Xiaomi's Xring O3 enters mass production on TSMC 3nm for September foldable

In-house silicon remains a niche proposition, not a volume replacement
Xiaomi's custom chips still represent a tiny fraction of its total phone shipments, limiting their near-term impact.
Mark

Why is Xiaomi building its own chips when Qualcomm and MediaTek already dominate?

Mimi

Control and differentiation. By designing their own silicon, Xiaomi can optimize the chip specifically for their phones and services—especially AI features and their own software stack. It's also a hedge against supply chain disruption and pricing pressure from chip suppliers.

Mark

But the numbers seem small. One million devices total, with only 150,000 phones. That's not moving the needle.

Mimi

Not yet. This is still the early phase. The O3 is going into a foldable, which is a premium segment where smaller volumes are acceptable. The real question is whether they can eventually move this technology down into mid-range phones where volume actually lives.

Mark

What about the autonomous driving chip? That seems like a different business entirely.

Mimi

It is, but it's part of the same strategy. Xiaomi is signaling that they're not just a phone company anymore. They're building an ecosystem—phones, tablets, smart home devices, and now vehicles. Custom silicon lets them tie all of that together.

Mark

Given that Huawei owns 68 percent of the foldable market, what's Xiaomi's realistic path here?

Mimi

Slow and incremental. They're not going to dethrone Huawei overnight. But foldables are still early. If Xiaomi can prove their chip works well, costs less than competitors, and enables features others can't match, they can carve out a meaningful slice. The real test is whether they can do this profitably.

Mark

Is the investment worth it? Twenty billion yuan already spent, and profits are falling.

Mimi

That's the gamble. In the short term, it's a drag on earnings. But if they succeed in building a competitive custom chip ecosystem, it could become a significant competitive advantage and margin driver over the next five to ten years. The question is whether they can survive the investment phase long enough to see the payoff.

  • Xiaomi's Xring O3 — packing 24 billion transistors and a rebuilt AI engine — has entered mass production, marking the company's most serious push yet to compete on its own silicon rather than Qualcomm's.
  • The stakes are enormous: over $3 billion invested, a 3,000-person chip team, and a decade-long spending commitment — yet custom chips have so far reached only 150,000 phones out of 65 million shipped in the first half of 2026.
  • Two additional chips — the O100 AI accelerator and the D100 autonomous driving processor — signal that Xiaomi is no longer just a phone company placing a chip bet, but a hardware platform company reaching into vehicles and large language models.
  • Huawei commands 68 percent of China's foldable market, and Xiaomi's September launch of the 18 Fold will ship just 200,000–300,000 units — a proof of concept in a battlefield already claimed by a dominant rival.
  • Profit fell 42.6 percent in Q2 2026 as component costs rose and unit volumes declined, tightening the window in which Xiaomi must prove that premium, in-house silicon can justify its extraordinary cost.

In the long arc of technological self-determination, Xiaomi has crossed a threshold that few consumer electronics companies attempt: the design and mass production of its own silicon. With the Xring O3 now moving through TSMC's 3-nanometer foundry and two companion chips aimed at AI and autonomous vehicles, the Chinese company is staking billions on the belief that controlling one's own chips is the foundation of controlling one's own future — even as the market reminds it that ambition and scale are not the same thing.

Xiaomi is betting its future on its own chips. On August 24, the company unveiled the Xring O3, its second-generation custom processor, already in mass production at TSMC's 3-nanometer foundry. The chip is bound for the Xiaomi 18 Fold, a foldable phone expected to ship in September in initial volumes of 200,000 to 300,000 units.

The O3 is a meaningful leap over its predecessor, the Xring O1. It squeezes 24 billion transistors into a 133-square-millimeter die — a 26 percent density increase — and raises memory bandwidth by 48 percent. The most significant redesign, however, is in the AI section. Xiaomi rebuilt the neural processing unit around its own MiMo models, rating it at 200 TOPS and claiming a 45 percent performance gain. Additional AI accelerators were embedded into both the main and graphics processors.

Xiaomi also launched two chips beyond the phone world. The Xring O100 is a 6-nanometer AI accelerator for large language models, using 3D stacking to hit 1.22 terabytes per second of bandwidth. The Xring D100 — described as China's first 3-nanometer autonomous driving chip — pairs a 20-core CPU with a 16-core neural processor and can run AI models with up to 200 billion parameters directly in a vehicle. Both will be manufactured by TSMC.

Yet the ambition sits uneasily against the numbers. Since the O1 launched, only about 150,000 phones with Xiaomi's custom silicon have shipped — a sliver of the 65 million handsets the company sold in the first half of 2026. The Xring program has consumed more than 20 billion yuan so far, with plans to spend 50 billion over a decade. Meanwhile, Huawei holds 68 percent of China's foldable market, and global smartphone shipments are projected to fall 14 percent in 2026. Xiaomi's own Q2 profit dropped 42.6 percent as costs rose and volumes declined.

The September foldable launch is less a market assault than a demonstration — proof that Xiaomi can design world-class silicon. Whether that proof can eventually translate into scale, and whether the investment can be justified in a market dominated by entrenched rivals, is the question the company has committed billions to answering.

Xiaomi is betting heavily on its own silicon. On August 24, the Chinese tech company unveiled the Xring O3, its second-generation custom processor, which has already moved into mass production at TSMC's cutting-edge 3-nanometer foundry. The chip is destined for Xiaomi's September foldable phone—the Xiaomi 18 Fold—with initial shipments expected between 200,000 and 300,000 units, according to people familiar with the company's plans.

The O3 represents a significant engineering step forward from its predecessor, the Xring O1, which launched in May 2025. The new chip packs 24 billion transistors onto a 133-square-millimeter die, a 26 percent jump in transistor density. Memory bandwidth has climbed 48 percent to 113.8 gigabytes per second, and the processor can now work with LPDDR6, the latest generation of phone memory. But the real redesign happened in the artificial intelligence section. Xiaomi rebuilt the neural processing unit—the part of the chip that runs AI tasks locally on the phone rather than sending them to a server—around its own MiMo models. The company rates the new NPU at 200 TOPS, a standard measure of AI throughput, claiming a 45 percent performance gain over the O1. Additional AI accelerators have been grafted onto both the main processor and graphics processor, two on one and eight on the other.

Xiaomi did not stop at phones. The company simultaneously launched two additional chips designed for different markets. The Xring O100 is a 6-nanometer AI accelerator built for large language models, using 3D wafer-on-wafer stacking to achieve 1.22 terabytes per second of bandwidth. The Xring D100, described as China's first 3-nanometer processor purpose-built for autonomous driving, pairs a 20-core CPU with a 16-core neural processing unit and can support up to 160 gigabytes of unified memory—enough to run AI models with up to 200 billion parameters directly on the vehicle. TSMC will manufacture both chips as well.

Yet the scale of this ambition sits in tension with Xiaomi's actual market position. Since the O1 launched, cumulative shipments of devices running Xiaomi's custom chips have passed one million units across smartphones, tablets, and watches. Of those, roughly 150,000 were phones. Against Xiaomi's first-half 2026 shipment base of 65 million handsets, in-house silicon remains a niche proposition—a premium offering, not a volume replacement for Qualcomm or MediaTek. The company has invested more than 20 billion yuan, roughly $3 billion, into the Xring program so far, with plans to spend at least 50 billion yuan over a decade. The semiconductor design team has grown to more than 3,000 people.

The foldable market itself is crowded and well-defended. Huawei shipped 1.6 million foldable phones in China during the second quarter alone, capturing 68 percent of the market. Honor held 13.7 percent. Xiaomi is entering this space with a custom chip in a phone that will ship in relatively small numbers at a high price point. The broader smartphone market is also contracting—global shipments are expected to fall 14 percent in 2026, according to IDC—while Xiaomi's own profitability has been squeezed. Second-quarter profit fell 42.6 percent, driven by rising component costs. The company sold 65 million phones in the first half of 2026 at an average price of 1,329 yuan, down from 84 million units at 1,141 yuan a year earlier. The pattern is clear: fewer phones, higher prices, thinner margins.

The September foldable will be the first real test of whether Xiaomi's massive chip investment can translate into meaningful market share. At 200,000 to 300,000 units, it is a modest beginning—a proof of concept rather than a volume play. Whether the company can scale from there, and whether custom silicon can justify its cost in a market dominated by established players, remains an open question.

Xiaomi has invested more than 20 billion yuan into the Xring program and plans to spend at least 50 billion yuan over a decade
— Xiaomi's stated investment commitment
Möchten Sie die ganze Geschichte? Das Original lesen bei CRN Asia ↗
Kontakt FAQ