On August 12, Tencent released its financial results for the second quarter of 2026.
The company generated RMB 204.8 billion (USD 30.3 billion) in revenue during the quarter, up 11% year-on-year (YoY). Operating profit rose 12% to RMB 67.3 billion (USD 10 billion), while non-IFRS operating profit increased 9% to RMB 75.6 billion (USD 11.2 billion). Excluding contributions from newer artificial intelligence products, including Hy, Yuanbao, CodeBuddy, WorkBuddy, and Xiaowei, further adjusted non-IFRS operating profit would have reached RMB 86.1 billion (USD 12.7 billion), up 19%.
Tencent’s core businesses already have mature business models and a track record of steady growth. Investors are therefore paying particular attention to its AI business and the capital expenditure required to support it. On the product side, Hy3 has attracted attention for its price-performance ratio, while WorkBuddy has recorded a sharp increase in users.
Tencent has also stepped up its investment in AI. R&D expenses reached RMB 27.3 billion (USD 4 billion) during the quarter, up 35% YoY. Capital expenditure came to RMB 52.8 billion (USD 7.8 billion), up 176% from a year earlier and 65% from the previous quarter. That was consistent with management’s earlier guidance that capex growth would become more pronounced in the second half of the year. After capital expenditure, Tencent’s free cash flow also turned negative for the quarter, an unusual occurrence for the company.
The day after Tencent reported its results, on August 13, its shares fell sharply at the open. The stock dropped as much as about 4.7% intraday before closing down 4.46%.
The post-earnings decline reflected investor concerns that Tencent’s sharp increase in AI spending could weigh on profit and cash flow in the near term, while returns from AI monetization have yet to fully materialize.
Stepping up capex for compute and memory
The rationale behind Tencent’s higher capital expenditure is relatively straightforward.
Its core businesses, including gaming and advertising, already have mature business models. They remain sizable and continue to grow steadily, but they are unlikely to return to the rapid growth rates they once delivered.
From a long-term perspective, Tencent is therefore using the cash generated by those businesses to invest in AI. The goal is to develop a new growth engine while also strengthening its existing products and services.
The financial data illustrate the challenge. Over the past five years, YoY revenue growth in Tencent’s gaming business has remained in the single digits for most quarters, exceeding 10% in only five of them. Given the current scale of Tencent’s gaming business, the 8% YoY growth recorded this quarter is still substantial. But it falls short of the kind of sustained growth investors may seek from a large consumer technology company.
Advertising has maintained stronger growth over a longer period. Beyond higher advertiser spending, however, part of that growth has come from incremental increases in ad load. Because Tencent also has to balance monetization with user experience, there is a limit to how far ad load can rise. As that ceiling approaches, advertising revenue could face greater constraints on further expansion.
The next question is where Tencent’s capital expenditure is going.
First is infrastructure. Management said during the earnings call that the company significantly increased its purchases of computing capacity during the quarter.
Second is the inference compute required for PAPI and coPAPI as Tencent upgrades its internally developed Hunyuan models.
Third is the continued development of AI capabilities across Tencent’s products and services.
Beyond computing chips, Tencent’s infrastructure purchases also require large quantities of high-specification memory and storage hardware. At the physical layer, the race for computing power is not simply about adding more GPUs. Memory and storage are critical parts of the data pipeline and help determine how efficiently large AI models can run.
For Tencent, Hy3 currently has a relatively modest parameter count. Tencent president Martin Lau said during the earnings call that Hy4 would have more parameters, while also setting the goal of outperforming larger models.
Combining a larger parameter count with stronger performance, while still delivering fast responses and a smooth user experience, requires more powerful computing infrastructure. That, in turn, raises requirements for memory and storage.
Memory is also a highly cyclical industry, with capacity expansion typically taking one to three years. Supply therefore responds to changes in demand with a significant lag.
The current increase in AI infrastructure spending has pushed utilization at major memory manufacturers such as Samsung Electronics, SK Hynix, and Micron higher. With capacity constrained in the short and medium term, supply remains tight.
Tight supply has kept memory prices elevated. Tencent’s decision to sharply increase spending on computing capacity, even as its reported free cash flow turned negative, suggests that securing infrastructure has become a priority.
The company is committing capital to ensure that its large AI models and cloud business are not constrained by hardware availability, while building enough capacity to support further growth in its AI-related businesses.
Tencent adds to the demand case for memory
Tencent’s capital expenditure this quarter remains below the spending levels of some large US technology companies, but it has risen sharply both year on year and quarter over quarter.
More importantly, management’s comments on the earnings call about large AI models, CodeBuddy, WorkBuddy, Xiaowei, and other businesses repeatedly emphasized returns. That is consistent with Tencent’s historically pragmatic approach to investment.
Given the needs of its AI business and current supply-demand conditions in memory, Tencent is unlikely to reduce infrastructure spending solely because of near-term pressure on profit.
In the near term, Tencent’s core businesses remain highly cash-generative, which helps offset the impact of a temporary increase in spending. Over the longer term, however, those hardware investments will still need to demonstrate their value through concrete use cases across Tencent’s ecosystem.
For memory companies, meanwhile, shares of major manufacturers including Samsung, SK Hynix, and Micron have undergone corrections of varying degrees since July.
From a fundamental perspective, however, the argument that rising AI capital expenditure could support earnings growth at memory companies remains intact. Tencent’s higher spending adds another source of demand.
Tencent also places considerable emphasis on investment returns. Its willingness to increase spending despite tight supply and elevated memory prices suggests that memory is becoming an increasingly important requirement for AI infrastructure.
At the same time, memory capacity takes time to expand. Continued demand growth could therefore keep the supply-demand balance tight and support industry pricing.
The recent correction in shares of leading memory companies may instead reflect factors such as profit-taking, sector rotation, and expectations surrounding Federal Reserve policy, rather than a clear deterioration in industry fundamentals.
In that sense, Tencent’s increased AI spending could help extend the current memory upcycle.
KrASIA features translated and adapted content that was originally published by 36Kr. This article was written by Huang Yida for 36Kr.
Note: RMB figures are converted to USD at rates of RMB 6.76 = USD 1 based on estimates as of August 14, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.