of advanced memory innovation in the AI sector. With $37 million in new equity and a clear technical roadmap, the company is positioned to address the bandwidth and power constraints facing frontier AI deployments. Its success could help unlock the next generation of AI infrastructure by removing one of the industry's most persistent bottlenecks.