Kingsoft Cloud Holdings Limited (KC) Earnings

Management highlights

- AI-driven business growth * AI cloud continues to be the primary driver of total revenue growth, with the business structure shifting steadily toward high-growth AI services * MaaS business on the Starflow platform grew extremely rapidly, with 120 models deployed online, supporting synchronized launches of major new models and serving over 230 enterprise customers * Deepened cooperation with leading customers in high-growth emerging sectors including embodied AI, autonomous driving, and AI for Science, delivering large-scale computing clusters to support rapid model iteration - Profitability improvement * Adjusted gross margin reached 15.4%, up 2.4 percentage points quarter-over-quarter and 0.5 percentage points year-over-year * Adjusted operating profit turned positive for the first time, with an adjusted operating margin of 4% (a new company record), driven by revenue expansion, higher margins, and improved operating efficiency * Adjusted net loss narrowed to RMB 136 million, down from RMB 300 million in the year-ago quarter; non-GAAP EBITDA reached RMB 1.1 billion, up 171% year-over-year, with a 36% non-GAAP EBITDA margin - Product and technology upgrades * Optimized high-concurrency inference deployment on Starflow Max, improved model throughput and granular usage management * Launched AgentKit, a full set of tools for building production-grade enterprise AI agents covering security, knowledge management, and governance * Enhanced the Starflow training and inference platform with more flexible resource scheduling, improving resource utilization and reducing development costs * Completed deep integration and full-lifecycle visual management for multiple mainstream domestic AI chips on the Galaxy Stack private AI infrastructure platform - Enterprise cloud strategic progress * Signed framework agreements for dedicated digital infrastructure projects including Jianghai Cloud for Yangtze River shipping and a strategic partnership with Wuhan Municipal Data Bureau for digital governance * Deployed a cloud-based integrated medical virtual surgery platform (developed under a national key R&D program) in over 30 hospitals nationwide * Proactively adjusted the enterprise cloud business structure, shifting from low-margin project-based system integration to recurring operating-based revenue

Guidance

- Full-year 2026 capital expenditure guidance (including capitalized assets from third-party financing and finance leases) is maintained at RMB 15 billion, unchanged from prior estimates. As of the first half of 2026, cumulative capex reached RMB 6.2 billion, accounting for over 41% of the full-year target, with visible acceleration in procurement growth - Management expects enterprise cloud revenue recognition to concentrate in the second half of 2026, with a strong project pipeline supporting full-year performance after the proactive business structure adjustment - Management expects overall ROIC (return on invested capital) to improve steadily as AI business scales, fixed costs are diluted, and business structure optimization continues

Segment performance

Kingsoft Cloud's total Q2 2026 revenue was RMB 3.072 billion, up 31% year-over-year and 40% quarter-over-quarter. - Public cloud services: Revenue of RMB 2.358 billion, up 45% year-over-year. Within public cloud, AI cloud revenue reached RMB 1.33 billion, up 82% year-over-year, accounting for 56% of public cloud revenue and 43% of total company revenue. MaaS (Model as a Service) revenue was more than 12 times the Q1 2026 level. - Enterprise cloud services: Revenue of RMB 714 million, down 1% year-over-year, accounting for approximately 23.2% of total revenue. Revenue from the Xiaomi-Kingsoft ecosystem reached RMB 810 million, up 28% year-over-year, accounting for 26% of total revenue, with 51% year-over-year growth in revenue from the top five non-ecosystem customers.

Risks & headwinds

- Chip and AI server supply constraints are expected to remain a long-term new norm in the industry, creating ongoing uncertainty for capacity expansion plans - MaaS business demand and profitability are subject to volatility from token price fluctuations, shifts in customer preference for newly launched models, and operating efficiency uncertainty - Enterprise cloud customer budget adjustments due to upstream hardware price hikes have delayed decision-making and contracting, creating near-term revenue growth pressure - The company continues to face high depreciation costs from large-scale AI infrastructure investments, which pressure near-term margins until capacity utilization improves

Analyst Q&A

• Q: How have improvements in open source large language model capabilities impacted Kingsoft Cloud's MaaS business, what is the usage growth trend, and will the company allocate more resources to MaaS relative to raw computing power leasing?

  A: Improved open source domestic model capabilities have driven increased demand for Chinese models, as domestic models are replacing overseas models in segments like web coding. Growing agent adoption has also increased demand for the company's AgentKit product, with price-sensitive routine workloads favoring domestic open source models, all of which benefit the company's neutral MaaS business. Management follows a balanced resource allocation strategy: raw computing power leasing has stable long-term contracted utilization, while MaaS is more volatile, so the two are dynamically balanced to complement each other.

• Q: What has been chip procurement progress since June, what is the updated full-year capex guidance, and how should we view the slow enterprise cloud revenue growth and its medium-term positioning?

  A: AI chip supply constraints have become a long-term industry norm, but the company has mitigated this by expanding its supplier base and increasing compatibility with domestic chips, which perform well for inference workloads. Full-year capex guidance is maintained at RMB 15 billion. For enterprise cloud, near-term slow growth stems from customer budget delays due to upstream price hikes and natural seasonality (revenue recognition concentrates in H2), and is also a result of proactive structural shift from low-margin project-based work to recurring operating revenue (reclassified to public cloud), so the segment's fundamentals remain solid.

• Q: What is Kingsoft Cloud's competitive positioning for MaaS relative to peers, and what is the latest pricing trend for AI cloud services and customer reaction to recent price adjustments?

  A: The company has a unique neutral positioning: unlike providers with proprietary in-house models, it is not required to push its affiliated models and can offer customers any leading model they prefer, which is a key strategic advantage. It also owns its computing power infrastructure, which is required to maintain strong MaaS profitability. The company recently raised prices for storage and AI computing power to pass through higher costs. Most customers accepted the increases easily, and the company was able to improve profitability on these services, thanks to its infrastructure and operational capabilities that deliver reliable high-performance service.

• Q: How does Kingsoft Cloud see its long-term market positioning relative to cloud providers with proprietary models and integrated ecosystems, and what sustainable margin level can it achieve?

  A: As a neutral cloud provider, the company maintains strong partnerships with all top model developers and labs, so it can always offer customers the best model for their use case. It also leverages its in-house technical capabilities to deliver higher SLA reliability than many competitors, and collaborates closely with model developers to optimize inference efficiency, often reaching performance levels close to the original model developers themselves. This neutral, customer-centric positioning supports sustainable long-term margins.