Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
IndustryInsight · On RadarSCMP Tech→ original

Tencent Spends 6x More on AI Servers, Adjusted Profit Falls 6%

When a team spends six times its usual amount on servers in a quarter, profit takes a hit. That is what happened at Tencent. Its capital expenditure on AI capacity grew more than sixfold to over $12 billion. Adjusted profit fell 6% to 66.5 billion yuan, even though revenue rose 8%. The takeaway for anyone building on AI: compute is expensive, and even the giants pay for it with their margins.

AI-processed from SCMP Tech; edited by Hamidun News
Tencent Spends 6x More on AI Servers, Adjusted Profit Falls 6%
Source: SCMP Tech. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceSCMP Tech ↗
Editor-in-Chief
Quarterly capital expenditureover $12 billion, up more than sixfold
Key metric #2Tencent's revenue rose 8% to 207.3 billion yuan
Key metric #3Profit of 66.5 billion yuan (about $10 billion) matched the forecast of Bloomberg analysts

Imagine a startup that suddenly starts renting six times more GPUs than before. It has more customers, but the cloud bill eats almost everything, and profit drops. That is roughly what happened at Tencent, only at the scale of China's largest technology company. In the third quarter, it increased capital expenditure more than sixfold, to over $12 billion, and paid for it with lower profit and lower free cash flow.

What happened in the earnings report

Capital expenditure (money spent on servers, accelerators and data centers) rose more than sixfold in the three months to 30 September and exceeded $12 billion. According to SCMP, Tencent is investing in computing capacity to catch up with competitors in China's fierce AI race.

Adjusted net profit for the same period fell 6% year over year to 66.5 billion yuan (about $10 billion). That matched the consensus forecast of analysts surveyed by Bloomberg. Revenue, meanwhile, rose 8% to 207.3 billion yuan.

It is a familiar picture: the business is growing, but profit is not keeping pace, because a significant share of the money goes into infrastructure. Free cash flow, meaning the money left after investment, also declined.

What this means in plain terms

Say you are building a chatbot on top of your own model. While you have few users, a handful of GPUs is enough. As load grows, you have to buy hardware in advance, before it starts generating revenue. For a month or two your margin is negative, but responses stay fast and you can make the model larger.

Tencent is in the same situation, only the bill runs to billions of dollars. The quality of AI products depends heavily on available compute. The more of it you have, the larger the model you can train, and the more people you can serve without queues.

The fact that profit matched analysts' expectations tells you the main thing: the market knew about this spending in advance and did not treat it as a surprise. Large investors are so far willing to tolerate a margin squeeze for the sake of the AI race.

Why this matters in practice

First: if you are building an AI product, factor the cost of compute into your economics from the start. Even a company with billions in profit visibly pays for it out of its own margin, and a small team has less of a cushion.

Second: if you work with Chinese clouds or models, expect capacity there to keep expanding. That follows from the scale of the investment, although the source gives no timelines or specific services.

Third: if you are assessing the market, look not only at revenue but also at the gap between AI spending and the income it generates. That gap shows when infrastructure investment will start to pay off.

What the data does not show

The available text has no breakdown by segment or product, so it is impossible to assess the return on AI separately from Tencent's other businesses. It is also unclear whether this pace of spending will continue in the coming quarters or whether this quarter was the peak. The source also gives no details on free cash flow: all we know is that it came under pressure.

Story

Key Questions Answered

What happened in the earnings report?

Capital expenditure (money spent on servers, accelerators and data centers) rose more than sixfold in the three months to 30 September and exceeded $12 billion. According to SCMP, Tencent is investing in computing capacity to catch up with competitors in China's fierce AI race. Adjusted net profit for the same period fell 6% year over year to 66.5 billion yuan (about $10 billion). That matched the consensus forecast of analysts surveyed by Bloomberg. Revenue, meanwhile, rose 8% to 207.3 billion yuan.

What this means in plain terms?

Say you are building a chatbot on top of your own model. While you have few users, a handful of GPUs is enough. As load grows, you have to buy hardware in advance, before it starts generating revenue. For a month or two your margin is negative, but responses stay fast and you can make the model larger. Tencent is in the same situation, only the bill runs to billions of dollars. The quality of AI products depends heavily on available compute.

Why this matters in practice?

First: if you are building an AI product, factor the cost of compute into your economics from the start. Even a company with billions in profit visibly pays for it out of its own margin, and a small team has less of a cushion. Second: if you work with Chinese clouds or models, expect capacity there to keep expanding. That follows from the scale of the investment, although the source gives no timelines or specific services.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Private AI Builders Community · Vibe Club

Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews

Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.

Join Vibe Club →Explore Hamidun Academy
What do you think?
Loading comments…