Advertisement. Remove ads.
Enfabrica, a Silicon Valley chip startup backed by Nvidia (NVDA), on Tuesday launched a new hardware-and-software system designed to reduce the cost pressures in artificial intelligence data centers, particularly those tied to high-bandwidth memory (HBM).
While Nvidia and Advanced Micro Devices (AMD) dominate the market for cutting-edge AI processors, their chips require increasingly expensive HBM modules to function efficiently. Those memory chips are supplied by companies such as SK Hynix and Micron Technology (MU).
Enfabrica’s new system, called EMFASYS – Elastic Memory Fabric System – uses a special networking chip that is designed to connect the AI computing chips directly to their boxes with an older, cheaper memory chip, called DDR5.
Using the company’s software to route data back and forth between AI chips and large amounts of lower-cost memory, Enfabrica is hoping its chip will keep data center speeds up but costs down as tech companies ramp up chatbots and AI agents.
By addressing memory bottlenecks, the company says its system could lower the total cost of running AI workloads.
“AI Inference has a memory bandwidth-scaling problem and a memory margin-stacking problem,” said Rochan Sankar, CEO of Enfabrica. “As inference gets more agentic versus conversational, more retentive versus forgetful, the current ways of scaling memory access won’t hold.”
The system is not officially available for purchase yet, but according to the company, several customers are currently sampling and piloting it.
Micron’s stock was unaffected by the news, rising nearly 1% in pre-market trade. On Stocktwits, retail sentiment around the company dipped lower within ‘bearish’ territory over the past day.
Nvidia’s stock rose 1.35% in pre-market trade amid broader market strength. On Stocktwits, retail sentiment edged higher within the ‘bullish’ zone as compared to a day ago.
For updates and corrections, email newsroom[at]stocktwits[dot]com.