Kingsoft Cloud's AI compute demand and margin improvement confirmed; price target raised to US$16

Overview

This report summarizes key takeaways from Kingsoft Cloud's earnings conference call and non-deal roadshow, focusing on 2026 guidance, AI compute, MaaS, and enterprise cloud. Morgan Stanley believes that AI demand, tight compute supply, and expense discipline jointly support the revenue and earnings outlook, and therefore reiterates its Overweight rating and raises its price target.

Core views

The company reaffirmed its 2026 revenue and margin outlook. Full-year revenue growth guidance remains at 30%-40%, and management believes there is a high likelihood of reaching the midpoint of the range, implying that revenue growth needs to accelerate in the second half of 2026. Full-year gross margin guidance is approximately 15%, with quarterly gross margin expected to remain stable or gradually rise; both AI cloud and MaaS are driving gross margin improvement. Fully depreciated GPUs that can still generate revenue are expected to provide additional depreciation benefits beginning in the second half of 2027. The company is highly confident that its full-year adjusted operating margin will turn positive and expects the margin to remain broadly stable or increase from 4% in the second quarter of 2026; both absolute expenses and expense ratios are under control. Full-year adjusted EBITDA margin guidance is approximately 35%, the RMB15 billion total investment plan is progressing on schedule, and the third quarter has started well. AI compute remains the primary growth theme. Based on current supply and demand conditions, the company believes that compute demand will continue to exceed supply for at least another year, and customers have absorbed price increases relatively well. In an environment of very high compute prices, Kingsoft Cloud prioritizes resource allocation to hyperscale customers that can sign full-lifecycle contracts and provide substantial upfront payments. This strategy sacrifices some short-term margin but provides greater long-term revenue visibility and cash flow. The company has also begun developing an asset-light, high-margin managed cloud business; AI cloud will additionally drive demand for traditional cloud services such as storage. Morgan Stanley forecasts 109% AI revenue growth in 2026, primarily driven by strong compute demand and RMB15 billion in server investment. The forecast table shows public cloud revenue rising from RMB6.633 billion in 2025 to RMB9.935 billion in 2026, RMB14.841 billion in 2027, and RMB20.817 billion in 2028. The MaaS business has integrated 120 models and serves 230 customers, who currently mainly use token services from domestic models in web programming scenarios. MaaS has the highest gross margin among the company's businesses. The report believes Kingsoft Cloud's advantage over model providers is its ability to offer the higher service-level agreements required for mission-critical enterprise processes; its advantage over major cloud providers is its neutrality toward different AI models and close connections with model providers. The company's model inference optimization capabilities are approaching those of the original model providers, while it is also considering expanding MaaS overseas. Enterprise cloud performance has been relatively weak, with only low- to mid-single-digit growth expected for the full year because rapidly rising hardware prices have caused some customers to postpone decisions. This is also the main reason Morgan Stanley slightly lowered its near-term revenue expectations. The company is internally training forward-deployed engineers, or FDEs; the report believes this model is well suited to meeting the AI deployment needs of large Chinese state-owned enterprises and government customers and could convert some project-based revenue into more recurring revenue. The forecast table shows enterprise cloud revenue rising from RMB2.925 billion in 2025 to RMB3.042 billion in 2026, RMB3.194 billion in 2027, and RMB3.322 billion in 2028, significantly slower than public cloud growth. Regarding model revisions, following the second-quarter 2026 results, Morgan Stanley lowered its 2026 revenue forecast by 0.9% and raised its 2027 and 2028 forecasts by 0.7% and 3.1%, respectively; the near-term reduction mainly reflects the enterprise cloud outlook. Due to an improved gross margin outlook and tighter expense control, EBITDA forecasts for 2026 through 2028 were raised by 2.5%, 4.9%, and 10.1%, respectively, while total operating expense assumptions were lowered by 9.2%-10.7%. Consequently, the 2026 normalized loss-per-share forecast narrowed by 26%; because the comparison base is very low, normalized EPS forecasts for 2027 and 2028 were raised by 764% and 84%, respectively. The report's financial forecasts show net revenue increasing from RMB9.559 billion in 2025 to RMB12.977 billion in 2026, RMB18.035 billion in 2027, and RMB24.139 billion in 2028; EBITDA rises from RMB1.707 billion to RMB4.850 billion, RMB9.082 billion, and RMB14.321 billion, respectively. The EBITDA margin in another key-drivers table rises from 24.4% in 2025 to 39.8% in 2026, 52.3% in 2027, and 60.8% in 2028, reflecting a different reporting basis from the approximately 35% full-year adjusted EBITDA margin guidance. The valuation continues to use 2027e EV/EBITDA. The base case applies 5.5x, corresponding to a US$16.00 price target; this multiple is below the 13x average and 10x median for US neocloud companies, reflecting Kingsoft Cloud's lower return on assets and greater uncertainty regarding supply-chain availability. Morgan Stanley selected EBITDA because depreciation policies vary across companies and EBIT is limited, making EBITDA more suitable for cross-company comparisons. The price target was increased by 7% from US$15.00 to US$16.00, primarily driven by the higher 2027 EBITDA forecast. The bull case remains at US$27.00 and 7.0x 2027e EV/EBITDA, assuming stronger demand for high-end GPUs, faster EBITDA growth, and positive progress for Xiaomi's Mimo model; the bear case remains at US$4.00 and 4.0x, based primarily on the assumption of slower AI development in China. The report also states that active institutional ownership is 53.7%, the hedge-fund-related industry long/short ratio is 2.1x, and net exposure is 33.6%.

Analysis framework

Morgan Stanley first used information from the conference call and non-deal roadshow to verify management's revenue, gross margin, operating margin, EBITDA margin, and investment guidance, and then separately analyzed the demand, competitive advantages, and revenue models of AI compute, MaaS, and enterprise cloud. The report subsequently incorporated its latest operating assessment into revenue, expense, EBITDA, and EPS forecasts for 2026 through 2028, and finally updated its price target and risk-reward assessment using 2027e EV/EBITDA and bull, base, and bear scenarios.

Methodology notes

• Valuation MethodEV/EBITDA valuation

  2027e EV/EBITDA valuation

  The report assesses valuation using enterprise value divided by EBITDA and applies 5.5x in the base case. This metric reduces the impact of differing depreciation policies and limited EBIT on cross-company comparisons.

• Valuation Method

  Bull, base, and bear scenario valuation

  The report applies 7.0x, 5.5x, and 4.0x 2027e EV/EBITDA, corresponding to US$27, US$16, and US$4, respectively, to present a valuation range under varying GPU demand, model development, and Chinese AI growth assumptions.

• Industry/Sector Analysis FrameworkSupply-demand framework

  AI compute supply-demand analysis

  The report assesses AI cloud revenue, contract visibility, and margin changes based on compute supply shortages, customer acceptance of price increases, GPU procurement, and resource allocation strategies.

• Company Fundamentals and Financial FrameworkOperating/Financial Leverage Analysis

  Transmission of revenue growth, gross margin, and expense control to EBITDA and EPS

  The report combines a higher AI business mix, gross margin improvement, and lower operating expense assumptions to explain why the increases in EBITDA and normalized EPS forecasts are significantly larger than the revenue revisions.

• Company Fundamentals and Financial Framework

  Morgan Stanley ModelWare forecasting framework

  Unless otherwise stated, the report's financial metrics and forecasts are based on the Morgan Stanley ModelWare framework, incorporating the latest quarterly results and management guidance into model revisions for 2026 through 2028.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Kingsoft Cloud Holdings (Kingsoft Cloud, KC.US)

  The report believes the company benefits from demand for high-end GPUs and AI compute, persistent supply shortages, and the expansion of its AI cloud and MaaS businesses.

  Strengths

  Ability to sign full-lifecycle contracts with hyperscale customers and obtain substantial upfront payments; MaaS offers high SLAs, model neutrality, broad model connectivity, and inference optimization capabilities approaching those of original model providers; strict expense control.

  Weaknesses

  Enterprise cloud is expected to grow only at a low- to mid-single-digit rate, with rising hardware prices causing customers to delay decisions; prioritizing compute resources for long-term contract customers sacrifices some short-term margin.

  Comparison

  The base-case valuation is 5.5x 2027e EV/EBITDA, below the 13x average and 10x median for US neocloud companies, reflecting lower returns on assets and greater supply-chain uncertainty.

  Risks

  Supply constraints may cause procurement shortfalls, financing costs may be affected by rising interest rates, and slower-than-expected development of Chinese AI models could weaken demand and valuation.

Key data

• 2026 revenue growth guidance30%-40%Unchanged, with a high likelihood of reaching the midpoint of the range, implying faster revenue growth in the second half.
• Full-year 2026 gross margin guidanceApproximately 15%Quarterly gross margin is expected to remain stable or increase, with contributions from both AI cloud and MaaS.
• Second-quarter 2026 adjusted operating margin4%The company expects it to remain broadly stable or increase thereafter and is confident it can turn positive for the full year.
• 2026 adjusted EBITDA margin guidanceApproximately 35%Full-year profitability guidance reiterated by management.
• Total investment planRMB15 billionProgressing as planned, with a good start to the third quarter.
• 2026 AI revenue growth forecast109%Driven by strong compute demand and server investment.
• MaaS platform scale120 models and 230 customersMaaS has the highest gross margin among the company's businesses.
• Revenue forecast revisions2026 -0.9%; 2027 +0.7%; 2028 +3.1%The near-term reduction mainly reflects a weaker enterprise cloud services outlook.
• EBITDA forecast revisions2026 +2.5%; 2027 +4.9%; 2028 +10.1%Reflecting an improved gross margin outlook and strict expense control.
• Operating expense assumption revisionLowered by 9.2%-10.7%Driving improvements in EBITDA and EPS forecasts.
• Normalized earnings forecast revisions2026 loss per share narrowed by 26%; 2027 EPS raised by 764%; 2028 EPS raised by 84%The substantial increases in 2027 and 2028 partly reflect a very low comparison base.
• Net revenue forecast2025-2028: RMB9.559 billion, RMB12.977 billion, RMB18.035 billion, RMB24.139 billionMorgan Stanley model forecasts.
• EBITDA forecast2025-2028: RMB1.707 billion, RMB4.850 billion, RMB9.082 billion, RMB14.321 billionMorgan Stanley model forecasts.
• Price target revisionRaised from US$15.00 to US$16.00Raised by 7%, primarily driven by the higher 2027 EBITDA forecast.
• Risk-reward scenariosBull US$27.00; base US$16.00; bear US$4.00Applying 7.0x, 5.5x, and 4.0x 2027e EV/EBITDA, respectively.
• Market price metricsClosing price US$12.37; 52-week range US$8.35-US$18.52Closing price as of August 19, 2026; the report states 29% upside to the price target.

Impact & implications

The report believes that the AI compute shortage and customers' acceptance of price increases enable Kingsoft Cloud to sustain strong revenue growth, while full-lifecycle contracts and upfront payments can improve long-term revenue visibility and cash flow at the cost of some short-term margin. MaaS, asset-light managed cloud, and demand for traditional storage services should improve the business mix and gross margin, while expense control enables EBITDA and EPS to improve faster than revenue. Enterprise cloud remains a near-term growth drag because hardware price increases have caused customers to delay decisions; if the FDE model can convert project-based business into recurring revenue, it could improve revenue quality. Morgan Stanley accordingly raises its medium- to long-term earnings forecasts and base-case price target, but continues to apply a significant valuation discount relative to US neocloud companies to reflect return-on-assets and supply-chain uncertainty.

Risks

• Supply-side constraints could result in insufficient procurement of GPUs and related equipment, affecting compute deployment and revenue realization.
• Higher-than-expected interest rates could increase funding costs and weaken investment returns.
• Slower-than-expected development of Chinese AI models could weigh on compute demand; the bear case explicitly assumes a slowdown in China's AI development.
• Lower returns on assets and uncertainty regarding supply-chain availability could continue to constrain valuation multiples.

What to watch

• Monitor whether revenue can accelerate in the second half of 2026 and bring full-year growth close to the midpoint of the 30%-40% guidance range.
• Monitor whether quarterly gross margin, adjusted operating margin, and the approximately 35% full-year adjusted EBITDA margin can be achieved.
• Monitor the RMB15 billion investment plan, GPU procurement progress, and whether the compute shortage persists for at least one year as the company expects.
• Monitor MaaS customer and model counts, overseas expansion, and progress in Xiaomi's Mimo model.
• Monitor delays in enterprise cloud customer decisions due to hardware price increases and the effectiveness of the FDE model in converting business into recurring revenue.
• Monitor the impact of interest-rate changes on financing costs and investment returns.