China’s AI Spending Race: How Big Tech Is Turning Capex Into New Revenue

A structural guide to understanding how Alibaba, Tencent, Baidu, JD.com, and Kuaishou are deploying capital in the AI era

What Is This About?

In mid-2026, five of China's largest internet companies — Alibaba, Tencent, Baidu, JD.com, and Kuaishou — reported earnings covering the same three-month window: April through June 2026. Taken together, these five companies generated roughly RMB 887 billion (approximately USD 122 billion) in revenue during that single quarter, making them a representative cross-section of China's internet economy.

What made this earnings cycle notable was not just the scale of revenue, but where the money is going — and increasingly, where it is coming back from. AI has moved from a line item in R&D budgets to a structural force reshaping capital allocation, cost structures, and revenue models across the industry simultaneously.

Why This Moment Matters

China's generative AI adoption rate reached 42.8% of internet users as of December 2025, up from 36.5% just six months earlier, according to CNNIC's 57th Statistical Report on China's Internet Development. That means generative AI has crossed from early-adopter territory into mainstream daily use — a threshold that changes the competitive calculus for every major platform.

When adoption is marginal, AI is an experiment. When nearly half the online population is using it regularly, it becomes infrastructure. Platforms that are not yet monetizing AI are not just missing an opportunity — they are watching their cost base expand without a corresponding revenue offset.

Where Is the Money Going? The Capital Expenditure Surge

The most immediate and measurable signal in the Q2 2026 earnings was a dramatic acceleration in capital expenditure (capex), driven explicitly by AI infrastructure demand.

Alibaba reported revenue of RMB 268.95 billion for the quarter, up 9% year-on-year. Its capex reached RMB 67.68 billion — a 75% year-on-year increase — equivalent to roughly one-quarter of quarterly revenue. The company attributed the increase directly to expanding AI infrastructure to meet growing customer demand.

Tencent reported revenue of RMB 204.8 billion, up 11%, with capex of RMB 52.78 billion — a 176% year-on-year surge, also approaching one-quarter of quarterly revenue. Tencent disclosed that its operating cash flow included substantial AI-related prepayments covering model upgrades, its WorkBuddy and CodeBuddy productivity tools, WeChat AI features, and cloud infrastructure.

Combined, Alibaba and Tencent spent more than RMB 120 billion on capex in a single quarter. Not all of that is attributable to AI — both companies maintain large cloud, data center, and general technology infrastructure — but both companies explicitly linked the acceleration to AI compute demand.

Baidu illustrates the pressure this creates at smaller scale. With quarterly revenue of RMB 31.33 billion (down 4% year-on-year), Baidu's capex of RMB 11.39 billion represented approximately 36% of quarterly revenue — a higher ratio than either Alibaba or Tencent. Its operating cash flow was only RMB 3.4 billion; free cash flow was negative RMB 7.95 billion. Baidu's core online marketing revenue fell 19%, meaning the traditional cash engine that historically funded investment is under structural pressure at precisely the moment AI spending is accelerating.

Why Capex Ratios Matter

Capital expenditure on servers and data centers does not disappear after it is spent. It converts into long-term assets that generate depreciation charges and ongoing operating costs across multiple future quarters. A company that spends heavily on AI infrastructure today is committing to a higher fixed-cost base for years. This is why cash flow durability — not just current profitability — is becoming a primary competitive variable in the AI race.

The Structural Divide: Who Can Afford to Stay at the Table?

The capex surge has introduced a new form of competitive stratification that goes beyond model quality or engineering talent.

Alibaba and Tencent have diversified revenue streams — e-commerce, gaming, advertising, and cloud — that generate sustained cash flow. They can absorb multi-year infrastructure investment without existential stress. Baidu must sustain AI investment while its legacy advertising business declines. Mid-tier internet companies face an even starker version of the same problem: the absolute cost of competitive AI infrastructure is largely fixed regardless of company size, but the revenue base available to fund it is not.

This dynamic is consistent with broader market data. IDC reported that China's intelligent computing cloud infrastructure market reached RMB 48.67 billion in 2025, growing 128% year-on-year, with projections to exceed RMB 100 billion within two years. Globally, IDC projected AI infrastructure spending to reach USD 487 billion in 2026, up approximately 53% year-on-year. GPU server deployment was the primary driver of a 30.7% year-on-year increase in global server market spending in Q1 2026.

The practical implication: AI competition has evolved from a technology race into a capital endurance test. Technical leadership can shift with a single model release. Infrastructure, cash flow, and capital capacity accumulate over years and are far harder to replicate quickly.

How AI Is Generating Revenue: Two Emerging Models

Beyond the spending side, Q2 2026 earnings showed that AI is beginning to produce measurable, recurring revenue — through two structurally distinct paths.

Path One: Selling Compute and Cloud Infrastructure to Enterprises

The clearest signal came from Alibaba and Baidu, both of which operate major cloud businesses.

Alibaba Cloud's AI cloud and compute services generated RMB 48.437 billion in revenue during the quarter, with both total and external customer revenue growing 45% year-on-year. Within that, AI-related product revenue reached RMB 12.376 billion — marking 12 consecutive quarters of triple-digit year-on-year growth. Alibaba cited Omdia data placing its share of China's AI cloud market at 38.1% in 2025.

Baidu's AI-driven business revenue reached RMB 12.5 billion in Q2, up 25% year-on-year, accounting for half of Baidu's core AI business revenue. Within that, AI cloud infrastructure revenue was RMB 7.3 billion, up 50%; GPU cloud revenue specifically grew 283% year-on-year, accelerating from 184% the prior quarter. AI application revenue was RMB 2.5 billion, up only 3%.

The contrast within Baidu's own numbers is instructive. Enterprise customers are currently spending most aggressively on compute, cloud infrastructure, and model runtime environments — the picks-and-shovels layer. AI application revenue, which requires customers to integrate AI into specific workflows, is growing far more slowly. This reflects where enterprise AI adoption currently sits: most organizations are acquiring AI capability before they have fully determined how to deploy it.

Path Two: Vertical AI Production Tools for End Users

Kuaishou's Kling AI video generation product took a different route. Kling generated more than RMB 850 million in Q2 revenue, up over 200% year-on-year, representing more than 2% of Kuaishou's total quarterly revenue of RMB 35.5 billion.

Video generation is commercially legible in a way that general-purpose AI assistants often are not. The value proposition is concrete: a business or creator can calculate how much a generated video costs versus the equivalent production spend, and make a straightforward ROI decision. The payment pathway is short, and the use case is tied to an existing production workflow rather than requiring behavioral change.

This is consistent with IDC data showing that China's AI application public cloud services market reached RMB 13.73 billion in 2025 — already larger than the RMB 7.94 billion market for large model training and inference cloud services. Enterprise AI procurement logic is shifting from "acquire model capability" toward "solve a specific business problem."

The Deeper Layer: AI Embedded in Core Business Operations

The most structurally significant — and most easily underestimated — dimension of AI monetization is not new AI products. It is AI improving the economics of existing businesses.

Tencent's marketing services revenue reached RMB 43.565 billion in Q2, up 22% year-on-year — nearly double the company's overall revenue growth rate of 11%. Tencent attributed this directly to AI-driven upgrades in its advertising recommendation models, its AIM+ automated ad placement tool, and enhanced closed-loop marketing capabilities within the WeChat ecosystem. The revenue is recorded under "marketing services," not under any AI-specific line item.

This matters for how the industry should be measured. AI's contribution to Tencent's advertising business cannot be isolated in a single revenue figure. It shows up as higher click-through rates, higher advertiser willingness to spend, and improved conversion — all of which flow into a traditional revenue category. Tencent's marketing services share of total revenue rose from 19% to 21% year-on-year, a structural shift that AI is driving without being directly labeled.

JD.com demonstrated a parallel dynamic in commerce and supply chain. Despite total revenue of RMB 346.4 billion declining 2.9% year-on-year, JD Retail maintained an operating margin of 4.6% (RMB 13.5 billion operating profit). JD deployed AI across procurement, fulfillment, and customer service: during its 618 shopping festival, its JoyInside smart device platform had partnered with nearly 200 brands, with cumulative connected devices growing more than threefold versus the prior year's Double 11 festival. JD Industrial deployed more than 70 AI Agents across its procurement-to-fulfillment chain in the first half of 2026.

JD Health's upgraded "Dr. Dawei" AI medical service — integrating online consultation, home testing, nursing, and pharmacy — served nearly four times as many users during the 618 period year-on-year. The model does not just answer questions; it initiates service requests, connects to products, and coordinates follow-up — a workflow in which transaction and fulfillment determine whether AI value is actually realized.

Industry data supports this trajectory. IDC reported that 27% of Chinese enterprises have already deployed AI Agents in production. A separate global survey found that 50% of organizations have deployed Agents across multiple business functions, with another 27% operating them in at least one function. Enterprise expectations are shifting from "what content can AI generate" to "what tasks can AI complete."

Why Existing Business Moats Are Being Amplified, Not Erased

A pattern visible across all five companies is that AI is not homogenizing the competitive landscape — it is reinforcing existing structural advantages.

Alibaba's AI investment is concentrated in cloud and e-commerce infrastructure, where it already holds dominant market position. Tencent's AI returns are flowing through advertising and the WeChat ecosystem, which no competitor can replicate. Baidu is applying AI to search, its core legacy business. Kuaishou is monetizing AI through short video and creator tools, where its user base and content ecosystem provide distribution. JD.com is embedding AI in supply chain and logistics, where its physical infrastructure creates defensible differentiation.

The logic is straightforward: AI models trained on proprietary transaction data, user behavior, and operational workflows produce better outputs than generic models — and the data required to train them is embedded in years of business operations. A competitor cannot purchase that data advantage; it must be accumulated. As AI systems handle more concrete tasks in the physical and commercial world, the feedback loops they generate further compound the advantage of incumbents with dense, high-quality data.

This suggests that the next phase of AI competition in China will be determined less by model benchmarks and more by the depth of integration between AI capability and core business operations — and by which companies can sustain the capital investment required to keep building.

What to Watch Going Forward

Several variables will determine how this competitive landscape evolves:

Capital durability. Which companies can sustain RMB 50–70 billion quarterly capex for multiple years without compromising financial stability? The answer will narrow the field significantly.

Revenue mix shift. How quickly does AI-attributed revenue grow as a share of total revenue for each company? Tencent's advertising and JD's supply chain efficiency gains suggest the most durable AI monetization may be largely invisible in AI-specific metrics.

Enterprise Agent adoption velocity. The transition from AI as a content tool to AI as a task-completion system (Agents) is the next major adoption curve. Companies with existing enterprise relationships, workflow data, and transaction infrastructure are best positioned to capture this shift.

Model commoditization pressure. As foundation model capabilities converge, differentiation will increasingly depend on application layer depth, data quality, and distribution reach rather than raw model performance.

Regulatory and geopolitical constraints. Access to advanced GPU hardware remains a structural variable for all Chinese AI companies, with implications for both training capacity and the cost of compute services.

Related Coverage:

Alibaba and Tencent Pour RMB 120B in a Single Quarter Into AI as China's Infrastructure Race EscalatesJD.com Invests Over 10 Billion Yuan in Comprehensive Push into Embodied AI RoboticsBaidu's GPU Cloud Surges 283% as Advertising Slumps 19% in Q2Kling AI Revenue Surges 200% as Kuaishou Pays the Price for AI Leadership