Xiaomi Xring O3 Chip Rides TSMC 3nm Capacity, Not Its Own Fab

What Xiaomi announced

On August 24, Xiaomi unveiled the Xring O3, the second generation of its in-house smartphone processor. The company presented it as a way to get more control over key components and depend less on outside chip suppliers.33 Two people familiar with the matter told Reuters that TSMC would make the chip on its 3-nanometre process. Neither Xiaomi nor TSMC immediately confirmed the manufacturing details.33 Xiaomi said the O3 was already in mass production. It also said two more chips had finished development and would be deployed next year: the Xring O100, a neural processing unit, and the Xring D100, a chip for autonomous driving.32

Reports agree on the basic specifications. The O3 has more than 24 billion transistors on a die of about 133mm², a 10-core CPU, a 16-core GPU and support for LPDDR6 memory.43 Several outlets name the node as TSMC's N3P, an improved version of the N3E process used for the first-generation Xring O1. The chip launched in the Xiaomi 18 Fold and the Xiaomi Pad 9 Pro Max in September.5 Since then, the 18 Fold has topped AnTuTu's September ranking of Android flagships.4

Much of the coverage reads the story as Xiaomi moving into chipmaking. For anyone following fab construction and manufacturing investment, that reading is wrong. Xiaomi designs these chips but builds no plants. The physical manufacturing, and the capital it requires, sits entirely with TSMC.

A design company renting leading-edge capacity

The sums Xiaomi has put in are large by design-house standards. A source told Reuters the company has spent more than 20 billion yuan (about $3 billion) on Xring development, and its semiconductor design unit has grown to over 3,000 people from 2,500 a year earlier.31 Xiaomi has publicly pledged at least 50 billion yuan over a decade for the effort, which it restarted in 2021.31

None of that money builds a fab. TSMC has set its 2026 capital spending at $60 billion to $64 billion, with about 70% to 80% going to advanced process technology.23 That one year of TSMC investment is roughly nine times Xiaomi's full ten-year chip pledge, which shows where the manufacturing risk actually lies.

Techwire Asia made the point most clearly: Xiaomi's goal of reducing reliance on suppliers applies to Qualcomm and MediaTek, not to TSMC. Reuters' sources place all three new Xring chips at that one foundry.45 The coverage differs on how this was disclosed. One Reuters version credits a single unnamed source for the TSMC contracts on the O100 and D100.31 Another version says Xiaomi itself confirmed it had contracted TSMC for both chips.36 Either way, Xiaomi's whole roadmap, covering phones, on-device AI and cars, now depends on TSMC wafer lines.

Why the volumes barely register at the fab

The shipment targets are small. One Reuters source said the folding phone carrying the O3 is aimed at 200,000 to 300,000 units.33 Of the more than one million devices shipped with the O1, only about 150,000 were smartphones.33 Xiaomi sold 65 million handsets in the first half of 2026 alone.31

On the manufacturing side, the numbers are smaller still. One independent analysis estimated that 300,000 O3 chips would need roughly 860 wafers. That figure assumes about 350 good dies per wafer and 3nm wafer prices of $18,000 to $20,000, which works out to about $50 of fabrication cost per chip. The author labelled this as their own estimate, not a Xiaomi figure.47 By comparison, TSMC's 3nm capacity is expected to reach about 180,000 wafers a month by the end of 2026 and about 210,000 by mid-2027.25 On those estimates, a full O3 production run would use well under 1% of a single month of TSMC's 3nm output.

That is why a few hundred thousand chips can count as mass production for Xiaomi while making no visible difference to TSMC's expansion plans. It also explains how Xiaomi gets capacity at all. The Next Web noted that leading-edge wafer allocation is contested by companies with much bigger order books, and that TSMC has not ruled out further price rises.41 Small, steady orders are easy to fit in.

The node choice is a cost decision

The outlets agree that Xiaomi deliberately stayed on 3nm and passed on 2nm. In January, Digitimes reported that Xiaomi would skip TSMC's 2nm process to save money.14 Baidu's summary of the reporting gave three reasons: limited early 2nm capacity already reserved by Apple, NVIDIA and Qualcomm, the much higher cost of 2nm, and pressure on phone component costs from rising memory prices.11 As a result, Xiaomi's chip sits on an older node than its rivals. Apple's A20 Pro uses TSMC's 2nm N2, while the O3 is on N3P.8

The 3nm choice has a cost of its own. Techwire Asia pointed out that Xiaomi did not move to a denser process between generations. It got its performance gains mainly by making the chip bigger: the transistor count rose 26% and the die area about 22%.45 On a leading-edge process, a bigger die means fewer chips per wafer and more exposure to defects, so each unit costs more.45 Even the die size is disputed. Geekerwan measured the O3 at 138.3mm², larger than Xiaomi's stated figure, and the chip has no built-in modem. Despite the size, GSMArena found it more efficient than the Dimensity 9500, which is built on the same N3P process.

In practice, Xiaomi is trading die area for wafer price. That works at low volumes on a mature node. It would cost much more if Xring chips went into tens of millions of phones.

TSMC's plant build-out is the backdrop

The capacity Xiaomi is buying comes from one of the largest fab expansions in the industry's history. In April, TSMC chairman C.C. Wei said the company was building new 3nm lines in the Southern Taiwan Science Park, due to start mass production in the first half of 2027. TSMC is also converting 5nm capacity in Taiwan to 3nm.29 Construction of TSMC's second Arizona fab is finished, with 3nm commercial production planned for the second half of 2027. Its second fab in Kumamoto, Japan, will also use 3nm, starting in 2028.29 Japanese media reported the Kumamoto investment at about $17 billion, a figure TSMC declined to confirm.30

Wei has said TSMC's 3nm demand comes from smartphones, AI and high-performance computing, automotive and IoT customers.27 Xiaomi's lineup touches almost all of those: a phone chip, an AI accelerator and a 3nm driving processor. Wei has also said a foundry needs two to three years to build a fab and another one to two years to reach mass production.29 That timescale is the main reason a phone maker rents capacity instead of building it.

Geopolitics decides who gets the wafers

Access to TSMC is not open to every Chinese company. The Next Web noted that Xiaomi is not covered by the US restrictions that cut Huawei off from TSMC, so it can buy leading-edge Taiwanese capacity while Huawei relies on what Chinese foundries can make.41 The Decoding the Dragon newsletter goes further. It argues that Xiaomi's position on the right side of US export controls matters more than its design skills, because Huawei has been stuck on SMIC's older 7nm process since being cut off in 2020.47

This is the main manufacturing risk, and Xiaomi's spending does not address it. A Chinese company designing advanced chips does not mean China making them. BYD went the domestic route for its 4nm driving chip,41 while Xiaomi has put its future products on a foundry in Taiwan.

Where coverage diverges, and the verdict

The biggest disagreement in the coverage is about what the O3 is for. Benchmark-focused outlets present it as proof that Xiaomi can beat Qualcomm.4 More sceptical analysts call it a showcase chip that loses money at current volumes.47 Techwire Asia argues Xiaomi only recovers its investment if the same engineering serves phones, tablets, watches and cars,45 and that is the purpose of the O100 and D100. A 6nm AI chip and a 3nm driving chip, both going to TSMC, make Xring look less like a phone project and more like an automotive and AI silicon program.

The O3 is a capable chip, but in manufacturing terms it is a small customer on a booked-up 3nm network that TSMC is spending tens of billions to expand. Xiaomi's investment goes into engineers and design work, and fab capital stays with TSMC. The things to watch are not plant announcements from Beijing. They are whether the O3 spreads beyond foldables and tablets, whether the D100 reaches Xiaomi cars next year, and whether US policy keeps Xiaomi's access to TSMC open.