The Information reported June 28 that Baidu's chip arm Kunlunxin is targeting a $50 billion valuation for its Hong Kong listing, more than triple the figure sources cited just two weeks ago, and is asking investors to commit to buying its semiconductors as a condition of participating in the deal.
That last part is worth sitting with. Kunlunxin isn't just pitching itself as an investment; it's packaging capital and procurement into the same handshake. Prospective IPO investors are being asked to sign up as customers at the same time, an arrangement that, if it holds, turns the fundraise into a revenue pipeline before the company has printed a single public quarterly result. You don't see that often in a conventional IPO. What it signals is that Kunlunxin's commercial ambitions outside Baidu are newer and more uncertain than the $50 billion target implies.
The valuation itself has moved at a pace that should give anyone pause. As recently as mid-June, South China Morning Post reported that Kunlunxin was targeting roughly $14.7 billion. In May, TrendForce cited a figure of HK$100 billion, around $12.8 billion at current rates. Now The Information puts it at $50 billion, a jump of more than three times in about six weeks. That kind of revision usually reflects one of two things: either private investor appetite has run well ahead of the fundamentals, or the company is benchmarking itself against peers and working backward. Given what happened when Biren, a Chinese AI chip designer, debuted on the same exchange in January 2026 at $717 million raised and then surged 119% intraday, the appetite is clearly real. But Biren was a startup. Kunlunxin is asking to be valued like a platform.
Kunlunxin's current generation, the P800, is a third-generation XPU that Baidu unveiled at its developer conference in April 2025. At the time, Baidu was running a 30,000-chip cluster on it, and the P800 became the first domestically built chip capable of standalone deployment of DeepSeek V3 and R1 at full 671 billion parameter scale. That's a legitimate technical milestone, not a marketing line. A chip that can run China's most-watched open model without relying on Nvidia hardware is exactly the kind of proof point that moves procurement conversations inside Chinese cloud providers and state-linked enterprises. The next generation, the M100 for inference, is scheduled for 2026, followed by the M300 for training in 2027.
Baidu holds a 57.67% stake in Kunlunxin, which it spun off from an internal division that dates to 2012. The company confidentially filed with the Hong Kong Stock Exchange in January 2026, with CICC, Citic Securities, and Huatai Securities as lead underwriters, and has simultaneously launched STAR Market listing guidance in Shanghai for what would be a dual listing. The dual-track approach is deliberate: Hong Kong gives it access to international capital, and the STAR Market gives it the domestic institutional validation that Chinese defense and state enterprise procurement officers tend to weight heavily.
The context for all of this is US export controls. Washington's successive rounds of chip restrictions have cut Nvidia's H100 and H20 off from Chinese buyers, and while Nvidia has reportedly been exploring export-compliant alternatives like its Vera CPU, the fundamental constraint on Chinese AI infrastructure has not loosened. That makes Kunlunxin's timing intelligible. The export control regime has created a captive market inside China that didn't exist three years ago, and every major Chinese tech company is racing to own a piece of its supply chain. The IPO is Kunlunxin's bid to capitalize on that structural shift before the window narrows or the competitive field gets crowded.
The investor-as-customer structure is the sharpest signal of where the company actually stands. Baidu has been Kunlunxin's anchor customer since the chip division's earliest days. The transition to third-party revenue is still underway, and the company needs both capital and committed buyers at the same time. Bundling them makes commercial sense if you can pull it off. The risk is that investors who've pre-committed to chip purchases will have a conflict of interest as shareholders if those purchase agreements ever become unfavorable, and any future governance question about procurement terms could turn ugly. Don't expect that tension to appear in the prospectus, but it's there.
At $50 billion, Kunlunxin would land in the neighborhood of major US chip designers like Marvell Technology, which trades around that range, and well above where most Chinese semiconductor firms have been valued publicly. The question is whether the domestic tailwinds from export controls, plus a product roadmap that reaches the M300 training chip by 2027, justify that multiple for a company still proving it can sell to customers who aren't its parent. The IPO market will answer that, probably by the end of this year.
Also read: Google DeepMind is losing the researchers who built its best work and the damage is already showing • Suno's Spark artist grants come with a permanent silence clause that says more about the lawsuit than the music • AI coding agents are turning code review into the next startup risk