KC Q1 2025: Xiaomi AI cluster delivery signals Q2 revenue boost
- Catalyst from AI business with key ecosystem partners: Management highlighted that a delivered 512-node cluster to Xiaomi and the training of Xiaomi’s 7-billion-parameter model on their infrastructure positions the company for strong future revenue pickup, especially as these figures will impact the next quarter's performance [doc 5].
- Improving margin profile as AI penetration deepens: Executives noted that the gross profit and operational margins are expected to improve in subsequent quarters as AI-driven projects and long-term contracts mature, supporting a more favorable margin profile going forward [doc 6][doc 7].
- Innovative CapEx financing and diversified investment strategy: The company is leveraging diversified financing methods—including cash spending, leasing, and off-balance sheet financing—to efficiently support its AI infrastructure investments, thereby preserving liquidity and reducing short-term cash outlay [doc 7].
- Seasonal and Revenue Delays: The Q&A revealed that both public and enterprise cloud segments face seasonality and delayed recognition of revenue, as key projects are delivered late in the quarter, potentially putting short-term growth under pressure.
- Chip Band Impact: Management acknowledged that while short-term revenue isn’t negatively affected, tighter chip restrictions in the mid-to-long term could increase costs and complicate AI infrastructure expansion.
- Margin Pressure from Pricing and Competition: There is a risk of margin compression due to aggressive price competition in AI cloud leasing services and the potential impact of open-source models like DeepSeek on traditional model training contracts.
Metric | Period | Previous Guidance | Current Guidance | Change |
---|---|---|---|---|
Revenue Growth | FY 2025 | accelerated revenue growth | no formal guidance provided (Q1 2025 did not issue specific annual guidance) | no current guidance |
Profitability (non‐GAAP operating profit) | FY 2025 | positive non‐GAAP operating profit | no formal guidance provided (Q1 2025 did not issue specific annual guidance) | no current guidance |
AI Investment | FY 2025 | RMB 10 billion | no formal guidance provided (Q1 2025 did not issue specific annual guidance) | no current guidance |
Revenue Contribution from Xiaomi and Kingsoft Group | FY 2025 | RMB 11.3 billion over FY2025–FY2027 | no formal guidance provided (Q1 2025 did not issue specific annual guidance) | no current guidance |
Non‐GAAP Operating Margin | Not specified | no prior guidance | Expected improvement in operating margin in the second half of 2025 (with EBITDA margin anticipated to recover faster) | no prior guidance |
AI Business and Revenue | Not specified | no prior guidance | Demand for AI cloud services expected to grow; large projects with Xiaomi likely to be deployed and booked from Q2 2025 | no prior guidance |
Capital Expenditure and Financing | Not specified | no prior guidance | Ongoing investments in AI infrastructure financed via cash, leasing and bank loans; updates on total spending to be provided later | no prior guidance |
Topic | Previous Mentions | Current Period | Trend |
---|---|---|---|
AI Revenue Growth | Consistently highlighted with strong year‐over‐year figures—e.g., triple‐digit growth in Q3 2024 (6.9x increase ), nearly 500% YoY growth in Q4 2024 and 10x YoY in Q2 2024 | Q1 2025 reported AI gross billing of RMB 525 million with over 200% YoY increase, making AI a key growth driver | Recurring strong growth with consistently positive sentiment, though growth multipliers vary over periods |
Infrastructure Investments | Emphasized heavily in Q2 (95% of CapEx on AI, RMB 654.8 million ) and Q3 (significant GPU server investments , planned related-party financing for RMB 11.3 billion cap ), with Q4 featuring a RMB 5 billion focus on computing power | In Q1 2025, CapEx of RMB 605 million was allocated to expanding AI infrastructure with acceleration of computing clusters and flexible financing arrangements | Continuous high investment focus with evolved flexible financing strategies |
Strategic Shift to AI Cloud and Solution Services | Discussed across Q2, Q3, and Q4 with focus on transforming from traditional IT to AI cloud services, launching new platforms, and targeting both enterprise and public cloud segments | Q1 2025 reinforced the shift by highlighting new platform features (e.g., the Xingliu training and inference platform) and deeper ecosystem integration with an expected service launch in Q2 2025 | Ongoing strategic reinforcement with expanding capabilities and clearer roadmap |
Deep Ecosystem Partnerships with Xiaomi and Kingsoft | Consistently mentioned: Q2 reported RMB 370 million (20% of revenue, balanced internal/external mix ), Q3 noted 36% YoY growth and 28% of public cloud revenue , and Q4 emphasized a revenue increase and a revenue cap plan (RMB 11.3 billion) | Q1 2025 reported RMB 500 million in revenue from the ecosystem (50% YoY increase) along with major deployments (e.g., a 512-node cluster for Xiaomi) | Persistent and deepening partnerships with rising revenue contributions and strategic cluster deployments |
High CapEx Investments and Innovative Financing Strategies for AI | Prior periods stressed heavy AI-related CapEx (e.g., 95% in Q2 , RMB 11.3 billion related-party transaction cap in Q3 , and RMB 5 billion investments with conservative depreciation in Q4 ) | Q1 2025 featured RMB 605 million in CapEx along with diversified financing approaches (financial leasing, bank loans, off-balance-sheet arrangements) to support AI infrastructure | Steady commitment to high investments with evolving and innovative financing strategies |
Margin Profile Improvement vs. Pricing Competition Risks | Q4 showed marked margin improvements (gross margin at 19.2%, EBITDA turnaround ); Q3 reported expanded margins driven by higher-margin AI services (gross margin to 16.3% ); Q2 highlighted a 17.1% grid with strong AI contribution | Q1 2025 management is optimistic about margin improvements in H2 2025 driven by AI revenues, while also flagging pricing pressures from demand dynamics and resource pool arrangements | Consistent margin improvement despite emerging pricing pressures, with vigilance toward competitive pricing dynamics |
Chip Supply Restrictions Impacting AI Infrastructure | Earlier periods (Q2, Q3, Q4) did not specifically mention chip supply constraints, aside from a brief Q4 note without detailed commentary | Q1 2025 acknowledged chip supply challenges; inventory stockpiling and strengthened domestic (Made in China) partnerships were highlighted to mitigate mid-to-long-term impacts | An emerging concern in Q1 2025 with proactive measures, representing a new focus compared to previous periods |
Seasonality and Revenue Recognition Delays | Not mentioned in Q2, Q3, or Q4 discussions | Q1 2025 explicitly described seasonal impacts (e.g., Chinese New Year affecting enterprise cloud revenues and employee bonus expenses) as well as revenue recognition delays (e.g., cluster delivery revenue deferred to Q2 2025) | A new topic in Q1 2025, reflecting short-term operational fluctuations and timing effects on revenue |
Dependency on Related-Party Revenues and Concentration Risks | Q2 detailed a 36.9% YoY increase in related-party revenue (RMB 370 million, 20% of total revenue ); Q3 noted a 28% contribution and significant related-party collaboration; Q4 emphasized revenue from ecosystem partners with a multi-year revenue cap (RMB 11.3 billion) | Q1 2025 reported significant ecosystem revenue (RMB 500 million) and major AI deployments for Xiaomi, implying continued dependency on a few key partners | A recurring reliance that is deepening, with strong strategic value but also an implied concentration risk |
Transparency in Pricing and Margin Disclosure | Q4 expressed caution and maintained strict confidentiality on pricing tactics (protecting business secrets) ; Q3 and Q2 provided limited disclosure due to the integrated nature of multi-segment services | Q1 2025 offered some insights into pricing pressures—mentioning the resource pool supply chain model affecting margins—but maintained overall opacity on detailed figures | A consistently sensitive topic, with management maintaining a cautious approach across all periods |
Competition from Open-Source Models and Large Language Model Dynamics | Q4 mentioned DeepSeek positively as a driver for lower costs and enhanced market adoption ; Q3 and Q2 did not address this topic explicitly | Q1 2025 discussed the impact of successful open-source efforts (DeepSeek R1) on demand from legacy large language model customers, yet reassured that long-term contracts remain intact | An emerging and increasingly relevant topic, with management noting competitive pressures but mitigating factors through long-term agreements |
Downscaling of Traditional CDN Business | Q2 detailed a reduction in the standard CDN contribution (from 23% to 19%, maintaining RMB 300 million minimum), Q3 explained proactive scaling down with significant IDC cost reduction (8.7% drop ), and Q4 reported a 2.6% cost drop due to strategic scaling | Q1 2025 did not mention traditional CDN business adjustments, indicating a possible deprioritization in favor of AI and cloud initiatives | Previously a key operational focus, but not mentioned in Q1 2025, suggesting a strategic shift toward higher-growth, higher-margin AI services |
-
Cloud Growth
Q: Weak cloud growth and Xiaomi model update?
A: Management noted that both public and enterprise cloud growth were tempered by seasonality—with enterprise delays due to Chinese New Year budgeting and public cloud projects having longer lead times, such as the 512-node cluster delivered to Xiaomi at March’s end. They also highlighted that Xiaomi’s 7-billion-parameter model is trained on Kingsoft Cloud’s clusters, signaling strong long‑term demand. -
CapEx & Chip Supply
Q: CapEx breakdown, chip issues, and model impact?
A: Haijian He explained that Q1 CapEx was RMB 605 million, financed partly through cash and leasing arrangements. Meanwhile, Tao Zou mentioned that though chip supply restrictions have a short‑term impact, inventory and alternative Made‑in‑China resources are cushioning any disruption, and demand remains robust for both large language models and smaller alternatives. -
Government AI & Margin Outlook
Q: Govt AI machine and margin outlook?
A: Zou described a jointly developed, integrated AI solution for government affairs that leverages both software and hardware—with pricing varying by configuration. Haijian He added that while Q1 margins suffered due to fixed expenses amid revenue seasonality, improved revenue in upcoming quarters should bolster margins. -
Margin & DeepSeek
Q: AI leasing margin and DeepSeek's training impact?
A: Tao Zou indicated that established long‑term contracts protect current margins, though newer resource‐sharing arrangements with partners slightly dilute margins. He also noted mixed effects from DeepSeek, with some large language model players reducing demand, countered by increased interest from other industry segments.
Research analysts covering Kingsoft Cloud Holdings.