Suanmiao raises ~$280M across two rounds in a year for 3D AI chips
What's the deal? Chinese chip startup Suanmiao TechnologyDealroom has a profile for this one. Try Dealroom → has closed its Series A, led by Beijing state-backed investor Yizhuang GuotouDealroom has a profile for this one. Try Dealroom → alongside other municipal and district state capital platforms. The round takes its total raised over the past year to nearly US$275M (roughly $280 million) across a Pre-A2 round and the Series A.
What's the endgame? Suanmiao designs and mass-produces 3D-architecture AI chips for cloud computing. It will use the funds to expand its research team, advance its core technology, and move its self-developed 3D TokenPU inference chip toward volume production.
Why now? The company is targeting what it calls a "trillion-level large-model inference market." Its first-generation product, the A4E, taped out in June, stacking eight layers of memory wafers on a logic wafer via 3D hybrid bonding to reach 16TB/s memory bandwidth and bypass the "memory wall."
Suanmiao has raised repeatedly over the past year, drawing state platforms including CDB CapitalDealroom has a profile for this one. Try Dealroom → and Zhongguancun Science CityDealroom has a profile for this one. Try Dealroom →, market firms such as Source Code Capital and Jade Value CapitalDealroom has a profile for this one. Try Dealroom →, and industry funds including Legend Capital. Some backers have reinvested.
The money behind the machine: The team's core members spent nearly seven years in 3D IC work before founding the company, building a wafer-on-wafer hybrid bonding supply chain in mainland China and logging mass-production experience across tens of thousands of 3D IC wafers — with cumulative revenue topping US$165M.
Founder and chief executive officer Wang Fuquan framed the bet on efficiency. "The explosion of large-model inference is essentially an extreme test of compute energy efficiency," he said (translated from Chinese), adding that the company chose the 3D TokenPU architecture to "bypass the traditional path of general-purpose GPUs."
The signal: State capital leading back-to-back rounds underscores China's push for homegrown AI compute built on domestic process nodes and supply chains. As inference demand scales, chipmakers betting on energy efficiency over raw general-purpose power are drawing serious money.
Read more: sohu.com
Image credit: Generated with Gemini
