NextFin News - Xiaomi on Monday unveiled three new in-house chips - a flagship 3-nanometer mobile processor, an AI accelerator, and an autonomous-driving silicon - taking its most direct swing yet at the Qualcomm and MediaTek duopoly that supplies the Android world.
The Xring O3, built on a 3nm process and already in mass production, is rated by Xiaomi at 5.22 million points on the AnTuTu benchmark with support for LPDDR6 memory, and is due to power the Xiaomi 18 Fold and Pad 9 Pro Max when they launch in China in September. Alongside it came the Xring O100, a 6nm neural processing unit for on-device AI, and the Xring D100, a 3nm chip for intelligent driving slated for 2027.
The timing is the point. Xiaomi reported second-quarter revenue of 108.9 billion yuan ($16.2 billion) and adjusted net profit of 6.2 billion yuan - down 42.6% year over year - as memory costs squeezed the very smartphone business it is trying to reinvent. In the first half of 2026 the company sold 65 million handsets at an average price of 1,329 yuan, up 16.5% from a year earlier but with unit volumes down 23% from 84 million. In a global market expected to shrink 14% this year, the only lever left for margin is the component bill - and the chip is the largest line item a phone maker can bring in-house.
That is why this launch is more than an engineering trophy. It is the latest escalation in a structural shift: the surviving Android giants are being forced to internalize their most strategic component or slowly lose pricing power to the silicon suppliers and to Apple. The question is whether Xiaomi can execute where Oppo's MariSilicon effort collapsed, where Intel and Nvidia never gained traction, and where even Samsung still hedges with Qualcomm.
The Chips: What Xiaomi Actually Built
The headline number is the O3's claimed 5.22 million AnTuTu score, which Xiaomi says makes it the first mobile SoC to clear the five-million-point barrier. The company rates the 10-core chip at up to 60% higher performance than its predecessor, with a new 16-core G2-Ultra NX GPU delivering up to 85% stronger graphics at 64% better power efficiency. For AI workloads, Xiaomi puts the platform at 200 TOPS of tensor performance and 3.13 TFLOPS of vector performance - a 45% NPU improvement - with acceleration spread across the CPU, GPU, ISP, DPU, and ADSP rather than sitting in a single dedicated core.
Two details matter more than the benchmark. First, the O3 is Xiaomi's first mobile processor to support LPDDR6 memory, at up to 113.8GB/s of bandwidth - a generational jump that matters specifically for the on-device AI inference Xiaomi is betting its software roadmap on. Second, Xiaomi says it spent 459 days developing the O3, roughly 15 months for a flagship mobile SoC. That is a credible cadence for a company that has now committed to annual releases.
"This is our first chip product. Going forward, we should most likely release a yearly upgrade," Xiaomi President Lu Weibing said in March at the Mobile World Congress trade show in Barcelona.
That sentence, delivered months before Monday's launch, is what turns a prototype into a strategy. The first-generation Xring O1 - also built on TSMC's 3nm node, with 19 billion transistors and a 10-core CPU - posted Geekbench 6 scores above 3,100 single-core and 9,600 multi-core, broadly competitive with Qualcomm's premium silicon. The O3 is the second step on the cadence Lu promised.
The O100 is the more strategically revealing chip. It is a 6nm NPU built around near-memory computing, stacking the processor vertically against memory with millions of high-speed channels to reach 1.22Tbps of bandwidth. Its job is to run Xiaomi's MiMo large language model locally on phones, cars, and robots - keeping inference off cloud servers. The D100, a 3nm automotive chip with a 20-core CPU, 16-core NPU, and support for up to 160GB of unified memory, can reportedly run AI models of up to 200 billion parameters on the vehicle itself. Both have completed development; the O100 deploys next year, the D100 in 2027. TSMC is the manufacturing partner for the lineup - a relationship that now carries geopolitical weight as U.S. authorities press the foundry to curtail business with mainland China customers.
Why This Hurts the Merchant Silicon Vendors More Than One Lost Design Win
The obvious reading is that Xiaomi is swapping one supplier's chip for its own. The deeper threat is what it does to the pricing model. Qualcomm and MediaTek do not just sell silicon; they capture a slice of every Android phone's value through chip pricing and patent licensing, and that slice grows as phones get more expensive.
Xiaomi's own numbers show why the incentive is so strong. Selling 23% fewer handsets at a 16.5% higher average price means revenue per unit rose while volume fell - the classic premiumization trap. In a shrinking market, every flagship shipped with an O3 instead of a Snapdragon keeps margin inside the company rather than paying it out, and hands Xiaomi a stronger negotiating position for the chips it still buys.
The exposure is concentrated. Apple, Samsung, and Xiaomi each accounted for more than 10% of Qualcomm's FY2025 revenue, and China represents roughly 63% of the company's sales. Xiaomi is not Qualcomm's largest customer - Apple is - but it is the most visible bellwether in the Android camp. If the world's number-three phone maker can credibly insource its flagship silicon, the same logic spreads to Samsung, which already straddles both worlds with Exynos, and to the smaller Chinese brands that currently have no choice but to buy from Qualcomm or MediaTek.
There is also a timing asymmetry working against the incumbents. Qualcomm is already losing Apple as a modem customer after the iPhone maker began replacing Qualcomm modems with its own in-house chip in 2025. The company's handset segment is under pressure from memory supply constraints and softer demand - it guided third-quarter revenue to a range of $9.2 billion to $10 billion, short of the $10.19 billion analysts expected, even as its automotive business grew 38% year over year. Losing share in the Android flagship tier while Apple exits the modem business is the kind of two-front squeeze that compresses valuation multiples, not just revenue.
MediaTek's vulnerability is different but real. Its shares trade at a price-to-earnings ratio above 63, versus roughly 14 for Qualcomm, a valuation that prices in continued dependence of the Android ecosystem on merchant silicon. A credible in-house alternative from a top-three OEM undermines that assumption at the margin.
Xiaomi's shares, which closed at 29.02 Hong Kong dollars on August 21 - up 4.5% that day and part of a run that has lifted the stock more than 4% on its latest quarterly results - had already been repricing on the back of earnings before Monday's silicon news. The chip announcement adds a second narrative: that Xiaomi's 33.1 billion yuan of R&D spending in 2025, and a fresh five-year pledge of 200 billion yuan after already delivering about 105 billion on its previous pledge, is starting to produce proprietary hardware rather than just features.
The Counter-Thesis: Why Most In-House Chip Bets Fail
The strongest case against Xiaomi is history. Intel and Nvidia both failed to break into smartphone processors. Oppo shut down its Zeku chip-design unit in May 2023, barely two years after unveiling the MariSilicon X, as the economics turned against it. Even Samsung - with decades more silicon experience and its own foundry - still relies heavily on Qualcomm's chips in its flagships because of better power efficiency and cellular connectivity.
That last point is the hard one. A smartphone SoC is not just a CPU and a GPU. The modem - the component that actually connects to cellular networks - is arguably the hardest piece of silicon on the planet to design, and Qualcomm's entire moat is built on decades of baseband patents and field-proven radios. Xiaomi has not announced an in-house modem. The O3 will still need a 5G modem, and until Xiaomi integrates one, it remains a customer of the very companies it is trying to displace - just for a smaller slice of the bill of materials.
There is also execution risk in the numbers themselves. A 5.22 million AnTuTu score is a company-claimed benchmark, not an independent review. The O3's real test comes in September, when the Xiaomi 18 Fold ships and third parties measure sustained performance, thermals, and battery life against the Snapdragon 8 Elite Gen 5 and MediaTek's Dimensity 9500. Benchmark parity at launch has never been the same as efficiency parity after a year of software updates.
And there is a geopolitical wildcard. If U.S. pressure on TSMC tightens to the point where Xiaomi's access to 3nm capacity is restricted, the entire roadmap - O3, O100, D100 - stalls regardless of design quality. Samsung, with its own fabs, is insulated from that risk in a way Xiaomi is not.
The counter-thesis is serious, but it proves too much if stated absolutely. Oppo failed with a narrow imaging-chip strategy during a smartphone downturn, with no ecosystem to amortize the cost. Xiaomi is doing the opposite: it is spreading the chip program across phones, tablets, cars, and AI infrastructure, and it has committed 200 billion yuan over five years - about 40 billion yuan a year, more than its entire R&D spend in 2025. Failure is the base case for a one-product bet; it is not the base case for a funded, multi-generational program embedded in a hardware ecosystem that already spans phones, scooters, appliances, and electric vehicles.
The Verdict: Structural Shift, Cyclical Headwinds
This is a structural shift in the Android supply chain overlaid on a cyclical downturn in handset demand - and the two forces point in opposite directions in the near term.
Structurally, the direction is clear. As AI moves from the cloud to the device, the chip becomes the product. The company that owns the silicon owns the power envelope, the AI feature set, and the upgrade cadence. Xiaomi's "human-car-home" ecosystem - phones, the SU7 electric vehicle, and smart-home devices all running HyperOS - only works if the silicon layer is under its own control. The O100 running MiMo locally and the D100 in the car are not side projects; they are the connective tissue of that strategy. Once a phone maker crosses the threshold of designing its own application processor, the marginal cost of extending it across the ecosystem falls, and the logic becomes self-reinforcing.
Cyclically, the next 12 to 18 months are the danger zone. Global smartphone shipments are forecast to fall 14% in 2026, the steepest contraction on record, with a further 1.1% decline expected in 2027 before a 5.5% rebound in 2028. Xiaomi's own volume - down 23% year over year in the first half - shows the pressure. In that environment, any stumble on the O3's yield, thermals, or real-world efficiency gets magnified, and the company cannot afford a high-profile failure in its premium foldable, the very product meant to showcase the chip.
The market has not fully priced the structural risk to the merchant silicon vendors. Qualcomm's shares closed at $160.75 on August 21, down from a 52-week high near $260 reached in late May, and the stock is pricing a transition story - automotive and edge AI offsetting a softening handset business. But that transition assumes the handset floor holds. Xiaomi's move, if it scales to Samsung, says the floor is lower than modeled.
What to Watch
Three signals will tell you whether this is a real inflection point or another false start. First, independent reviews of the Xiaomi 18 Fold in September: if the O3 matches Snapdragon 8 Elite Gen 5 performance in sustained workloads at equal or better efficiency, the thesis is confirmed. Second, Xiaomi's modem roadmap - an announcement of an in-house 5G modem, or a multi-year modem supply deal with Qualcomm, will reveal how much of the bill of materials Xiaomi actually controls. Third, Samsung's response: if the world's largest Android maker accelerates its own Exynos roadmap in response, the merchant silicon duopoly is genuinely under siege; if it stays quiet, this remains a Xiaomi-specific story.
The falsifying signal is specific: if Xiaomi's flagship phones revert to buying Qualcomm's premium Snapdragon for the 2027 cycle, or if the O3 ships but is confined to a single mid-range device, the independence narrative is broken and the shares of Qualcomm and MediaTek should recover on the realization that Android insourcing has peaked.
Xiaomi is not trying to beat Qualcomm at being Qualcomm. It is trying to make the Qualcomm question irrelevant to its own products - and in a shrinking market, that is the only move that matters.
Explore more exclusive insights at nextfin.ai.
