Demand for chips and cloud capacity is accelerating, but Tencent, Baidu and SMIC show why investors must separate AI adoption from durable returns.
The development
China’s artificial-intelligence build-out has moved beyond the model-release contest and into a more demanding phase: proving that heavy spending can produce cash flow.
The latest earnings from Tencent, Baidu and Semiconductor Manufacturing International Corporation, or SMIC, show three different parts of the same cycle. Tencent is generating stronger advertising and cloud demand while sharply increasing capital expenditure. Baidu’s AI businesses are growing, but not yet fast enough to offset pressure in its legacy advertising engine. SMIC, meanwhile, is benefiting from tight domestic chip capacity and firmer pricing.
That combination is evidence of genuine demand. It is not yet proof that every company participating in the boom will earn attractive returns. The central question for investors is no longer whether China will spend on AI, but where the economics of that spending will settle.
The money is reaching infrastructure first
SMIC offers the clearest near-term signal. The foundry reported second-quarter revenue above $3 billion for the first time, up 36% from a year earlier, while profit attributable to owners more than tripled to $479 million. Capacity utilisation reached 93.7%, and both shipments and average selling prices increased.
The headline needs qualification. SMIC said AI demand was supporting orders, but much of the new volume is not limited to the most advanced processors used to train frontier models. AI systems also require power-management chips, connectivity components, storage controllers, sensors and other mature-node semiconductors. This broadens the industrial opportunity beyond one category of accelerator.
It also explains why China’s AI push is becoming intertwined with industrial policy. The State Council’s “AI Plus” programme is intended to spread AI into manufacturing, consumption, public services and other parts of the economy. When adoption moves from chatbots to factories, vehicles and robots, demand shifts from software alone toward chips, equipment, data centres, electricity and networking.
For suppliers, that can produce revenue before the final applications have proved their own profitability. It is the familiar economics of an infrastructure cycle: companies selling capacity may monetise earlier than businesses still searching for a winning consumer product.
Tencent shows both sides of the ledger
Tencent’s second-quarter revenue rose 11% to RMB204.8 billion. Marketing-services revenue increased 22%, helped by AI improvements to advertising recommendations, while fintech and business-services revenue rose 9% amid demand for cloud and AI services.
The cost of building that capability is rising faster. Capital expenditure reached RMB52.8 billion, up 176% from a year earlier. Tencent can fund the programme from strong gaming, advertising and payments businesses, but the outlay changes the investment case. Revenue growth alone is no longer enough; investors must ask how quickly AI spending lifts margins, engagement and free cash flow.
There are early signs of commercial value. Better ad targeting can raise the price and effectiveness of inventory already embedded in Weixin, while cloud customers can rent computing capacity rather than build it themselves. Those uses are economically more concrete than a stand-alone chatbot with an uncertain willingness to pay.
The risk is timing. Compute capacity is purchased before demand is fully visible, depreciation follows, and competition can reduce prices. If utilisation stays high and AI features improve Tencent’s existing products, the spending could strengthen a formidable platform. If monetisation lags, the same investment will weigh on cash generation.
Baidu illustrates the transition risk
Baidu’s results show why an AI label cannot erase weakness in the business funding the transition. Second-quarter revenue fell 4% to RMB31.3 billion, and online-marketing revenue declined 19%. Against that, revenue from Baidu Core’s AI-powered businesses reached RMB12.5 billion, while operating cash flow remained positive at RMB3.4 billion.
The AI figures are meaningful. Baidu reported rapid growth in GPU cloud revenue and continues to invest in models, autonomous driving and its Kunlun chip business. Yet the company is trying to build new revenue streams while search advertising faces structural pressure and a soft domestic demand environment.
That creates a harsher test than Tencent faces. Baidu must show not only that AI products can grow, but that they can become large and profitable quickly enough to replace a shrinking legacy pool. Investors should therefore focus on the mix and quality of revenue, not simply the growth rate of a newly defined AI category.
What the headline misses
China’s advantage may be less about matching the absolute volume of US private AI investment and more about applying AI across a large manufacturing base. Stanford’s 2026 AI Index estimated that US private AI investment reached $285.9 billion in 2025, compared with $12.4 billion in China. That comparison is important, but incomplete: it excludes the full effect of state support, corporate capital expenditure and policy-directed infrastructure.
The constraint remains access to leading-edge compute. Export controls have encouraged Chinese companies to use domestic chips, improve model efficiency and develop specialised systems, but substitution is not frictionless. Domestic foundries still face limits in advanced manufacturing equipment, yields and scale. A shortage can support supplier pricing in the short run while slowing customers’ deployment.
There is also a measurement problem. Companies increasingly describe products, cloud workloads and business lines as AI-related. Definitions differ, and high percentage growth can start from a small base. Reported AI revenue should be tested against total revenue, margins, capital intensity and cash conversion.
Why it matters for investors
The investable lesson is to treat China’s AI sector as a value chain rather than a single theme.
Infrastructure suppliers may benefit first from scarcity and capacity expansion, but they carry high fixed costs and policy risk. Cloud platforms have recurring-revenue potential, yet require sustained investment and face price competition. Internet platforms can use AI to improve advertising, commerce and productivity, although returns depend on whether those gains exceed the additional compute bill. Robotics and autonomous systems offer a larger long-term market, but commercial deployment remains uneven.
For Asia-focused investors, the effects extend beyond mainland technology shares. Semiconductor equipment, memory, data-centre power, cooling, optical networking and contract manufacturing across the region may gain from higher capital spending. Singapore-listed companies are more likely to feel the theme indirectly through data-centre infrastructure, banks financing expansion and industrial supply chains than through direct exposure to Chinese foundation-model developers.
None of those links guarantees attractive returns. Valuation, currency, regulation, export controls and access to foreign capital can overwhelm sound operating progress. Company quality and investment price remain separate questions.
What to watch next
Alibaba’s next earnings report will provide the nearest test of whether rapid AI-cloud growth is continuing. In its March quarter, cloud revenue rose 38% to RMB41.6 billion and external-customer revenue increased 40%, driven partly by AI products.
Beyond one quarter, four indicators matter: cloud-revenue growth relative to capital expenditure; utilisation and pricing at domestic foundries; evidence that AI lifts advertising or commerce margins; and the pace at which robots and autonomous systems move from demonstrations into paid deployment.
The downside case is a prolonged spending race in which capacity grows faster than profitable demand. The more constructive case is that cheaper domestic compute and efficient models accelerate adoption across China’s industrial base, spreading returns beyond the largest platforms.
The bottom line
China’s AI industry is no longer a speculative story resting only on policy ambition or model benchmarks. Revenue is appearing in chips, cloud services and AI-enhanced advertising. But the latest results also show that the benefits are uneven and the cash requirements substantial.
Finance World’s read is that the strongest evidence currently sits in infrastructure demand and in AI features attached to established businesses. The weakest part of the case is the assumption that fast adoption will automatically produce high returns for every participant. Investors should follow cash conversion and utilisation—not the number of models launched—to judge whether China’s AI build-out is creating durable economic value.