Xiaomi Group (HK01810) released its 2025 financial report, achieving another record performance: total revenue for 2025 increased by 25% year-on-year to RMB 457.3 billion, surpassing the RMB 400 billion mark for the first time. Adjusted net profit surged 43.8% year-on-year to RMB 39.2 billion, also setting a new historical high.

Following the earnings release, Xiaomi Group Partner and President Lu Weibing, along with Vice President and CFO Lin Shiwei, attended the subsequent earnings conference call to explain key financial highlights and respond to questions from several analysts.

The following is a transcript of the analyst Q&A session from the conference call:

**Morgan Stanley Analyst Andy Meng:** Congratulations to Xiaomi on achieving record highs in both revenue and profit for 2025. I have two questions. First, regarding memory components: since Q4 of last year, we have seen significant price increases for memory, which is a major risk concern for investors in the smartphone business. Recently, we've noticed competitors have started raising prices on some phone models to pass on cost pressures. However, Xiaomi's smartphone prices have remained stable. Does this contrast indicate that Xiaomi has a clear advantage in memory inventory and supply chain management compared to rivals, allowing it to perform better amid industry challenges this year? Could President Lu share your latest views on memory storage costs and Xiaomi's strategy for coping this year?

My second question concerns our new energy vehicles. We noted the successful launch of the new-generation Xiaomi phone last week on the 19th. In post-launch communications with investors, some interpreted Xiaomi's decision to only release order lock-in figures instead of pre-order numbers as an indication that pre-orders fell short of expectations—a relatively negative view. Other investors see it as a more rational, measured sales strategy, suggesting that the new Xiaomi Su7 will achieve better sales—a more positive interpretation. Could management share your perspective on these two investor viewpoints?

**Lu Weibing:** I'll address your first question. Memory costs have been a topic of concern each quarter since last year, and I've shared similar views on multiple occasions. Summarizing my past points: first, I believe this is a long cycle, likely extending into 2027. Second, the magnitude of the price increase could be substantial. Based on these assessments and subsequent developments, I now feel the cycle might be even longer than initially judged. Driven by surging AI server demand, the extent of memory cost inflation has exceeded our original forecasts significantly. My previous predictions were already considered aggressive within the industry, but reality has been even more aggressive. This is the current reality—a long cycle impacting all consumers, not just smartphones, though phones receive more attention. Some product categories with smaller memory requirements have already faced supply shortages or even halted production. This is the industry's current state.

We have a simple way to gauge the impact: comparing the memory's Bill of Materials (BOM) cost to the total device BOM cost. If the memory BOM cost is high relative to the total, the impact is greater. In our existing product lines, smartphones, tablets, and notebooks have relatively high memory BOM cost ratios. Of course, phone prices vary—high-end models have a smaller memory cost proportion, while low-end phones have a larger one. If a company has a high proportion of low-end phones, the impact will be greater. This is a simple qualitative assessment.

Regarding competitors raising phone prices recently due to memory cost pressures, I fully understand these moves. No phone manufacturer would raise prices unless forced to an unbearable point. However, the timing of price hikes may vary, but I believe they are ultimately unavoidable. For Xiaomi, we are under significant pressure, but we aim to absorb the costs for consumers as long as possible. When we can no longer sustain this, we will have to raise prices. I hope consumers will understand when that happens—we simply held out longer. Ultimately, price increases are an inevitable trend.

To address this, Xiaomi's advantage lies in our business diversity. Categories like major appliances have much smaller memory requirements, though automotive needs are substantial due to smart cockpit demands. However, compared to the total vehicle BOM, the proportion is smaller than for phones. Our diversified product portfolio helps us manage this issue better. That said, Xiaomi's memory demand remains high globally across phones, tablets, notebooks, TVs, and vehicles. We have strong, trusted relationships with global memory suppliers and long-term supply agreements. Thus, we currently face minimal shortage risks and may even have some relative price advantages.

Additionally, my previous pessimistic forecasts led us to be more aggressive in stockpiling, giving us better inventory positioning. Overall, however, the impact on costs for our core smartphone-centric devices is substantial, and short-term pressure remains.

**Lin Shiwei:** I'll answer the second question. Following the Su7 launch last week, we observed strong user interest: over 50,000 test drives in the first three days. We released two figures: 15,000 orders within 34 minutes, and order lock-ins exceeding 30,000 after three days. We fulfilled our commitment to start deliveries on the fourth day post-launch, having pre-produced vehicles to address previous-generation users' long wait times. These steps are iterations based on our first-generation experience.

Regarding why we only release lock-in numbers instead of pre-orders: we believe lock-in figures more fairly represent actual demand, as they indicate users' firm intent to take delivery and enter the production cycle. While industry practices vary, we will stick to this approach.

Additionally, I'll share some user insights from the lock-in data, which show improvements over the previous generation. First, most lock-in users are new, not previous-generation owners. Second, the proportion of female users is higher. iPhone users, previously around 50% of first-gen data, now approach 60%. Lock-in progress is also faster. Nearly 60% of users selected paid color options, indicating higher penetration rates across lock-in speed, female user ratio, iPhone user share, and premium choices compared to the last generation.

**Goldman Sachs Analyst Timothy Zhou:** I have two AI-related questions. First, over the past two years, Xiaomi has built foundational models and multimodal capabilities. How do these multimodal AI abilities benefit the overall ecosystem? How does the recently launched MiClaw differ from tools developed by internet platforms? How do you view the interplay with internet platforms, and what significance does this hold for the upcoming AIOS?

Second, more quantitatively, how does Xiaomi plan service models or commercialization for AI serving internal operations, ecosystem users, and third-party developers? What KPIs are set for the team regarding large model applications? CEO Lei Jun previously mentioned a three-year investment plan of RMB 60 billion. How will this be allocated between OPEX and CAPEX?

**Lu Weibing:** Back in 2023, we spent considerable time planning Xiaomi's AI future. We believe it unfolds in stages: the first layer is AI infrastructure, including computing power and servers. After infrastructure comes application explosion, serving practical uses. We identified 2026 as the year for AI application breakout. Ultimately, AI will bridge the virtual and physical worlds, a key judgment guiding our strategy.

Based on this, we began building relevant teams in 2024, invested in large models in 2025, and delivered milestones including language and voice models. Last year, we focused on future AI agents and their role in personal smart devices, enabling previously impossible tasks. With MiClaw's launch, we've achieved rapid phone deployment—a global first for integrating such capabilities into mobile devices. Feedback has been very positive.

We believe the next major milestone is AI entering autonomous driving and robotics—two key vectors. Xiaomi has made corresponding investments, aiming to serve our entire retail ecosystem. This is our fundamental direction, and you've seen our delivered outcomes.

Regarding MiClaw's differentiation: its key advantage is system-level permissions. As a hardware company, when we integrate our own Agent, it has inherent system access. This tests model capability, permission availability, data capacity, and integration depth. Xiaomi's strength lies in having models, system permissions, data, users, and integration capabilities—advantages third parties lack due to limited system access and data. Our security is also stronger. We see MiClaw as a prototype for future AI paradigms. As for commercialization, it's still early; we have no concrete plans or team KPIs yet, as those require business maturity.

**Lin Shiwei:** The RMB 60 billion investment includes R&D spending and CAPEX. R&D costs encompass both current expenses and prior CAPEX amortization. It can be divided into three parts: current R&D expenses (over 70% of the 2026 allocation), current CAPEX costs, and amortization of past CAPEX. Over the next three years, the R&D expense proportion may be slightly lower than 70%, with CAPEX and its amortization taking a higher share.

**CICC Analyst Hanjing Wen:** I have two questions. First, regarding IoT: 2025 performance was strong in both revenue and connected devices. Considering positive factors like smart major appliances hitting record highs, but also concerns about domestic subsidy phase-outs, what is your outlook for the domestic IoT market this year? Also, how is IoT overseas expansion progressing under the new retail strategy?

Second, on automotive: 2025 automotive operations achieved profitability. With the new Su7 launch and subsequent plans, what are the profit and earnings targets for the automotive business this year?

**Lu Weibing:** For IoT, given the wide product range, I'll separate China and overseas. In China, a key opportunity is IoT premiumization. While our IoT scale is large, our average selling price remains relatively low. Despite progress last year, we still lag behind targets. For instance, our watches and hair dryers have significant ASP gaps versus top players. With past R&D and manufacturing investments, we expect substantial high-end advancements this year—our biggest opportunity.

In major appliances, our market share is still low: 4-5% for refrigerators and washing machines, around 10% for air conditioners. We see ample room for growth in both scale and ASP. New retail expansion supports this—we reached about 18,000 stores by end-2025, greatly boosting IoT development.

Overseas, IoT potential is huge, with the overseas market about three times the size of China's. Even achieving half of our domestic penetration would mean significant multiples of growth. We are still in early stages of blank market coverage. By deploying teams, channels, and localized products—having addressed adaptation and compliance—we see massive potential. Premium, high-ASP products also have great upside overseas, supported by new retail expansion: from 450 stores last year to over 1,000 by year-end, focusing on mid-to-high-end products. My visit to a London Xiaomi Store was impressive—it primarily sells premium items. Categories like cleaning products have strong unmet demand, and even rice cookers sell well in the UK. Overall, IoT has substantial growth potential both domestically and internationally.

**Lin Shiwei:** On automotive, we delivered over 410,000 vehicles in 2025, exceeding our initial target. CEO Lei Jun set a 2026 delivery goal of 550,000 units, so we expect growth this year. Despite some market pressure, we are confident based on our projections and launch data.

Regarding profitability: this segment is part of our "7000-level AI New Business" unit, which includes AI investments and other new ventures. Thus, it's not solely automotive. At this stage, new businesses are still in investment phase. As mentioned, we will continue increasing AI investments, including in robotics. The segment's performance should be viewed in two parts: automotive growth plus investments in other areas. Last year's strong results will be followed by automotive growth and further investments this year.

**Citigroup Analyst Tina Wang:** Two questions. First, given the Middle East situation, is there any impact on overseas operations (e.g., phones), or increased costs for materials/logistics? Second, on overall gross margin trends: for phones, is there a principle regarding profit level maintenance or price adjustments? For automotive, despite industry pressure, do you have principles to outperform peers or ensure certain profit levels? For AIoT, premiumization is a strategy—will this support margin improvement this year?

**Lu Weibing:** Regarding the Middle East, we hope for a peaceful resolution to conflicts nobody wants to see. Impacts on the global economy and industry are significant. For Xiaomi specifically, Middle East revenue is a single-digit percentage of total sales, so even if affected, the impact on our overall business is relatively small. We view the situation as manageable. However, we already see indirect global effects, such as rising costs for oil-related bulk materials.

On margins for phones, autos, and IoT, plus pricing principles: quantifying memory cost impacts is challenging. Even with high base prices, further increases—however small in percentage terms—translate to large absolute figures, reducing pricing visibility and lead time. Nonetheless, market share scale remains crucial for phones. Our principle is to protect market share while making trade-offs, maintaining our product positioning. As few phone makers are listed, peer comparisons are difficult, but this reflects our competitive approach.

For vehicles, memory costs are significant in absolute terms but a smaller portion of total vehicle BOM versus phones, so the impact is relatively smaller. For IoT, memory usage is minimal, and our long inventory cycles mitigate effects. Overall, phones, tablets, and notebooks face the greatest impact; autos are affected but less so; IoT impact is minimal.

**CITIC Securities Analyst Jinchi Li:** I had the pleasure of testing MiClaw—impressive performance. What preconditions would trigger consideration of commercialization for large models or MiClaw, linking to revenue? Second, any updates on smart driving capabilities and chip resources this year?

**Lu Weibing:** On MiClaw, as a beta tester, you've seen its promising aspects. However, the product is still young and needs refinement—I use it daily and provide feedback, such as on the user interface needing improvement. Many desired features aren't yet available, but progress is rapid, with version updates every few days. We have high expectations. For Xiaomi, AI ultimately serves our "Human-Vehicle-Home" ecosystem, enhancing user experience and interaction. Pure commercialization is premature at this stage; while our large models show high efficiency and specialization, absolute costs remain high. It's too early to discuss commercialization.

On smart driving, our XLA model significantly enhances capabilities, as highlighted at the launch. The full OTA version will take a few more months, but internal tests are very promising. I alternate between Su7 and Yu7, relying heavily on assisted driving and frequently submitting improvement suggestions. I feel our smart driving advancing steadily. We have deployments in driving solutions, models, and related chips. End-to-end integration will increasingly improve user experience. I encourage you to upgrade and experience the progress.

**UBS Analyst Hongyi Su:** Two questions: first, on new business investments, with accelerated Mimo model iterations and reaffirmed AI commitments, does the priority of in-house chip development adjust within new business spending? Second, for IoT—tablets and PCs also face memory cost pressures. Will strategies resemble phones, prioritizing share and consumer experience?

**Lu Weibing:** We are indeed increasing R&D spending this year. Chip development remains a critical long-term strategic capability—a platform essential for many products, especially with AI and large models increasing chip dependency. While boosting AI investments, we haven't reduced chip spending; instead, many chips are transitioning rapidly to AI needs. Chips are integral to our broader AI strategy, so I have no concerns.

For PCs and tablets, strategies will be similar to phones. Note our recently launched notebook—Xiaomi's first in a decade—is facing significant demand exceeding supply, despite factoring in memory cost hikes during development. Strong product power allows some price tolerance. This year's environment tests innovation and technological capabilities. While memory costs pose pressure, we have methods to partially offset increases. Despite challenges, I believe our team can deliver satisfactory results for 2026.