Alibaba and Tencent’s $18 Billion AI Spending Spree Signals China’s Compute Race Has Begun

Alibaba and Tencent’s second-quarter results offer a clear signal about China’s AI strategy: the competition is no longer centered only on which company releases the strongest model. It is increasingly about who can secure enough computing power, chips, data-centre capacity and cloud distribution to train and run AI systems at scale.

Together, the two companies spent about 120.5 billion yuan on capital expenditure in the quarter—roughly $18 billion. 52 That is a remarkable amount for a single three-month period, and it shows that China’s largest technology platforms are willing to accept near-term pressure on profits and cash flow in pursuit of a longer-term infrastructure advantage.

The numbers reveal a deliberate trade-off

Alibaba spent 67.68 billion yuan on capital expenditure in the quarter, up 75% year over year. Its net profit fell 75%, even as revenue increased 9%. 33 The result is a stark illustration of the company’s choice: preserve short-term profitability, or build capacity while demand for AI cloud and computing services is accelerating.

Tencent’s capital expenditure reached 52.8 billion yuan, up 65% from the previous quarter and 176% from a year earlier. The figure also exceeded an analyst estimate of 32.1 billion yuan cited in reporting on the results. 2223 Tencent’s revenue rose 11% to 204.8 billion yuan, but net profit increased only 0.7% to 56 billion yuan. 17

The combined picture is not simply that both companies are spending more on research. They are committing capital to physical and commercial infrastructure that can support model training, inference and cloud services over several years.

Alibaba is betting that cloud demand will pay for the buildout

Alibaba’s quarter provides some evidence that its infrastructure is already connected to customer demand. Reuters reported that cloud revenue rose 45% as AI demand supported the business. 33 Other reports use different definitions for Alibaba’s AI and cloud categories, so the precise growth rate should be treated carefully; the consistent signal is that the cloud business is growing while investment is surging.

That growth does not eliminate the financial risk. Alibaba’s profit decline shows how expensive the expansion is before the additional capacity produces a full return. The company has therefore made a classic infrastructure bet: spend ahead of demand, capture customers and usage, and recover the investment over time.

The strategy also gives Alibaba several ways to monetize the same asset base. Computing capacity can support its own models and applications, serve enterprise customers through Alibaba Cloud, and potentially be sold as additional capacity when internal demand does not use the entire system. The more customers depend on that infrastructure, the stronger Alibaba’s position may become—but only if utilization and pricing remain high enough to justify the spending.

Tencent’s quarter shows both monetization and cash-flow pain

Tencent’s figures make the cost of the AI buildout especially visible. Heavy infrastructure investment pushed free cash flow to a negative 13.8 billion yuan in the quarter. Tencent said the cash flow included large AI-related prepayments for infrastructure supporting its models and services. 30

At the same time, Tencent is not investing without any commercial evidence. Its advertising business benefited from AI-driven improvements: marketing-services revenue rose 21.8% year over year to 43.565 billion yuan, according to reporting on the results. 18 Reuters likewise described advertising gains as a driver of the quarter’s revenue growth. 17

That matters because advertising gives Tencent a nearer-term route to recover AI investment. Better recommendation, targeting, creative tools and conversion can improve an established business before standalone AI products become a major revenue stream. The gains are encouraging, but they do not yet demonstrate that the full infrastructure program will earn an attractive return.

Why compute has become a strategic asset

The companies’ spending suggests that computing capacity is being treated as a bottleneck with strategic value. A shortage of chips, servers, power and data-centre capacity can limit the speed at which a company improves its models or serves customers, regardless of how strong its software may be.

That creates an incentive to secure capacity early. A company that waits for demand to become obvious may find that the necessary hardware is unavailable or more expensive. By investing sooner, Alibaba and Tencent can try to protect their own AI roadmaps while also building cloud platforms capable of selling computing services to other businesses.

This is why the spending should not be read as ordinary product development alone. It is also a bid for future infrastructure revenue, customer relationships and technological flexibility. The downside is that capacity bought too early can become underutilized, while falling prices or more efficient models can reduce the value of the hardware.

ByteDance adds pressure to the arms race

The competitive field extends beyond the two listed companies. Bloomberg reported that ByteDance was discussing capital expenditure of as much as $70 billion in 2026 for data centres and other AI infrastructure. 1 That figure was described as a consideration under discussion, not a confirmed final budget, so it should not be treated as guidance.

Even as an unconfirmed scenario, the report illustrates the strategic pressure facing China’s largest platforms. If rivals build substantially more capacity, a company that spends cautiously could lose access to compute, developers, enterprise customers or distribution channels. That makes the decision asymmetric: a few quarters of weaker cash flow may appear more manageable to management than falling permanently behind in infrastructure.

The central question is still return on investment

The spending surge is evidence of conviction, not proof of success. Alibaba’s cloud growth and Tencent’s advertising gains show that AI is already supporting parts of their existing businesses. They do not establish that the companies can generate sufficient revenue from the much larger pool of chips, data centres and computing capacity now being assembled.

Several risks remain:

• Demand could lag capacity. If enterprise adoption or consumer usage grows more slowly than expected, expensive infrastructure may be underused.
• AI services could become commoditized. Stronger open models or more efficient systems could reduce the prices that cloud providers can charge.
• Hardware constraints could raise costs. Chip availability, export restrictions and supply-chain limitations can make each unit of usable compute more expensive.
• Profit pressure could last longer than expected. Management’s investment timelines are targets, not guarantees, and the payback period depends on utilization, pricing and continued AI demand.

The most defensible conclusion from the second-quarter results is therefore not that Alibaba or Tencent has already won. It is that both companies believe compute access is important enough to justify unusually heavy upfront spending—and that China’s AI competition is entering a capital-intensive phase.

For investors and technology companies, the next test is conversion. The key metrics will be whether cloud usage, advertising performance and AI applications grow quickly enough to absorb the new capacity, and whether those revenues eventually translate into sustainable free cash flow rather than simply larger infrastructure bills.