Xiaomi said on March 19 that it will invest at least RMB60 billion ($8.7 billion) in AI over the next three years, a day after Reuters reported that the anonymous Hunter Alpha model circulating on OpenRouter was in fact an early internal test of Xiaomi’s MiMo-V2-Pro. Xiaomi’s official MiMo page says the model has more than 1 trillion total parameters, 42 billion activated parameters and a 1 million-token context window. Taken together, those disclosures turn Xiaomi’s AI push into something more concrete than a spending slogan: the Chinese phone, EV and IoT giant is trying to turn its hardware ecosystem into an AI distribution platform powered by a serious in-house model stack.
The spending pledge matters more because Xiaomi already has a model to point to
Big AI investment promises are easy to overstate, especially when they come from companies that already know how to command attention at launch events. That is why Xiaomi’s March 19 number matters in this case. Reuters reported that Lei Jun said Xiaomi will invest at least RMB60 billion in AI over the next three years, while follow-up coverage from Chinese outlets added a second anchor: AI research and capital expenditure in 2026 alone are expected to exceed RMB16 billion. Those figures are large enough to matter, but the more important point is what Xiaomi says the money is for. According to reporting from Zaobao, Yicai and domestic event coverage, the plan spans foundation models, embodied intelligence and AI applications rather than one isolated lab effort.
That framing changes the story. If Xiaomi had announced only a large spending figure, the safer interpretation would have been that another Chinese consumer-technology company wanted to be seen as participating in the AI race. Instead, the pledge landed just after the company was forced into a different kind of disclosure: it had to explain that a mystery model already attracting developer attention was its own. The spending line therefore reads less like a vague future ambition and more like an attempt to scale a capability Xiaomi believes it already has.
It is still important to keep the caveat clear. Xiaomi has not already spent RMB60 billion. The number is a forward commitment, not a completed outlay, and investors or readers should not confuse the promise with audited realized expenditure. But in editorial terms, this is exactly why the Reuters headline worked. The money matters because there is now a model, a product identity and a broader strategy for the market to evaluate.
Hunter Alpha gave Xiaomi an unusual kind of proof
The Hunter Alpha part of the story is what makes the whole package travel beyond China’s domestic tech press. Reuters reported on March 18 that the model, which had generated buzz on OpenRouter and even led some outside observers to wonder whether it might be a DeepSeek successor, was actually tied to Xiaomi. Our earlier coverage, Xiaomi unmasks Hunter Alpha as MiMo-V2-Pro, bringing a hardware giant into the frontier agent race, captured that identity reveal and why it mattered for Xiaomi’s global AI profile. Xiaomi’s MiMo team and the official MiMo page then confirmed that Hunter Alpha had been an early internal test version of MiMo-V2-Pro. In other words, Xiaomi did not just unveil a model on stage and ask the market to trust a benchmark slide. It had a model that first drew attention under partial anonymity and only later got folded back into the company narrative.
That sequence matters because anonymous attention can function as a kind of market signal. It does not prove the model is category-defining, and it certainly does not replace third-party verification, but it does show that Xiaomi had something interesting enough to spark developer curiosity before the full corporate branding was attached. For a company better known internationally for smartphones, wearables and now electric vehicles, that is a meaningful shift in perception.
The official model details also give the story more weight than a generic “AI reveal” would have had. Xiaomi says MiMo-V2-Pro has more than 1 trillion total parameters, 42 billion active parameters and a 1 million-token context window, with particular emphasis on longer-context handling and agent stability. Those are the kinds of specifications that signal platform ambition rather than a lightweight feature model. They do not prove superiority over every rival model, and they should still be treated as company-disclosed claims unless independently validated. But they are specific enough to support a stronger argument: Xiaomi wants to be taken seriously as a builder of large-model infrastructure, not merely as a downstream integrator of other companies’ AI.
This is not another Xiaomi EV feature story
That distinction is critical because Xiaomi has already generated a different kind of headline this week in electric vehicles. Our earlier Xiaomi coverage focused on the refreshed SU7 and the logic of China’s EV specs war in Xiaomi Gives the SU7 Standard LiDAR at a Near-Flat Price: LiDAR, pricing discipline, safety positioning and margin pressure. This round is not that story. The better frame is that Xiaomi is trying to elevate AI from one more product feature into a company-wide operating layer.
The reason that matters is Xiaomi’s corporate structure. Xiaomi is not approaching AI from the same starting point as Alibaba or Tencent, whose AI narratives are more closely tied to cloud, platform services and enterprise software monetization. Xiaomi’s advantage, if it can turn it into one, is distribution across consumer endpoints. It already sells phones, tablets, wearables, home devices and electric vehicles. If the company can place a capable in-house model behind those touchpoints, then AI becomes less a stand-alone app and more a connective layer across the hardware environment where Xiaomi already has user reach.
That is why the article should not be reduced to “Xiaomi joins the AI race.” Xiaomi was already in the race in a loose sense, just as many large Chinese technology groups are. What changed this week is that the company made the race legible in corporate terms. It linked spending, model identity and ecosystem scope into one narrative. That is a different claim from saying a new car has better features or that a phone maker is experimenting with chatbots.
Xiaomi may be trying to build a consumer-tech AI distribution machine
The most useful analytical lens for international readers is that Xiaomi may be trying to become China’s consumer-tech AI distribution machine. That phrase matters because Xiaomi’s path is not a copy of the Alibaba or Tencent playbook. Alibaba can talk about cloud growth and enterprise AI demand. Tencent can talk about AI workloads, gaming, social traffic and cloud profitability. Xiaomi’s opportunity is more device-centric. If it can connect models to phones, cars, smart-home devices and agent-style workflows, then it can distribute AI through physical products people already use every day.
This is where Lei Jun’s spending language and the MiMo disclosure begin to reinforce each other. A giant capex-style pledge on its own might sound defensive, as if Xiaomi were simply trying to avoid being left behind by the rest of China’s AI wave. But pairing that pledge with a model that has already built some mystique gives Xiaomi a different kind of story. It can argue that AI is becoming a core layer in the same ecosystem logic that previously connected its phones, IoT devices and now EVs.
There is also a strategic implication for how Chinese AI competition is discussed in English. Much of that conversation has focused on cloud leaders, startup labs and geopolitics around chips. Xiaomi adds another category: the consumer-electronics and hardware company that wants to own not only endpoints, but also the model layer that mediates what those endpoints do. That does not make Xiaomi the next OpenAI or Google, and it would be a mistake to oversell the parallel. What it does suggest is that China’s AI contest is broadening beyond cloud-first incumbents and pure-model startups.
What changed, and what comes next
What changed this week is that Xiaomi turned scattered signals into a company-level AI thesis. On March 18, the company effectively acknowledged that the viral anonymous Hunter Alpha model was an early test of MiMo-V2-Pro. On March 19, Lei Jun put capital behind that reveal with a pledge to invest at least RMB60 billion in AI over the next three years and more than RMB16 billion in AI research and capex in 2026 alone. The combination gave the market a cleaner message: Xiaomi is not treating AI as a side project attached to phones or cars; it is trying to build a foundational AI capability that can sit across its broader hardware and software ecosystem.
What comes next is harder than making the announcement. Xiaomi now has to prove that MiMo-V2-Pro can become more than a technically interesting model page, that the company can translate large spending commitments into visible product integration, and that its AI push can create real leverage across devices, cars and agent-style experiences. The most careful conclusion is not that Xiaomi has already joined the top tier of global model companies. It is that Xiaomi has made itself harder to dismiss as a peripheral AI participant.
That nuance matters. The spending commitment is still a promise. The model’s strongest claims still rely heavily on Xiaomi’s own disclosures and early market buzz. And the commercialization path is still less defined than the announcement energy might suggest. Even so, the editorial takeaway is clear enough: this was not another Xiaomi EV configuration battle, and it was not just another headline about corporate AI spending. It was Xiaomi showing that it wants to turn an anonymous model hit into a company-wide AI strategy and, eventually, into a consumer-tech AI distribution platform with Chinese characteristics.
Sources
- Reuters — Xiaomi to invest at least $8.7 billion in AI over next three years, CEO says (2026-03-19)
- Reuters — Mystery AI model revealed to be Xiaomi’s following suspicions it was DeepSeek V4 (2026-03-18, updated 2026-03-19)
- Xiaomi MiMo — Xiaomi MiMo-V2-Pro (2026-03-18)
- Yicai Global — China’s Xiaomi to Invest Over USD8.7 Billion in AI in Next Three Years, CEO Says (2026-03-20)
- Zaobao — Xiaomi plans to invest at least RMB60 billion in AI over the next three years (2026-03-20)
Editorial caveats: Treat RMB60 billion as a forward-looking minimum commitment, not spending already completed. Treat MiMo-V2-Pro’s model positioning as partly company-disclosed and only partly market-tested; the anonymous Hunter Alpha attention is an interesting validation signal, not definitive independent proof of top-tier performance. Keep the core judgment focused on Xiaomi trying to turn its phone-plus-car-plus-IoT ecosystem into an AI distribution platform, rather than overstating the story as a direct challenge to every global frontier-model leader.
