--- title: "Xiaomi (Trans): Memory Price Upcycle Lasts Longer Than Expected; AI Monetization Still Early---" type: "Topics" locale: "en" url: "https://longbridge.com/en/topics/39486168.md" description: "Below are Dolphin Research's notes from Xiaomi Corp.'s 2025 Q4 earnings call. For our take, see 'Xiaomi: From heaven to hell—what sustains the faith?'.I. $XIAOMI-W(01810.HK) core takeaways. Shareholder returns: ~HKD 6.3bn of open-market buybacks in 2025; ~HKD 4.7bn repurchased YTD 2026; an automatic buyback program announced in Jan 2026 with a cap of HKD 2.5bn.Guidance: 2026 auto deliveries target of 550k units. Planned 2026 R&D spend of over RMB 40bn..." datetime: "2026-03-24T15:08:17.000Z" locales: - [en](https://longbridge.com/en/topics/39486168.md) - [zh-CN](https://longbridge.com/zh-CN/topics/39486168.md) - [zh-HK](https://longbridge.com/zh-HK/topics/39486168.md) author: "[Dolphin Research](https://longbridge.com/en/news/dolphin.md)" --- > Supported Languages: [简体中文](https://longbridge.com/zh-CN/topics/39486168.md) | [繁體中文](https://longbridge.com/zh-HK/topics/39486168.md) # Xiaomi (Trans): Memory Price Upcycle Lasts Longer Than Expected; AI Monetization Still Early--- **Below is Dolphin Research's transcript of Xiaomi Corp.'s Q4 2025 earnings call. For the earnings take, see '**[**Xiaomi: From Peak to Trough — What Still Anchors Conviction?**](https://longbridge.cn/en/topics/39484676?channel=SH000001&invite-code=294324&app_id=longbridge&utm_source=longbridge_app_share&locale=zh-CN&share_track_id=f9eec4a2-fe15-4e37-b3df-7358b0f36c8b)**'.** **I.** $XIAOMI-W(01810.HK) **Key Takeaways** 1\. **Shareholder returns**: Repurchased approx. HKD 6.3 bn in 2025; from early 2026 to date, repurchased approx. HKD 4.7 bn. Announced an automatic buyback plan in Jan 2026 with a cap of HKD 2.5 bn. 2\. **Guidance**: **2026 auto delivery target of 550k units; 2026 R&D spend plan of over RMB 40 bn, with more than RMB 16 bn in AI and embodied intelligence**. AI investment of RMB 60 bn over the next three years. 3\. **Key financials**: Full-year revenue of RMB 457.3 bn (+25% YoY), a record high; adj. net profit of RMB 39.2 bn (+44% YoY). Blended GPM at 22.3% (+130 bps YoY), an all-time high; Smart EV and AI innovation segment delivered its first full-year positive OP (RMB 0.9 bn). 4\. **CapEx**: **2025 CapEx reached RMB 18.2 bn (+73% YoY), with smart EV and AI innovation accounting for over 66%**. Cumulative R&D spend over the past five years was RMB 105.5 bn, with more than RMB 200 bn expected over the next five years. **II.** $Xiaomi Corporation(XIACY.US) **Call Details** **2.1 Management Highlights** 1\. **Smartphones** a. Global shipments of 165 mn units, top-three globally for 22 consecutive quarters with a 13.3% share, and a top-three position for five straight years. b. Mainland China premium mix (RMB 4,000+) reached 27.1% (+3.8 pp); 6,000–10,000 RMB band share at 4.5%, nearly doubling YoY. c. Smartphone ranking in Mainland rose to No. 2; shipments ranked No. 2 in LatAm and SE Asia, and No. 3 in Europe and Africa. d. Launched the first Leitzphone for global markets in Feb 2026, priced at EUR 1,999, marking a new milestone in the overseas premium push. e. Despite higher memory prices in 2025, full-year smartphone GPM held at a relatively healthy 10.9%. 2\. **IoT** a. Full-year revenue of RMB 123.2 bn (+18.3% YoY), surpassing RMB 120 bn for the first time, with record highs at home and abroad. b. GPM of 23.1% (+280 bps YoY), a record high. c. Major appliances revenue grew over 23%, a record; wearables (bands) ranked No. 1 globally, TWS earbuds No. 2; tablet shipments rose 25.2%, ranking No. 5 globally. d. AI glasses launched in Jun 2025, ranking No. 3 globally and No. 1 in China. e. Tech-driven home appliances entered Europe, now covering Spain, France, Germany and Italy. 3\. **Internet services** a. Full-year revenue of RMB 37.4 bn (+9.7% YoY), a record; Q4 revenue was RMB 9.9 bn. b. Global MAU reached 750 mn (+7.4% YoY); Mainland MAU 190 mn (+10.1% YoY). c. Advertising revenue of RMB 28.5 bn, a record; overseas internet revenue of RMB 12.6 bn (+15.2%), accounting for 33.8%. d. Full-year internet GPM at 76.5%, stable. 4\. **Smart EV** a. 2025 deliveries reached 411,082 units, well above the initial 300k target; Q4 deliveries were 145,115 units, a quarterly record. b. Avg. post-tax ASP of RMB 251,171 (+7% YoY). c. As of Feb 13, 2026, cumulative deliveries exceeded 600k; 2026 delivery target is 550k. d. The new-gen SU7 launched on Mar 19, with over 15k locked orders in 34 minutes and over 30k in three days. e. Vision GT concept car shown at MWC, the first time a Chinese brand was invited. 5\. **AI and core tech** a. Foundation models entered the global open-source leading cohort: launched MiMo-V2-Pro (over 1 tn parameters, supporting a 1 mn context window), MiMo-V2-Omni, and MiMo-V2-TTS. b. MiMo-V2-Pro ranked No. 8 globally and No. 5 on brand dimension in Artificial Analysis' LLM Intelligence Index. c. In Mar 2026, the phone AI agent 'Xiaomi miclaw' entered limited testing, the first attempt by an OEM to deploy such an agent on smartphones. d. The new SU7 is equipped with the XLA cognitive foundation model, enhancing parking, navigation, complex understanding, and voice control. e. In Nov 2025, launched the Miloco smart home solution, giving smart homes 'eyes, brain, hands, and feet' for the first time. 6\. **Embodied intelligence/robotics** a. In 2026, released a tactile-driven precision grasping fine-tuning model and open-sourced the Xiaomi-Robotics-0 VLA foundation model, achieving multiple SOTA results. b. Xiaomi's embodied robots began internships in auto plants; success rate of synchronous nut installation reached 90.2%, fastest takt time of 76 seconds, achieving three hours of autonomous operation. c. A large number of embodied robots are expected to work in Xiaomi factories in the next five years. **2.2 Q&A** **Q: Memory prices keep rising, and peers have begun to raise prices. What advantages does Xiaomi have in supply-chain management and inventories, and how will you address this year's challenges?** A: **On memory inflation, this has been a recurring topic each quarter. I previously said this is cyclical and we need to look out to 2027**. That said, **this upcycle may last longer and run higher than I expected**, with AI-driven demand pushing memory prices up and cost pressure exceeding prior assumptions across the industry. This will significantly impact all consumer devices, not just smartphones. Some smaller-capacity categories are even seeing supply cuts. The impact varies by memory's share in COGS — the higher the share, the greater the impact. **In our portfolio, smartphones, tablets and laptops are most affected, while memory accounts for a relatively smaller share in premium smartphones**. **We have already seen some competitors raise mid-range smartphone prices in the past two weeks**, which I fully understand. For any OEM, it is hard to absorb such hikes without passing costs to consumers. We face significant pressure too, and will try our best to absorb costs to protect consumers; **when that becomes untenable, we will have to adjust prices and hope for understanding. We may move slower on price hikes, but that does not mean we are immune**. Xiaomi has several advantages: **first, home appliances are less affected, and smart EVs are impacted but contribute less to COGS than phones**, so a diversified mix helps spread risk. Second, as a global leader in smartphones, tablets, laptops and smart EVs, **we have strong long-term partnerships and trust with global memory suppliers, backed by long-term supply contracts, with no current risk of supply disruption**, which is an edge vs. rivals. Third, **having held a more cautious view on memory inflation, we built inventories more aggressively, resulting in higher coverage**. Overall, though, end-product costs are meaningfully affected, and near-term pressure is a fact. **Q: How are sales trending after the new-gen Xiaomi car launch? Why disclose locked-order data instead of reservations?** A: From the user side, we see several signals. First, **we had a strong start — over 15k locked orders in 34 minutes and over 30k in three days**. We also delivered on the promise to start deliveries from day four post-launch, as we prepped manufacturing early and iterated on lessons from the first-gen where users waited too long. As for why we disclose locked orders: we believe it is fairer. **Locked orders are not reservations or large deposits; they are firm orders that will lead to delivery and directly inform our manufacturing cycles**. Practices vary across the industry, but we think this is the right approach. Some structural details on locked orders: investors care whether these are upgrades from existing owners or new customers — **the vast majority are new customers, not first-gen upgrades. Around 50% are first-gen users**, and about 60% of new buyers are iPhone users. Locking has progressed faster than the first-gen with larger volumes, and about 60% of buyers chose paid options, a higher penetration than the previous generation. Female buyer mix, iPhone user penetration, and color options are all better than last time. **Q: How has Xiaomi's foundation model capability evolved over the past two years? What is miclaw's positioning and its relationship with IoT? How do you view AI across users, developers, internal use and monetization? What KPIs does the foundation model team have? How will the RMB 60 bn over three years split between OpEx and CapEx?** A: In 2023, we spent substantial effort defining our AI strategy. We started from the AI infrastructure — algorithms and other foundations — on which applications could scale exponentially, and last year we already said 2026 would be the breakout year for AI applications. AI is moving from virtual to physical, consistent with my earlier view. We began investing in foundation models in 2024 and made major progress in 2025. On agents, we focus on how AI can achieve more on a single device, enabling users to accomplish things previously not possible. Through open claw, we rapidly rolled out the miclaw beta, and early feedback has been very positive. The larger opportunity is AI entering the physical world — autonomous driving, robots, humanoids — where Xiaomi is already investing, ultimately serving the broader ecosystem. **miclaw is a self-developed AI agent that tests our model prowess, boundary of capabilities, deep-integration, and data strengths**. Xiaomi will integrate our own user data while using cloud data, with robust system integration and safety measures. miclaw is a prototype for future AI agents but remains early; there is no concrete monetization model yet, and no team KPIs until the product matures. **On the RMB 60 bn allocation: it includes both R&D expenses and CapEx, with R&D consisting of current-period spend and amortization of past R&D**. For **the RMB 16 bn planned in 2026, about 70% is current-period R&D, with the balance being CapEx plus prior-period amortization**. Over the next three years, annual CapEx will include amortization from prior CapEx, so current-period R&D may be below 70%. **Q: IoT delivered strong results in 2025, but investors worry about domestic macro pressure on home appliances. How do you plan overseas expansion for IoT? With autos profitable in 2025, what is the profit outlook for 2026?** A: IoT spans multiple categories, and we look at China and overseas separately. In China, the opportunity is premiumization. While IoT is large in scale, ASPs remain low; we made significant progress last year, but we are still far from the goal — categories like watches and hair dryers still have low ASPs. We will push premiumization this year and expect many positive developments. For major appliances, our market share is ~4 pts in washers and fridges and ~10 pts in air conditioners, so there is still room to raise prices. Overseas, the runway is substantial. Our business has been China-centric; entering North America would give us a market three times China's size, and fully tapping it could be six times or more than the domestic market. We will deploy teams and products. At the Mi Store in London, I saw most products on sale were premium and the assortment was comprehensive; overseas stores grew from ~4,500 to 10,000 last year, giving high-ASP products significant potential. **For autos, we delivered over 410k units in 2025 vs. a 300k target, and aim for 550k in 2026**, and we are confident in achieving it. On profitability, note that this is the 'Smart EV and AI innovation' segment, which includes autos as well as AI and new businesses that remain in investment mode. We will continue to invest in AI and robotics this year. So assess on two dimensions: the car business performance, and the AI plus new initiatives; the segment did well last year and should remain encouraging as autos grow and other areas start to bear fruit. **Q: How does the Middle East situation impact Xiaomi's overseas business (IoT, phones)? Any effect on logistics and raw material costs? How do you protect smartphone margins, and what safeguards are there for high-end autos and AIoT?** A: The Middle East conflict is something we do not want to see, and we hope for a resolution. It has broad impacts on global industry and the economy. **For Xiaomi, the Middle East contributes only a single-digit percentage of revenue and an even smaller share of profit, so the impact is manageable**. We also note some effects on prices of oil-related materials such as plastic pellets, but overall the situation remains under control. On category margins and pricing: **actual memory cost inflation keeps exceeding expectations** — each quarter's anticipated peak has been surpassed. Because memory is a large share of product COGS, even small increases have a significant impact, making it hard to forecast precisely as before. For smartphones, given our scale and share, protecting our market position is paramount; we seek a balance between share and profit, with market standing being critical. **The degree of impact from memory inflation varies: smartphones, laptops and tablets are most affected, smart EVs next, and IoT least**. **Q: At what maturity will miclaw commercialize, and what revenue model do you expect? Any progress in the chip biz this year?** A: I have spoken a lot about miclaw. As a user, you should already have tried it — it brought many surprises but is still a new product with ample room for improvement; I have also received plenty of feedback and we will keep optimizing. Iteration will be very fast, with new versions every few days. **Overall, Xiaomi's AI is currently focused on serving users. It is too early to discuss monetization**. While our models are efficient and we have some foundation for token monetization, absolute costs are still high, so now is not the time to talk about commercial models. In addition, the XLA cognitive model is already deployed on the new SU7, and internal testing results are very strong. I am also a user and personally test ADAS; any issues are fed back directly to the team. In autonomous driving, we have deployments across both models and chips, and end-to-end integration will bring users a new experience once in place. **Q: With higher AI investment, will chip spending be adjusted? Any new chips coming? Will tablets and laptops follow phone-like pricing to protect UX?** A: R&D is rising significantly this year, but chips are a long-term strategic capability — a platform that underpins multiple product types and categories. Although AI spend is growing, we are not cutting chip investment; many chips are part of the broader AI strategy, and we will continue to invest with conviction. PCs and tablets will broadly follow a strategy similar to phones. Note our newly launched Xiaomi laptop, refined over four years, is selling very well with demand exceeding expectations. We knew memory would rise at launch, but given strong product capability, users still accepted higher prices. So product innovation and technical strength are key — even with higher memory costs, we can keep products attractive; with the team's efforts, we are confident in delivering solid results. ### Related Stocks - [Samsara Inc. (IOT.US)](https://longbridge.com/en/quote/IOT.US.md) - [XIAOMI-W (01810.HK)](https://longbridge.com/en/quote/01810.HK.md) - [XIAOMI-WR (81810.HK)](https://longbridge.com/en/quote/81810.HK.md) - [Xiaomi Corporation (XIACY.US)](https://longbridge.com/en/quote/XIACY.US.md) - [Xiaomi Corporation (HXXD.SG)](https://longbridge.com/en/quote/HXXD.SG.md) - [SIASUN (300024.CN)](https://longbridge.com/en/quote/300024.CN.md) ## Comments (2) - **我爱说真话 · 2026-03-24T15:52:10.000Z**: GLJ Field Control GP - **我爱说真话 · 2026-03-24T15:52:08.000Z**: You really shorted G LJ