Xiaomi has moved its chip program from a promised next step to a three-part product roadmap. On Aug. 24, the company unveiled the Xuanjie O3 smartphone processor, the O100 AI accelerator, and the D100 processor for autonomous-driving workloads. The trio is notable not simply because Xiaomi designed the chips itself, but because it places mobile devices, localized AI compute, and vehicles inside one semiconductor strategy. As Caixin reported, the launch is part of Xiaomi’s effort to build more control over a technology stack that has long depended on outside suppliers.
That is a meaningful change from the posture Xiaomi described only days ago, when it said that its next self-developed chip was near. The company can now point to actual processors with differentiated roles rather than a broad ambition. The test will be whether it can turn that portfolio into durable products, usable software tools, and volumes large enough to justify the investment.
Xiaomi’s Three-Chip Design Separates Mobile, AI, and Driving Tasks
The O3 is the centerpiece. Caixin describes it as a 3-nanometer flagship smartphone system-on-chip with 24 billion transistors, fabricated on TSMC’s process technology. Caixin says the initial O3 device will be Xiaomi’s 18 Fold, due in September. That schedule matters because a foldable handset is a limited-volume product compared with Xiaomi’s broader phone range. It gives the company a high-end setting in which to test the chip, its thermal behavior, software compatibility, imaging pipeline, modem integration, and AI features before making a larger commitment.
The O100 takes the strategy in a different direction. Rather than treating AI as one function folded into a phone processor, Xiaomi is presenting a dedicated accelerator for larger workloads. Caixin says the O100 is scheduled to reach the market in 2027. The D100 is also expected in 2027 and is intended for autonomous-driving applications. Together, the chips describe a company seeking to place its own silicon at several points where AI is shifting device design: on the handset, at the edge, and inside the vehicle.
There is no guarantee that vertical integration will lower Xiaomi’s costs quickly. Custom chips require years of design work, fabrication access, software engineering, validation, and a product pipeline large enough to spread those costs. Caixin puts Xiaomi’s R&D outlay above 21 billion yuan and describes a silicon organization approaching 3,000 employees. Those figures illustrate why only a handful of Chinese consumer-technology groups can attempt this kind of program at scale.
Domestic Design Still Depends on a Global Manufacturing Chain
The launch is a reminder that semiconductor autonomy is not a binary condition. Xiaomi designed the O3, but Caixin says it uses TSMC’s 3-nanometer process. That combination is important. It demonstrates that Chinese companies can deepen their role in architecture, system integration, and AI optimization while continuing to rely on an international fabrication network for leading-edge production.
The distinction is especially relevant as China’s consumer-electronics groups try to make AI features more immediate and less dependent on remote data centers. An on-device processor can help with responsiveness, privacy choices, and power management, but the experience still depends on models, memory bandwidth, operating systems, developer tools, and app partnerships. Xiaomi’s existing hardware reach gives it an opportunity to test that chain across phones, cars, and connected devices.
The company is not the only Chinese business trying to turn AI hardware into a consumer proposition. The recent RayNeo iO test of China’s AI-glasses market showed how much the commercial question extends beyond model quality. Devices must be comfortable, reliable, affordable, and useful often enough to change everyday behavior. Xiaomi faces a parallel challenge: a chip launch is strategically significant, but customers will ultimately judge faster applications, better battery life, camera performance, driving assistance, and the consistency of AI features.
The Commercial Question Is Bigger Than a Benchmark Score
Xiaomi has a reason to be cautious about how quickly a new chip can reshape its supplier relationships. Caixin cites analysts who question the immediate commercial impact of its in-house program, particularly when early devices are produced at limited volumes. The company’s earlier O1 processor exceeded one million shipments, according to the report, but the handset volumes attached to it were much smaller. That gap captures the central business problem: chip shipments alone do not establish a broad consumer platform.
The 2027 timing for the O100 and D100 also suggests that Xiaomi is sequencing the program rather than claiming an instant replacement for established chipmakers. The company still needs to prove that a standalone accelerator can support attractive AI services and that a vehicle processor can meet the demanding requirements of automotive safety, power efficiency, software maintenance, and supply assurance.
For China’s AI industry, the broader lesson is that the competitive field is widening. Leading models remain essential, but the companies that distribute them through devices, cars, and edge systems will increasingly determine where AI is useful and who captures value. Xiaomi’s three-chip announcement does not settle that contest. It does, however, make clear that the company wants a larger role in deciding how Chinese AI reaches users beyond the cloud.