Positron AI secures $875M for HBM alternative chip at $5B valuation
Positron AI has raised $875 million to develop Asimov, an inference chip using commodity LPDDR5X memory instead of high-bandwidth memory (HBM). The company claims its approach achieves over 90% memory bandwidth utilization, targeting a market where HBM remains scarce. The funding values the startup at $5 billion.
Sources: techtimes.com
“The raise underscores how startups are betting on memory innovation to sidestep HBM supply constraints in AI inference hardware.”
What it means
While HBM dominates high-end AI training, its scarcity and cost have pushed startups to explore alternatives. Positron’s bet on LPDDR5X could disrupt the inference market if it delivers on performance claims, but success hinges on overcoming thermal and bandwidth trade-offs. Competitors like Groq and Tenstorrent are also pursuing memory-efficient designs, signaling a broader shift toward cost-effective AI hardware.
Read the original reporting
This is a short summary. The outlets below did the reporting.
This brief was drafted automatically from the sources above and published under our editorial policy. Spotted an error? Tell us.