Xinxin Hangtu: China's State AI Fund Is Buying Into Startups Practically From Birth
China's state-backed AI industry development fund has become a shareholder in Xinxin Hangtu, a startup based in Suzhou. The company is less than a year old, but its focus on integrated circuit design and AI software made it a target for strategic investment. Its registered capital grew only slightly. What matters here is who invested, not how much. Beijing keeps building a closed semiconductor ecosystem by backing small teams that can produce alternatives to Western technology. The message is clear: China is willing to bet on risky early-stage companies if that helps close the gaps in its technology stack.
AI-processed from 36Kr (36氪); edited by Hamidun News
While the rest of the world watches quarterly earnings from giants like Nvidia, a real war of attrition over the future of semiconductors is playing out away from the headline numbers. Squeezed by sanctions and export controls, China has switched to "carpet investing" in young talent. The latest high-profile move, though not the biggest in dollar terms, came when the National AI Industry Development Fund took a stake in Xinxin Hangtu.
The Suzhou startup appeared on the radar just a year ago, in December 2023, and has already earned a "state seal of approval" in the form of direct investment. Context matters here: the National AI Fund is not another venture player chasing quick returns. It is an instrument of state policy, built to keep China's AI industry from suffocating without access to Western architectures.
When an entity like this invests in a company with registered capital of less than 20 million yuan, it signals that Xinxin Hangtu's technology stack has been deemed critical. The company specializes in integrated circuit design and software development, which today form the foundation of any AI product. Without homegrown chips, every Chinese neural network remains a layer on top of someone else's technology, one that Washington can switch off with the stroke of a pen.
Geography matters too. Suzhou has long been one of China's main high-tech hubs, competing with Shenzhen and Beijing. Backing Xinxin Hangtu in this particular region confirms that authorities want to diversify where development happens.
CEO Li Jun and his team now operate under close state scrutiny, and that brings them not only funding but also access to government contracts and infrastructure that ordinary startups can only dream of. In China, investments like this are often called "angel money with political weight." Why should we care?
Because China is methodically building a parallel reality in hardware. While Western companies optimize existing architectures, Chinese engineers are being forced by circumstance to find unconventional ways around the restrictions. Chip design and its practical application in software is exactly where China could make a leap.
If Xinxin Hangtu and companies like it manage to build efficient application-specific accelerators (ASICs) for narrow AI tasks, dependence on top-tier GPUs could decline. That won't happen tomorrow, but the groundwork is being laid right now, through targeted capital injections like this one. Ultimately, this is a long-term strategy in action.
Beijing isn't simply handing out money; it is building a network of hundreds of small companies that together form a sanctions-proof shield. Even if half of these startups fail, the rest will form the backbone of a new industry. Xinxin Hangtu is just one piece of this enormous puzzle, but the fact that a state fund is willing to work with companies at such an early stage says a lot.
This is a long game, and the stakes are an entire country's technological sovereignty. Bottom line: Time will tell whether Xinxin Hangtu becomes China's answer to Western chip design houses or remains just a line item in the fund's report, but Beijing clearly has no intention of skimping on the "brains" for its machines.
- Nov 1, 2026Voice bots built on LLMs turn to Jev, a fast classifier, for decisions in 200–500 ms
- Nov 1, 2026Local AI models agreed to pay scammers, while simple rules blocked all 42 schemes
- Oct 31, 2026A 22B-Parameter Video Model Runs on an RTX 3080 with 10 GB of VRAM
- Oct 30, 2026Mistral opens Large 4 weights: strong at code, but Chinese models score 32 to its 26
Build & Ship Production AI Agents Without Boilerplate Code
Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.