Alibaba Group just posted one of the sharpest cash flow declines in its history, and the reason is straightforward: the company is spending enormous sums to build out AI and cloud infrastructure at a pace it has never attempted before. In the quarter ending March 31, 2025, Alibaba’s free cash flow plunged 76% year-over-year, driven almost entirely by data center construction and computing capacity for AI workloads. For the full fiscal year, free cash flow fell 53%. Meanwhile, Chinese AI startup DeepSeek has released its V4 model, intensifying a domestic technology race that is forcing the country’s largest companies to spend more while earning less.
Alibaba’s financial squeeze
The March quarter numbers were dramatic, but they did not arrive in isolation. In the December 2025 quarter, Alibaba’s revenue grew just 2% year-over-year while adjusted EBITA dropped 57%. That combination, near-flat revenue alongside collapsing profitability, illustrates the core tension in Alibaba’s strategy: the company is absorbing massive upfront costs for AI infrastructure while the revenue those investments are supposed to generate has not yet materialized at scale.
Alibaba’s leadership has been explicit about the bet it is making. In its annual report filed with the SEC for the fiscal year ended March 31, 2025, the company stated that its planned investment in cloud and AI infrastructure over the next three years will exceed the total it spent on those categories in the previous decade. CEO Eddie Wu told analysts on the May 2025 earnings call that AI-related cloud revenue was growing at triple-digit rates, framing the spending as a land grab in a market where early infrastructure advantages could lock in enterprise customers for years. But the gap between that optimism and the current financial reality is wide. Investors are being asked to accept years of compressed margins on the promise of future payoffs that remain unquantified in Alibaba’s forward guidance.
Tencent feels the same pressure
Alibaba is not alone. Tencent Holdings, China’s other technology giant, reported its own version of the same squeeze when it released first-quarter 2025 results in May. Tencent’s capital expenditure surged as the company expanded GPU clusters and cloud infrastructure to support its AI ambitions, including its Hunyuan large language model and a growing suite of AI-powered tools integrated into WeChat and its enterprise services. Revenue growth in Tencent’s cloud and business services segment accelerated, but the company’s overall profit margins narrowed as infrastructure spending outpaced top-line gains.
Tencent’s situation differs from Alibaba’s in important ways. Its core gaming and social media businesses still generate substantial cash flow, giving it a financial cushion that Alibaba’s more commerce-dependent model does not provide to the same degree. But the direction is the same: both companies are channeling billions into AI infrastructure at a moment when returns on that spending remain uncertain, and both are doing so under competitive pressure that makes slowing down risky.
DeepSeek V4 raises the stakes
That pressure has a name. DeepSeek, the Hangzhou-based AI startup backed by quantitative trading firm High-Flyer, released its V4 model in recent weeks, as confirmed by the Associated Press. The release matters because DeepSeek has built its reputation on producing high-performance models at a fraction of the cost that larger rivals spend. Its earlier V3 and R1 models drew attention for matching or approaching the capabilities of systems from much larger organizations while requiring significantly less compute to train.
V4 represents the latest escalation. While independent technical benchmarks and detailed adoption figures have not yet been published, the model’s arrival adds a concrete competitive option for Chinese enterprises evaluating AI providers. For Alibaba’s Qwen model family and Tencent’s Hunyuan, DeepSeek’s cost-efficiency narrative is particularly uncomfortable: it suggests that brute-force infrastructure spending may not be the only path to competitive AI, and that a leaner challenger can keep pace with or outperform incumbents who are spending at historic rates.
The timing sharpens the tension. DeepSeek V4 arrived just as Alibaba’s earnings laid bare the financial cost of its own AI buildout, giving potential customers and investors a direct comparison between the incumbent’s capital-intensive approach and the startup’s more efficient one.
The chip constraint underneath it all
Both Alibaba and Tencent are making these investments under a constraint that does not apply to their American counterparts: U.S. export controls on advanced semiconductors. Washington has progressively tightened restrictions on the sale of high-end AI chips to China since late 2022, with additional rules finalized in 2024 and early 2025 that limit access to Nvidia’s most powerful GPUs and the equipment needed to manufacture comparable chips domestically. The restrictions have forced Chinese companies to rely on less advanced hardware, stockpile chips acquired before cutoff dates, or develop workarounds, all of which add cost and complexity to AI infrastructure projects.
DeepSeek’s efficiency-first approach is partly a response to this reality. Training capable models on less powerful hardware is not just a business strategy; it is a necessity in an environment where the most advanced chips are increasingly difficult to obtain. For Alibaba and Tencent, the chip constraints mean that their massive infrastructure spending may yield less compute per dollar than equivalent investments by U.S. or European rivals, making the financial math even more demanding.
What the spending race means for investors
The core question for anyone watching Alibaba and Tencent is whether the revenue will follow the spending. Alibaba’s management has pointed to triple-digit growth in AI-related cloud revenue, but that growth is coming off a small base and has not yet offset the margin compression from infrastructure investment. Tencent’s AI monetization is similarly early-stage, with most of its AI tools still being integrated into existing products rather than generating standalone revenue streams.
The three-year investment pledge in Alibaba’s annual report sets a clear timeline. If AI-driven cloud revenue does not accelerate meaningfully by fiscal year 2028, the company will have spent more than a decade’s worth of prior investment with limited financial return. Tencent faces a softer version of the same test, cushioned by its gaming and advertising businesses but still exposed to the risk that AI infrastructure becomes a cost center rather than a growth engine.
DeepSeek’s presence complicates the picture further. If the startup continues to deliver competitive models at lower cost, it could erode the pricing power that Alibaba and Tencent need to recoup their infrastructure investments. The incumbents are betting that scale, enterprise relationships, and integrated ecosystems will protect their positions. DeepSeek is betting that efficiency and speed can overcome those advantages. As of mid-2026, neither side has proven its case, and the financial stakes for both keep climbing.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
