Tencent's second-quarter results revealed a stark strategic shift: the company is pouring profits into compute infrastructure, and the immediate cost is visible in its cash flows. Revenue rose 11% year on year to ¥204.8bn ($30.4bn), beating expectations, but net profit of ¥56bn fell short of the ¥61.8bn analysts had projected, according to CNBC. The decisive number was capital expenditure, which jumped 176% from the same period last year to ¥52.8bn, roughly $7.8bn. That splurge swung free cash flow to an outflow of ¥13.8bn, a rarity for a company that has long been a cash generator.
These figures describe one decision: Tencent is channeling its earnings into building AI compute capacity and asking investors to wait for the payoff. The company is not simply upgrading servers; it is reshaping its entire business model around artificial intelligence, with all the risks and potential rewards that entails. For a firm known for the reliability of its gaming and advertising revenues, this quarter marks a departure from the cautious playbook that has defined it for decades.
The number worth doing by hand
Put the two cash figures side by side. Tencent spent ¥52.8bn on capital expenditure in the quarter. Free cash flow landed at minus ¥13.8bn, as reported by Bloomberg. That means operations did not generate enough cash to cover the infrastructure bill and the company's other obligations. It is a striking headline for a company that has printed cash for twenty years. WeChat, games, advertising, and cloud services together could not cover the cost of compute in the quarter.
Analysts and investors have been parsing this for signs of whether the aggressive spending is prudent or reckless. Tencent's leadership frames it as a necessary investment in the future, pointing to opportunities in generative AI, cloud services, and advertising technology. The company has not given a multiyear capital expenditure target, unlike Alibaba, but it has said its AI product investments will double this year. That ambiguity has left the market guessing about the long-term trajectory of both spending and returns.
The operating business, however, remains healthy. Domestic games revenue grew 17% to ¥47.3bn, driven by titles such as Delta Force and Valorant, a sharp acceleration from the 6% growth seen in the first quarter. Marketing services revenue rose 22% to ¥43.6bn, which Tencent attributes to AI-driven ad targeting and better measurement. Cloud revenue grew in the low twenties, and the company raised prices for some services. International games slipped 0.8% on a currency-neutral basis but grew 4% in constant currency terms, a mixed result reflecting foreign exchange headwinds.
Two ways to say the same figure
The 176% year-on-year increase in capital expenditure is the most dramatic number, and the South China Morning Post led its coverage with it. But the quarter-on-quarter increase is also telling: capital expenditure rose 65% compared with the first quarter. Both figures come from the same earnings release, and the gap between them shows how quickly the company is accelerating its spending. Most media outlets picked one or the other, but the progression matters more than the single-period snapshot.
Operating costs climbed 22.6% in the quarter, according to Bloomberg Intelligence. That outpaced revenue growth, which is why adjusted operating profit growth slowed to 9.2%. Adjusted net income reached ¥68.4bn, up 9% and roughly in line with market expectations, but the rising cost base reflects the heavy investments in hardware, data centers, and model development. Tencent is also spending on research and development for its own AI models, including the newly released Hy3 and the scheduled Hy4 later this year.
The company has expanded its compute capacity at a pace rarely seen in its history. Executives have mentioned partnerships with chip suppliers, including a reported $3bn memory deal with CXMT, a Chinese memory chipmaker. This aligns with a broader push by Chinese tech giants to secure domestic supply chains amid US export controls. Tencent is not just buying off-the-shelf hardware; it is working to ensure a stable and affordable supply of advanced components, which could give it a competitive edge over rivals that are more dependent on imported chips.
The fallback plan is becoming a landlord
During the earnings call, analysts pressed for details on how Tencent would achieve returns on its massive outlays. The answers were unusually specific. Chief strategy officer James Mitchell said Tencent could achieve a “decent return in an immediate timeframe” if it simply rented out all of its compute capacity. That would essentially turn the company into a cloud infrastructure provider, a real and profitable business. But Tencent is choosing to build its own models instead, for what Mitchell called “superior economic returns over the longer term”.
President Martin Lau reinforced the same point but framed it as a floor rather than a ceiling. In the worst case, which he stressed the company does not expect, the infrastructure can be rented out at cost recovery. “There is also clear downside protection,” he told analysts. In other words, the building boom is not a gamble that could wipe out shareholders; even if the AI models fail to outperform, Tencent can sell compute capacity to other companies and recover its costs. That reassurance is meant to calm investors who fear the company is entering an arms race with no clear finish line.
Read that plainly: the downside case for a $7.8bn quarterly compute bill is that Tencent becomes a neocloud. That is a viable business, offering server rental and AI processing services, but it would be a very different company from the one known for game skins, advertising, and consumer apps. The shift toward infrastructure could dilute the consumer focus that has long been Tencent's strength. The company is betting that its integrated approach, combining consumer apps, gaming, advertising, and cloud, will generate higher returns than being a pure infrastructure provider.
Punished for spending, punished for not committing
Tencent finds itself in a bind. Alibaba has pledged more than $50bn over three years for AI infrastructure, sending a clear signal to the market that it is all-in on artificial intelligence. Tencent, by contrast, has set no multiyear target. It has said only that investment in AI products will double this year. That vagueness has given the market room to interpret the spending as both reckless and insufficient, depending on the observer.
The stock is down 26% this year, erasing roughly $170bn in market value. Part of the decline reflects worries that the capital expenditure is uncontrolled and may not produce returns for years. Another part reflects the opposite worry: that Tencent is not spending enough and lacks a flagship AI model at the frontier of the industry. Among China's biggest tech companies, Tencent is the only one without a model widely recognized as best-in-class. Moonshot's Kimi K3 matched industry leaders on far fewer resources, according to benchmarks. Alibaba's Qwen3.8-Max topped several major charts. Tencent has responded by hedging: it participated in DeepSeek's debut funding round and integrated DeepSeek's V4 model across its own products.
The lack of a frontier model is a competitive disadvantage, but Tencent argues that its distribution advantage matters. WeChat, with 1.44 billion monthly users, is a massive distribution channel for AI assistants. The company's WorkBuddy product has become China's most popular AI office tool, with 21 million monthly visits in June. Xiaowei, the voice assistant inside WeChat, is still in a small-scale prototype test. These are early but not negligible efforts.
Tencent's gaming business remains a reliable cash cow, but it is not immune to competition. The company has been cutting studio investments in Japan and trimming staff at its LightSpeed and TiMi units, signaling a more disciplined approach to content development. International games revenue slipped slightly on a currency-neutral basis, though constant-currency growth was positive. The core WeChat ecosystem continues to expand slowly, with monthly active users up 2% to 1.44 billion, while QQ Mobile fell 2% to 520 million.
What is actually working
Despite the market's concerns, the operating business had a solid quarter. Domestic games revenue rose 17% to ¥47.3bn thanks to strong performances from Delta Force and Valorant, which have captured player attention and monetized well through in-game purchases and seasonal events. That accelerated sharply from the 6% growth in the first quarter, showing that Tencent's gaming pipeline remains healthy. Marketing services revenue rose 22% to ¥43.6bn, driven by AI-powered ad targeting that improves click-through rates and advertiser ROI. Tencent credits the ad growth to its ability to leverage WeChat's vast data trove and machine learning algorithms.
Cloud revenue also posted solid growth in the low twenties, which is notable in a highly competitive Chinese market. Tencent raised prices for some cloud services, indicating strong demand and pricing power. The company's focus on enterprise AI and hybrid cloud solutions appears to be paying off, though it still lags Alibaba Cloud in market share. The combined effect of these operations was not enough to cover the capital expenditure, but the underlying business is growing at a healthy clip.
Tencent's own AI products are early but showing promise. WorkBuddy, an AI office assistant, leads the Chinese market with 21 million monthly visits in June, according to data cited by the company. Xiaowei, the assistant embedded in WeChat, is being tested with a small user base. The company has also released the Hy3 model, a foundation model for natural language processing, with a larger successor Hy4 expected later this year. If Hy4 can compete with the best models from DeepSeek and Moonshot, Tencent would have a credible answer to its critics. If not, the compute bill may have bought a seat at someone else's breakthrough.
Why Amsterdam cared
The result moved a European stock. Prosus, the Amsterdam-listed group that holds a large Tencent stake, fell 6% on the day. That reaction shows how a quarterly report from Shenzhen can become a European market event, a link that is rarely visible until a day like this. Prosus derives a significant portion of its value from its Tencent holding, so any disappointment in Tencent's results or outlook directly affects its share price. The decline also reflects the broader sentiment toward Chinese tech stocks, which have been volatile amid regulatory changes and geopolitical tensions.
Tencent has been quietly acquiring the hardware needed for its AI push. Apart from the reported memory deal with CXMT, the company has been strengthening its relationships with semiconductor suppliers and building out data centers in key regions. Founder Pony Ma told analysts that the company is “making substantial progress toward building a new AI-empowered Tencent”. That vision, however, will take time to materialize. The immediate test arrives in about three months, when Tencent plans to launch Hy4, the larger successor to the Hy3 model it shipped last month. That launch will be a critical moment: either Hy4 competes with the best from DeepSeek and Moonshot, validating Tencent's spending, or it falls short, leaving the company with a massive infrastructure bill and no frontier model to show for it.
In the meantime, Tencent's leadership is trying to thread the needle between aggressive investment and financial discipline. The company has not announced a buyback dividend increase, focusing instead on reinvestment. That approach has unsettled some investors who preferred the old Tencent, which returned substantial cash to shareholders. But the management team argues that the AI transition is a once-in-a-generation opportunity, and missing it would be far costlier than overspending now.
The long-term picture is genuinely uncertain. Tencent could emerge as a dominant AI player, leveraging its massive user base and distribution strengths to deploy models at scale. Or it could find itself locked in a brutal price war with rivals that have deeper pockets or more advanced models. The discount to its historical valuation, trading at a multiple below the broader market, suggests that investors are skeptical. Yet Tencent has a track record of adapting to disruption, from mobile internet to cloud computing. The AI push is the biggest bet yet, and the second-quarter figures show just how much the company is willing to pay for a place at the table.