Alibaba's financing indicates a longer AI capex cycle, but the value loss priced in by the market far exceeds the transaction's actual cost

Overview

Using Alibaba's August 2026 share placement as its starting point, the report analyzes how China's three leading internet cloud service providers are financing AI capital expenditure. JPMorgan's core conclusion is that the financing indicates Alibaba's elevated capital expenditure may persist longer than previously expected, but the transaction's economic cost is insufficient to explain the unusual share-price decline; Alibaba, Tencent, and Baidu differ significantly in their balance sheets and financing paths and should not be priced uniformly.

Core views

On August 24, 2026, Alibaba issued 710 million new shares at HK$112.70 per share, raising HK$80 billion (US$10.3 billion). The issue price represented an 8.4% discount to the previous Friday's closing price, and the new shares were equivalent to 3.7% of the existing share count, with settlement on August 26. The proceeds are intended to build full-stack AI capabilities, including infrastructure. This was Alibaba's first major share placement since its 2019 Hong Kong listing, and the deal was upsized due to demand from long-term investors. Alibaba's Hong Kong shares fell 8.5% on August 25, while Tencent declined 2.2% and the Hang Seng TECH Index fell 2.1%. After decomposing the move, the report estimates that the discount cost existing shareholders about US$0.9 billion, peer-market movements explained about US$6.7 billion, and an additional US$18 billion to US$19 billion represented an unusual revaluation. After adding the US$10.3 billion in cash, pro forma market capitalization remained approximately US$286 billion. In other words, the additional value erased by the market was about twice the cash raised, while the purely arithmetic effect of the issuance discount accounted for only about 0.3% of Alibaba's share-price movement. If the entire unusual decline were attributed to the intended use of proceeds, it would imply that every RMB100 raised destroyed approximately RMB185 of value; even a total loss of all proceeds would explain only about US$10 billion. JPMorgan therefore interprets most of the decline as the market repricing the duration and scale of the entire AI capital-expenditure program rather than the financing directly causing an equivalent loss of value. The report does not change its forecast for Alibaba's FY27E capital expenditure, which remains RMB200 billion, and treats the roughly RMB270 billion annualized pace in the June quarter as a stress case. What has changed is the assessment of cycle duration: raising US$10.3 billion in permanent capital five days after the earnings release indicates that management expects elevated investment to continue for more than one procurement cycle. At RMB200 billion of capital expenditure, Alibaba has an annual internal funding gap of approximately RMB100 billion; under the stress case, the gap is approximately RMB170 billion. The share-placement proceeds can cover about 8.4 months of the base-case gap and about five months of the stress-case gap. The company also has approximately RMB150 billion of discount-adjusted available liquidity and substantial debt capacity, with borrowing costs of about 2% to 3%, so the report believes the core issue is neither liquidity nor debt-servicing capacity. Net bank borrowings increased by RMB11 billion in the June quarter, while combined current and non-current prepayments rose by approximately RMB69 billion, potentially including prepayments for chips and computing capacity, indicating that actual cash commitments are advancing faster than reported capital expenditure. The report expects additional debt financing over the next 12 months and potentially financing at the cloud-business level; another parent-company share placement would be viewed as a negative signal that the return outlook has weakened. JPMorgan evaluates Alibaba's share placement through four independent tests. The first is whether computing infrastructure creates economic value. The model assumes that each RMB1 of capital expenditure generates RMB1 of annual revenue at full load, with a 45% incremental EBITDA margin, a 20% tax rate, a five-year economic life, straight-line depreciation, and utilization ramping through 60%, 90%, and 100%. Each RMB100 of capital expenditure produces after-tax cash flows of negative RMB100, RMB25.6, RMB36.4, RMB40, RMB40, and RMB40, yielding a 22% project IRR, a 2.9-year payback period, and RMB36 of net present value when discounted at an approximately 10% cost of capital. This is consistent with management's stated payback period of less than three years and blended return on invested capital above 13%. However, not all proceeds are allocated to infrastructure: training, application inference, and customer-acquisition expenses are recorded under AI Labs and Applications, a business that currently incurs a quarterly loss of RMB13.9 billion and has no auditable return. The report therefore uses 60% of proceeds allocated to infrastructure as its base case and assigns no measurable return to the remaining 40%; under this blended allocation, every RMB100 raised still generates approximately RMB22 of net present value. The second test is earnings per share. If at least 80% of the proceeds are used for infrastructure, the transaction becomes EPS-accretive from the second year onward; under the base-case allocation of 60% to infrastructure, FY28E EPS is diluted by approximately 0.5% and is broadly neutral thereafter. Expressed in return-on-invested-capital terms, once most of the proceeds have been deployed, third-year accretion requires a mature ROIC of approximately 15%; prolonged idling of the funds would cause dilution under all scenarios. The third test is the financing method. The report estimates that a debt-financing alternative would produce EPS about 3 to 3.5 percentage points better than the share placement in every period because the 2% to 3% coupon on RMB bonds is far below the stock's approximately 12% earnings yield. The rationale for choosing equity is not near-term EPS, but obtaining offshore currency for procurement, preserving ratings and covenant headroom, and securing permanent capital that does not require refinancing during the capital-expenditure cycle while consolidated free cash flow is negative. These benefits are real, but they are not costless. The fourth test focuses on the intrinsic value per share for existing shareholders. Based on a pre-transaction fair value of HK$205, if all proceeds are invested in infrastructure, post-transaction value per share is HK$203.1, down 0.9%; under the 60% infrastructure allocation, value per share is HK$202.6, down 1.2%. The transaction becomes accretive to per-share value only if pre-transaction fair value is below approximately HK$153 and 100% of the proceeds are invested in infrastructure, or below approximately HK$137 under the 60% allocation. The report's conclusion is therefore not that the share placement is entirely free, but that its approximately 1% transfer of intrinsic value and conditional EPS impact are far too small to explain the approximately US$19 billion unusual revaluation. The three companies have made similar strategic judgments on AI but use different financing models. Following its share-price revaluation, Alibaba uses parent-company equity supplemented by debt, at the cost of direct dilution. Tencent relies on operating cash flow, bonds, and an RMB875 billion investment portfolio: first-half operating cash flow was RMB154 billion, or about RMB310 billion annualized, against the report's estimated RMB200 billion of capital expenditure, RMB41.6 billion of dividends paid, and approximately RMB45 billion of share repurchases. Quarterly net cash declined from RMB146.9 billion to RMB58.2 billion, mainly because dividends coincided with RMB51 billion of computing-capacity prepayments. Tencent also issued US$2.45 billion and RMB15 billion of long-term notes in June, while short-term borrowings increased to RMB86 billion and accounts payable rose from RMB121 billion last December to RMB176 billion. Supporting this is a portfolio comprising RMB487 billion of listed investments and RMB388 billion of unlisted investments. The report believes Tencent's risk of parent-company equity financing is close to zero, but divesting investments would forgo assets and returns that the market may undervalue and would incur tax leakage; whether this is justified depends on whether the approximately 20% capital-expenditure ROIC can clearly exceed the portfolio's high-single-digit returns. Baidu's June-quarter capital expenditure was RMB11.4 billion, while operating cash flow was only RMB3.4 billion, with funding mainly coming from RMB17.4 billion of financing inflows. Short-term loans increased from RMB7.6 billion to RMB26.3 billion, long-term loans from RMB3.4 billion to RMB20.8 billion, and right-of-use assets under leases doubled from last December to RMB18.6 billion. After discounting long-term deposits and excluding iQIYI, approximately RMB100 billion of liquidity can still cover the base-case funding gap for years. The report believes Baidu's constraint is not cash, but its low parent-company valuation and unwillingness to dilute the parent directly; its equity-financing path is at the Kunlunxin subsidiary level. This approach reduces parent-company dilution but requires existing shareholders to surrender part of their interest in a core asset embodying the AI investment thesis. The resulting financing principles are as follows: parent-company equity can be issued when the share price has rerated and project IRR exceeds the stock's earnings yield; an investment portfolio can be divested when its size exceeds project funding needs; and financing can be raised at the subsidiary level when the parent is undervalued and AI assets are concentrated in a subsidiary. “Not issuing parent-company equity” does not mean there is no economic cost. Expanding capital expenditure does not necessarily increase returns proportionally; unit economics determine the outcome. If utilization and pricing are squeezed by expanding supply, project IRR declines rapidly: most sensitivity scenarios remain above the 10% cost of capital, but IRR is only approximately 11% when 80% utilization is combined with a 10% price decline, and approximately 3% when 70% utilization is combined with a 20% price decline. RMB270 billion is therefore a risk case rather than an upside case. Under the same cohort economics model, total depreciation associated with one year of base-case capital expenditure is equivalent to 28% of Alibaba's FY27E net profit, 13% of Tencent's, and 23% of Baidu's. However, incremental EBITDA exceeds depreciation from the first full year onward, so the report believes the real income-statement risks are EBITDA, utilization, and pricing rather than depreciation itself. In terms of relative positioning, Tencent's higher valuation reflects an internally fundable investment plan and an RMB875 billion investment portfolio; Alibaba has removed some financing uncertainty, while the market's pricing of the negative signal from the share placement appears excessive to the report, with the stock valued at approximately 12 times FY27E earnings and 8 times FY28E earnings; Baidu's financing is more likely to occur at the Kunlunxin level, so parent-company dilution risk remains low. The report cites BYD's and Xiaomi's March 2025 share placements at 6% to 8% discounts, after which their shares initially declined before the moves were absorbed by the prior upward trends, only as background and not as a valuation anchor for Alibaba. JPMorgan accordingly maintains Overweight ratings on all three companies and recommends gradually building positions during pullbacks ahead of Alibaba's September-quarter earnings release.

Analysis framework

The report first separates the arithmetic of Alibaba's placement discount, peer-market movements, and unusual revaluation, and then distinguishes annual internal funding surpluses or gaps, immediate liquidity, incremental debt or equity capacity, and the coverage period provided by each funding pool. It subsequently uses a project cash-flow model to test the net present value and internal rate of return of computing-capacity investment, while separately evaluating EPS, the differences between debt and equity financing, and intrinsic value per share for existing shareholders. Finally, it compares Alibaba's, Tencent's, and Baidu's cash flows, balance sheets, and financing paths and performs sensitivity analysis on utilization, pricing, use of proceeds, and capital-expenditure scale.

Methodology notes

• Valuation MethodologyDCF Discounted Cash Flow

  Computing-capacity project cohort net present value, internal rate of return, and payback-period model

  The report forecasts annual after-tax cash flows for every RMB100 of computing-capacity capital expenditure and discounts them at an approximately 10% cost of capital, calculating a 22% project IRR, a 2.9-year payback period, and RMB36 of net present value to determine whether the infrastructure investment itself creates value.

• Corporate Fundamentals and Financial FrameworkROIC–WACC spread

  Comparison of mature return on invested capital with cost of capital

  The report compares project returns with an approximately 10% cost of capital and notes that third-year EPS accretion requires a mature ROIC of approximately 15%, assessing whether capital expenditure covers funding costs and equity dilution.

• Corporate Fundamentals and Financial FrameworkFree cash flow analysis

  Analysis of internal funding gaps, liquidity, and financing coverage periods

  The report subtracts uses such as capital expenditure, dividends, and share repurchases from operating cash flow, then incorporates available liquidity and financing capacity to measure the degree to which the three companies require external funding and how long their funding pools can last.

• Industry/Sector Analysis FrameworkVolume-price decomposition

  Sensitivity of computing-capacity utilization and cloud-computing pricing

  The report adjusts mature utilization and service pricing simultaneously to observe changes in project IRR, illustrating how supply, utilization, and pricing pressure weaken unit returns as capital expenditure expands.

• Valuation MethodologyPE/PEG valuation

  Relative positioning based on forward P/E

  The report uses Alibaba's approximately 12 times FY27E and 8 times FY28E P/E multiples to show that the market has already reflected substantial financing and capital-expenditure concerns and incorporates this into its relative-positioning assessment.

Asset mapping & comparison

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

• Alibaba (9988.HK/BABA)

  Funds AI capital expenditure through a parent-company share placement and subsequent debt; the financing has removed some funding uncertainty but creates direct dilution.

  Strengths

  Approximately RMB150 billion of available liquidity, substantial debt capacity, positive net present value indicated by the computing-infrastructure model, and a current valuation of approximately 12 times FY27E and 8 times FY28E earnings.

  Weaknesses

  At RMB200 billion of capital expenditure, it has an annual internal funding gap of approximately RMB100 billion, while FY28E EPS is still diluted by about 0.5% under the 60% infrastructure allocation.

  Comparison

  Its structural funding gap is larger than Tencent's and Baidu's, and the cost of its share placement is the most visible and easiest for the market to price.

  Risks

  Slow deployment of proceeds, a lower infrastructure allocation, weaker utilization or pricing, expanded capital-expenditure commitments, and another parent-company share placement within 12 months.

• Tencent (0700.HK)

  Primarily supports AI investment through operating cash flow, bonds, and investment-portfolio disposals; the report believes the risk of a parent-company share placement is close to zero.

  Strengths

  First-half operating cash flow of RMB154 billion and an additional RMB875 billion portfolio of listed and unlisted investments provide the most balanced funding sources.

  Weaknesses

  Investment-portfolio disposals would forgo potential returns and incur tax leakage, while capital expenditure, dividends, and share repurchases occurring simultaneously would reduce net cash.

  Comparison

  Compared with Alibaba's direct dilution, Tencent's higher valuation is accompanied by internal funding capacity and support from a large investment portfolio.

  Risks

  If capital expenditure, dividends, and share repurchases in a single quarter collectively exceed operating cash flow without corresponding asset disposals, the funding balance will weaken.

• Baidu (9888.HK/BIDU)

  Relies on borrowing and plans to raise subsidiary-level equity financing through Kunlunxin to avoid diluting the parent company at a low valuation.

  Strengths

  Approximately RMB100 billion of adjusted liquidity can cover the base-case funding gap for years, and the risk of a direct parent-company share placement is low.

  Weaknesses

  Quarterly operating cash flow is below capital expenditure, creating high reliance on additional borrowing; subsidiary financing will surrender part of the equity in a core AI asset.

  Comparison

  Its self-funding capacity is weaker than Tencent's, but unlike Alibaba, its equity financing is more likely to occur at the AI-subsidiary level.

  Risks

  Kunlunxin's listing is delayed beyond 2026, or the actual financing amount is materially insufficient to support the announced capacity plan.

Key data

• Alibaba share-placement size710 million shares, HK$80 billion (US$10.3 billion)Issue price of HK$112.70, an 8.4% discount to the prior closing price, with the new shares equivalent to 3.7% of the existing share count
• Alibaba's August 25 share-price performance-8.5%Tencent declined 2.2% and the Hang Seng TECH Index fell 2.1% over the same period
• Unusual value revaluationApproximately US$18 billion to US$19 billionApproximately twice the cash raised; discount cost of about US$0.9 billion, with peer movements explaining about US$6.7 billion
• Alibaba FY27E capital expenditureRMB200 billionForecast unchanged; the RMB270 billion annualized level is treated as a stress case
• Alibaba's annual internal funding gapBase case approximately RMB100 billion; stress case approximately RMB170 billionThe share placement can cover approximately 8.4 months and five months, respectively
• Computing-capacity project returns22% project IRR, 2.9-year payback period, and RMB36 of net present value per RMB100For infrastructure funding, with an approximately 10% discount rate
• Base-case use-of-proceeds mix60% infrastructureWhen no measurable return is assigned to the remaining funds, every RMB100 raised still generates approximately RMB22 of net present value
• Base-case EPS impactFY28E dilution of 0.5%Broadly neutral thereafter; accretive from the second year when infrastructure accounts for 80% or more
• Relative EPS advantage of debt financingApproximately 3 to 3.5 percentage pointsRMB bond coupons are approximately 2% to 3%, below the stock's approximately 12% earnings yield
• Impact on Alibaba's intrinsic value per shareDecline of 0.9% to 1.2%Based on pre-transaction fair value of HK$205, post-transaction value per share is HK$203.1 or HK$202.6
• Tencent investment portfolioRMB875 billionIncluding RMB487 billion of listed investments and RMB388 billion of unlisted investments
• Baidu's June-quarter cash investmentCapital expenditure of RMB11.4 billion and operating cash flow of RMB3.4 billionReceived RMB17.4 billion of financing inflows during the same period
• Project-return stress casesIRR of approximately 11% or 3%Corresponding respectively to 80% utilization with a 10% price decline and 70% utilization with a 20% price decline
• Depreciation relative to FY27E net profitAlibaba 28%, Tencent 13%, Baidu 23%Calculated based on total depreciation associated with one year of base-case capital expenditure and the same cohort economics model

Impact & implications

The report believes the more important implication of Alibaba's share placement is an extended AI capital-expenditure cycle, rather than a debt-servicing crisis or an inevitable destruction of substantial value by the proceeds. The three companies' financing paths express their costs differently: direct dilution for Alibaba, investment-portfolio opportunity costs and tax leakage for Tencent, and the surrender of equity in an AI subsidiary for Baidu. Alibaba's financing therefore does not directly imply that Tencent or Baidu will also conduct parent-company share placements. Subsequent returns will depend on the proportion of proceeds actually invested in infrastructure, deployment speed, cloud-business growth, utilization, and pricing.

Risks

• If the mature ROIC on Alibaba's use of financing proceeds falls below approximately 15%, the report's conclusions regarding accretion and value creation will weaken.
• If infrastructure's share of AI spending falls to approximately 40%, more funds will flow to training, application inference, and customer-acquisition expenses that currently have no auditable return.
• If Alibaba's three-year capital-expenditure commitment materially exceeds RMB380 billion without raising its external-cloud growth guidance above 50%, the investment-return thesis will come under pressure.
• If most of the financing proceeds remain undeployed for more than 12 months and continue to sit in treasury balances rather than fixed assets and prepayments, the transaction will continue to dilute earnings.
• If losses at AI Labs and Applications do not decline sequentially in the September quarter, it may indicate that infrastructure's share of AI spending is lower than expected.
• If easing computing-capacity supply causes external-cloud prices to decline, utilization and pricing pressure could bring project IRR close to or below the cost of capital.
• If Alibaba conducts another parent-company equity financing within 12 months, the report would view it as a negative signal for investment returns.
• If Tencent's capital expenditure, dividends, and share repurchases in a given quarter collectively exceed operating cash flow without corresponding investment disposals, its internal funding balance may be disrupted.
• If Kunlunxin's listing is delayed beyond 2026 or the financing amount is materially insufficient, Baidu's ability to expand AI capacity will be constrained.

What to watch

• Monitor whether technical selling pressure subsides after Alibaba's share placement settles on August 26, 2026.
• Monitor the impact of Baidu's extraordinary general meeting on the same day on arrangements related to southbound trading.
• Alibaba's September-quarter results are the main test, with a focus on whether external-cloud growth can exceed 50%, whether AI Labs losses can decline sequentially, and whether capital expenditure aligns with the advance financing.
• Observe Alibaba's use of proceeds over the next two quarters to verify the infrastructure allocation and deployment speed required for EPS accretion.
• Monitor whether Alibaba increases debt or introduces financing at the cloud-business level over the next 12 months, and whether it conducts another parent-company share placement.
• Monitor evidence from Tencent's Hunyuan 4 regarding the effectiveness of AI investment, as well as the quarterly balance among capital expenditure, dividends, share repurchases, and operating cash flow.
• Monitor the timing and financing size of Kunlunxin's listing, as this event will change Baidu's funding capacity for expanding AI capacity.