Micronano Core raises $300M as AI chip startup sells services before silicon ships
What's the deal? Chinese AI chip startup Micronano Core has raised US$275M (roughly $300 million) over six months, including more than US$275M in the past two months, the company said on October 28. The funds will back development, tape-out, and mass-production preparation for its three-dimensional compute-in-memory AI inference chips.
What's the endgame? Micronano Core is betting on compute-in-memory architecture to ease the memory-bandwidth and energy bottlenecks that slow AI inference. It aims to serve phones, PCs, and data centres.
Why now? Its first three chips have all been taped out, with customer sampling and volume supply expected in early 2027. That leaves a roughly two-year gap in which device makers must adapt algorithms and deploy models before silicon arrives.
The company says it has completed multiple rounds of simulator validation with phone platforms, original design manufacturers, and embodied-intelligence firms, with some projects in deep joint debugging. That pre-launch work has opened an asset-light business: teams skilled in model deployment and chip adaptation can sell migration and performance-tuning services to end-device makers, charging per project.
What could go wrong? The chips remain unproven. No product has reached mass production, and the raise reflects capital's bet on inference compute and a domestic chip roadmap rather than validated commercial success. The metrics that matter — test performance, customer adoption, and actual delivery — are still ahead.
The signal: Large cheques for pre-revenue, pre-silicon chip startups show how far investors are willing to run ahead of proof, wagering on both AI inference demand and China's push for homegrown chips.
Read more: sohu.com
Image credit: Generated with Gemini
