Tencent · 0700.HK · Executive Summary
A value investing approach: buy a wonderful business at a fair price and hold it for as long as the business remains wonderful.
1. Investment summary
Warren Buffett teaches us to buy not stocks but stakes in businesses, and to ask three questions: is the business understandable, is it protected from competitors by a durable “moat”, and is it selling below its intrinsic value? Tencent answers yes to all three. This is the company that owns WeChat, an app no one among China’s 1.42 billion users can live without; the world’s largest gaming business; China’s second-largest payment system; and an investment portfolio the size of a major sovereign fund. Yet the market values the company’s core business at roughly seven to eight times annual operating profit — a multiple usually reserved for declining companies, not growing monopolies.
Author’s verdict
BUY below HKD 500 per share. Base-case fair value is HKD 580–630 per share. Expected price upside over a 12–24 month horizon: 20–30 percent, plus roughly 2–3 percent a year returned through dividends and buybacks. The aggressive-buy zone is below HKD 450, where the margin of safety exceeds 30 percent.
The author’s opinion as of the publication date; not an individual investment recommendation. See the disclaimer at the end.
Key figures for the first half of 2026
Metric | Value | YoY change |
|---|---|---|
Revenue | RMB 401.2 bn | +10.1% |
Gross profit | RMB 229.7 bn (57.2% margin) | +11.8% |
Net profit attributable to equity holders | RMB 114.1 bn | +10.3% |
Non-IFRS profit (excluding non-cash items) | RMB 136.3 bn | +9% |
Earnings per share (basic) | RMB 12.64 | +11.2% |
Operating cash flow | RMB 154.1 bn | +1.8% |
Total assets | RMB 2,159 bn | +5.9% year-to-date |
Net cash position | RMB 58.2 bn | positive |
Investment portfolio (three buckets) | approx. RMB 897 bn | 15 + 15 + 11% of assets |
Why now. The company is in the middle of the largest investment cycle in its history, buying RMB 59 billion of AI computing capacity per quarter. The market sees free cash flow falling and is nervous. Investors should focus on something else: these are temporary costs of building a fourth “moat”, while the three existing ones (social network, games, payments) continue to deliver record profits. Moments like this — when a great business looks worse than it really is — are what create entry points.
2. What kind of business is this
Imagine a company that simultaneously owns WhatsApp, Instagram, PayPal, Steam, Netflix and a venture fund the size of Berkshire Hathaway — all inside a single country of one and a half billion people. That is Tencent. Revenue comes from four segments.
Segment | Share of revenue | What it is, in plain terms |
|---|---|---|
Value-added services: games and social (VAS) | 49% | The world’s largest gaming business (in-game purchases) plus paid subscriptions to music, video and reading |
FinTech and business services | 30% | WeChat Pay payment fees, wealth management, lending, cloud servers for rent |
Marketing services (advertising) | 19% | Advertising inside WeChat and other services — the fastest-growing segment, thanks to artificial intelligence |
Other | 2% | Film, licensing, merchandise |
Moat one — WeChat, China’s only super-app. 1.42 billion monthly active users. Every Chinese person who needs to communicate has to be on WeChat — that is where their family, work, bill payments, doctor’s appointments and transit pass live. The network effect is absolute: leaving WeChat in China means dropping out of society. Millions of mini-programs run inside the app — shops, services and games from which Tencent collects fees. It is a monopoly that neither Alibaba nor ByteDance has been able to shake.
Moat two — “evergreen” games. Tencent has set strict criteria: a game counts as evergreen if more than 5 million people play it daily (mobile) or 2 million (PC), and its annual gross receipts exceed RMB 4 billion. The company has more than a dozen such games — they live for decades, like TV series that never end, and generate predictable revenue. On top of that come stakes in the world’s best studios: Epic Games (Unreal Engine and Fortnite), Larian (Baldur’s Gate 3), FromSoftware (Elden Ring), Ubisoft, and Korea’s KRAFTON and Shift Up.
Moat three — WeChat Pay. China’s number-one payment platform: more than a billion commercial transactions a day and over 70 million active merchants. It operates in 74 countries outside mainland China (more than 100 including cross-border commerce mini-programs), serving Chinese tourists (overseas spending growing +60% a year) and foreign visitors to China (+50%). Every transaction earns a fee. It is the country’s financial infrastructure, built into a messenger. Tencent has radically simplified payments for foreign tourists in China: Visa, Mastercard, JCB and Discover cards can now be linked directly to a Weixin Pay wallet without opening a Chinese bank account. Tencent is also rapidly scaling Weixin Palm Pay (payment by palm scan), already deployed on metro lines (including Beijing and Guangzhou), on university campuses and in major retail chains.
Moat four, under construction — artificial intelligence. The next section is devoted to it, because it is what will determine the company’s value over the coming decade.
3. Artificial intelligence — the fourth pillar
The company’s founder, Pony Ma, has openly admitted that Tencent missed the first wave of AI hype and lost the battle of benchmark rankings. But the company chose a different war — not for records, but for application: connecting AI models to the real-world needs of a billion users. The strategy rests on three products.
Tencent’s foundation language model. The third generation consistently ranks among the three most-used models in the world without being the largest — the secret is the best ratio of price to answer quality. The fourth generation is being trained now and will be larger. Thanks to the ecosystem’s unique proprietary data, the model leads in multimodal tasks — generating 3D objects and images — which is monetized directly in games and advertising. Capital expenditure priority number one is training new generations: without a top-tier model, nothing else works.
A platform for office workers and small businesses: the user states a task once, and an orchestration layer chooses the right models and tools on its own — Tencent’s or third-party. The skills library has passed 70,000 and keeps growing thanks to external developers, on an app-store model in which Tencent is the operator and fee collector. Important for investors: paying WorkBuddy users already deliver the same gross margin as the cloud business as a whole — the unit economics work; this is not burning cash to follow a trend.
A personal assistant for 1.4 billion users: order delivery, buy tickets, make a payment — with a single sentence, through thousands of existing mini-programs. It runs on WeLM, a specialized model optimized for privacy and low-cost compute. The next step is agent-to-agent interaction: merchants will get their own digital agents that negotiate with buyers’ agents without human involvement. Every such deal goes through WeChat Pay — transaction growth without growth in headcount costs.
A token is a unit of work for a language model, roughly one word of text. Over three years, global token prices have fallen by more than 99 percent — which sounds like a disaster for sellers. But Tencent’s cost of producing tokens is falling just as fast, so margins hold. And the market has sobered up: server memory has become more expensive (DDR4 up 1,800 percent), endless price dumping is no longer possible, and weak players are exiting. The only factor the market cannot eliminate is state-owned telecom companies with subsidized prices, but their quality of service is not comparable. Management also has a safety net: if new products fail to take off, the purchased computing capacity can be leased out immediately and at a profit.
Through stakes in robot makers (Unitree, Agibot, UBTECH, Pudu) and partnerships, Tencent connects its online ecosystem to the physical world: robots receive voice commands via Tencent’s speech-recognition technology and report to operators through WeChat. The company has taken the lead in investing in “embodied AI”, a sector China intends to dominate globally.
4. Revenue — a decade of growth with one pause
Year | Revenue, RMB bn | Growth |
|---|---|---|
2016 | 151.9 | — |
2017 | 237.8 | +57% |
2018 | 312.7 | +31% |
2019 | 377.3 | +21% |
2020 | 482.1 | +28% |
2021 | 560.1 | +16% |
2022 | 554.6 | −1% (regulatory shock) |
2023 | 609.0 | +10% |
2024 | 660.3 | +8% |
2025 | 751.8 | +14% |
First half of 2026 | 401.2 | +10.1% (Q2 +11%) |
The only decline of the decade came in 2022, when the state was restricting the gaming industry and the country was in lockdowns. The cause was external, not in the business — and as soon as the pressure eased, growth returned and has now been accelerating for a fourth year in a row. Advertising leads the growth: around +20% annually for four straight years, and the growth is “engineered”: an AI model improves the targeting of every ad impression, while AIM+ automation and the closed purchase loop inside WeChat (see an ad, buy without leaving the app) monetize traffic that used to go to waste. Cloud has accelerated to over +20% on demand for AI compute — and demand exceeds capacity. Management is candid that about two percentage points of advertising growth is a one-off effect, and keeps expectations in check.
Forecast: 10–12 percent growth in 2026 (about RMB 830–845 billion). Confidence is high: revenue is recurring (subscriptions, fees, in-game payments), two of the three drivers do not depend on the state of the economy, and the regulatory climate has turned favorable — Tencent is now a national AI champion.
5. Costs, gross profit and margins — a quiet revolution
Cost of revenues is growing more slowly than revenue (+7.9% vs. +10.1% for the half-year), and gross margin reached an all-time high of 58 percent in the second quarter. For a sense of scale: in 2022 it was 43 percent. Fifteen percentage points in four years is an exceptionally rare achievement for a company of this size.
Segment | Gross margin Q2 2026 | A year ago | Reason for the change |
|---|---|---|---|
Games and social services | 64% | 60% | Shift toward in-house games — no licensing royalties to pay |
Advertising | 57% | 58% | Depreciation of AI equipment eats into part of the gain |
FinTech & business services | 52% | 52% | Stable; cloud is growing on existing infrastructure |
Other | −35% | +14% | This is where losses from new AI products are booked |
The optimization is real, but its essence is reallocation, not cutting: staff costs did not grow at all (−0.7% against revenue +10%), license amortization is falling, while technical infrastructure costs nearly doubled (+81%) and server depreciation rose 46%. The company is funding its transition to AI out of internal efficiency. Management discloses the price of the bet openly: without new AI products, operating profit would have grown 19%; with them, 9%. The difference — about RMB 10.5 billion per quarter — is a deliberate investment in the future, not a loss of cost control.
Margin outlook: a plateau of 57–59% over the next eighteen months — the tailwind from the games mix will be offset by a wave of depreciation on purchased servers. A break above 60% is possible by 2028 if the new AI products turn profitable.
6. Earnings per share — two layers of one story
Year | EPS, RMB | Growth |
|---|---|---|
2016 | 4.38 | — |
2018 | 8.34 | +90% vs. 2016 |
2020 | 16.84 | +102% vs. 2018 |
2021 | 23.60 | +40% |
2022 | 19.76 | −16% (non-cash one-off items) |
2023 | 12.19 | −38% (one-off gains rolled off) |
2024 | 20.94 | +72% |
2025 | 24.75 | +18% |
First half of 2026 | 12.64 | +11.2% |
Average EPS growth over nine years has been 21 percent a year — a 5.6x increase. The 2022–2023 dip is an accounting illusion: in 2021–2022 profit was inflated by one-off gains from distributing JD.com and Meituan shares to shareholders, and against that base 2023 looked like a decline. Operating profit grew steadily throughout. The same effect appeared in Q2 2026: reported EPS rose only 1.5% because of a RMB 10 billion paper loss (the revaluation of preferred shares in a portfolio company whose value had risen — yes, risen!). Non-IFRS profit grew 9–10%.
The hidden engine: the company buys back and cancels its own shares — the share count is down 6 percent from its peak. The buyback is also pre-emptive: shares for employee compensation are bought on the market in advance, so there is no dilution, while share-based pay preserves the company’s cash. Every year a shareholder owns a slightly larger share of the business without buying a single additional share.
7. Cash flows — the kind of machine Buffett loves
Buffett calls insurance float — customers’ money that can be used for free — the secret of Berkshire. Tencent has its own float: deferred revenue of RMB 136 billion. Players and subscribers pay in advance — the company receives cash today and delivers the service later. It is an interest-free loan from a billion users, which grew by RMB 23 billion over the half-year and serves as a leading indicator of future revenue.
Metric | 2024 | 2025 | H1 2026 |
|---|---|---|---|
Operating cash flow, RMB bn | 258.5 | 303.1 | 154.1 |
As % of revenue | 39% | 40% | 38% |
As % of net profit | 130% | 132% | 131% |
Every yuan of profit is backed by 1.3 yuan of cash — earnings quality is impeccable. For an ordinary business, converting 15–20% of revenue into cash flow is an excellent result; Tencent consistently converts 40%. In Q2 2026 free cash flow turned negative for the first time (−RMB 13.8 billion): the company prepaid about RMB 50 billion for computing capacity. Excluding these prepayments, free cash flow would have been +RMB 37.6 billion. This is not a sickness but construction: by 2027–2028, as purchases normalize, free cash flow will rebound sharply.
8. Balance sheet — a fortress with rising walls
Total assets exceeded RMB 2.159 trillion. Equity funds 57% of assets — for comparison, a level above 50% is considered a sign of strong financial stability, and this is in the midst of an investment cycle. Property, plant and equipment reached a record RMB 199 billion (9.2% of assets) and will keep growing aggressively: a huge pipeline of equipment prepayments (RMB 91 billion in advances) has been built up and will convert into servers on the balance sheet.
Total liabilities rose 17.4% over the half-year to RMB 937 billion — let us look at why.
What increased | Amount | Is it bad? |
|---|---|---|
Payables to server suppliers | +RMB 54.5 bn | No — ordinary invoices for AI equipment, paid on schedule |
Borrowings (mostly short-term, in RMB) | +RMB 48.0 bn | No — rates of 0.75–3.5%, below the yield on the company’s own reserves |
User prepayments (deferred revenue) | +RMB 23.3 bn | Quite the opposite, excellent — this is future revenue |
Notes payable (June 2026 issue) | +RMB 17.3 bn | No — 10- to 30-year RMB notes at 2.5–3.1% |
Key fact: even after all the purchases, the net cash position is positive — cash and deposits exceed all debt combined by RMB 58 billion. The company is funding the largest capex program in its history with cheap debt, without selling either its investments or its future. Interest expense (about RMB 13.5 billion a year) is covered many times over by interest income on reserves alone (RMB 16.9 billion).
What does NOT appear on the balance sheet deserves a separate word: ten years of investment in in-house R&D worth RMB 500 billion (including RMB 50 billion in the half-year) has been expensed immediately — under accounting rules it is not capitalized. The Hunyuan model, engines, recommendation algorithms — all of it sits on the balance sheet at zero. The company’s real value exceeds its book value by hundreds of billions.
9. The investment portfolio — a “second Tencent” inside Tencent
When you buy a Tencent share, you buy two businesses at once: the operating machine (WeChat, games, advertising, cloud) and a shadow investment holding company worth ~RMB 900 billion (roughly USD 125 billion) — the size of the world’s largest funds. The portfolio sits in three accounting buckets.
Bucket | Size | What is inside | Impact on financial statements |
|---|---|---|---|
Associates (significant influence, board seats) | RMB 324 bn | Pinduoduo, Epic Games, WeBank, Universal Music, Kuaishou, Ubisoft, AI-chip maker Enflame | Share of profit — in the income statement (~RMB 6.4 bn of profit per quarter excluding non-cash items) |
Equity investments without influence (FVOCI) | RMB 325 bn | Sea, Spotify, Snap, Reddit, KRAFTON, Bilibili, AI labs Moonshot and Baichuan, GPU maker Moore Threads | Revaluation bypasses profit and goes straight to equity |
Venture bucket (FVPL) | RMB 248 bn | DeepSeek, robot makers Agibot and Pudu, Kling AI (video generation), Argentine bank Ualá + treasury investments of ~RMB 50 bn | Revaluation hits profit every quarter (+RMB 10.7 bn in the half-year) |
Fact one — hidden value. The market value of listed associates exceeds their carrying amount by RMB 51 billion. In total: the fair value of listed stakes is RMB 487 billion, and the carrying amount of unlisted ones is RMB 388 billion. Chief Strategy Officer James Mitchell puts it plainly: “Our portfolio is a liquid strategic reserve. If the market temporarily misprices an asset, we know how to wait.”
Fact two — the pivot to artificial intelligence. Over the half-year, the carrying amount of unlisted startups rose by RMB 24.8 billion, with the main jump in the second quarter. Tencent has assembled positions across Chinese AI: chips (Enflame, Moore Threads — a hedge against US sanctions on GPUs), models (DeepSeek, Moonshot, Baichuan), applications (Kling AI), and robots (Unitree, Agibot, Pudu, UBTECH — the whole landscape). Upcoming IPOs of these companies are a ready-made catalyst for a re-rating.
Fact three — the portfolio works. It generates ~RMB 25 billion of profit a year (a tenth of group profit), pays dividends (RMB 4.4 billion in the half-year), and is being monetized: RMB 42 billion of sales from the equity bucket in the half-year, and part of a listed associate was sold in July. The listed portfolio’s drawdown in the first half (−RMB 48.9 billion through equity) is market fluctuation that does not affect profit: in 2024 the same line contributed +RMB 94.6 billion.
10. Dividends and buybacks — a growing return of capital
Payment year | Dividend per share, HKD | Growth |
|---|---|---|
for 2019 | 1.20 | +20% |
for 2021 | 1.60 + distribution of JD.com shares | — |
for 2022 | 2.40 + distribution of Meituan shares | +50% |
for 2023 | 3.40 | +42% |
for 2024 | 4.50 | +32% |
for 2025 | 5.30 (paid in May 2026) | +18% |
The dividend has grown 4.4x in six years and is raised every year. The current yield is modest — about 1% — but it is only the visible part of the return. The company pays out only 17–18% of profit as dividends, but buys back and cancels shares (HKD 24.4 billion in the half-year, HKD 112 billion in 2024). The total shareholder return is 2–3% a year plus a growing share of the business. A special trump card is dividends “in kind”: in 2021–2022 shareholders received JD.com and Meituan stakes worth ~HKD 290 billion — the largest distributions in the history of Asian markets. A portfolio of RMB 487 billion in listed shares makes it possible to repeat the trick. The dividend’s safety margin is huge: it is covered many times over by operating cash flow, even in a quarter of record purchases.
11. Ratios
Ratio | Value | What it means in plain terms |
|---|---|---|
PEG (P/E to growth) | ~1.0 | The company’s price exactly matches its earnings growth rate. For a global technology monopoly that is clearly cheap: such companies usually trade at one and a half to two times that |
Free cash flow yield | approx. 4.5% (2025) | How much cash the business generates per dollar invested. US tech giants yield 1.5–2.5%. A 4.5% level is an anomaly of cheapness |
Enterprise value to EBITDA (EV/EBITDA) | 10.7 (reported) and 10.0 (adjusted) | Buying the business pays back in ten years of operating cash flows — fair for a mature company, cheap for a growing one |
Same, excluding the investment portfolio (Core EV/EBITDA) | 7.9 and 7.3 | Strip out the portfolio and cash, and the pure operating business is valued at seven years of profit — the price of a declining company for a growing monopoly |
Return on equity (ROE) | 21% | Every RMB 100 of shareholders’ equity earns RMB 21 of profit a year. Above 15% is excellent; 21% on trillion-yuan equity is the mark of a monopoly |
Return on capital employed (ROCE) | 20% | Also accounts for borrowed money — a high figure proves that profit is not bought with leverage |
Same, operating business only (Core ROCE) | approx. 107% | A phenomenon: the business itself (servers, software, teams) requires only RMB 225 bn of capital and generates RMB 240+ bn of operating profit — it pays for itself within a year |
Return on assets (ROA) | 13.6% | Good for a company where half the assets are a conservative investment fund |
Cash flow margin | 40% | Out of every yuan of revenue, 40 fen stays as cash. The norm for a healthy business is 15–20% |
Equity ratio | 57% | More than half of the empire belongs to shareholders rather than creditors — a financial fortress |
Takeaway from the dashboard: on every efficiency metric without exception, Tencent is a company of the highest tier in the global economy, while on every price metric it is valued like an average company. That combination is what undervaluation means.
12. Who owns the company and who runs it
The structure is simple and fair: a single share class, one share — one vote, no special privileges for the founders. The largest shareholder is the Dutch Prosus (22–23%), a strategic investor since 2001 that does not interfere in management. Founder Pony Ma holds ~8% — his wealth is in the company’s shares, so his interests are aligned with investors’. President Martin Lau holds ~0.5%, earned over twenty years with the company. Management behaves like co-owners: in 2024–2025, seeing no projects with exceptional returns, it returned cash to shareholders like a classic value company; in 2026, seeing returns in AI, it redeployed capital there — with an explicit caveat that spending is tied to measurable returns, not fashion. CFO John Lo confirmed that the infrastructure built so far is already paying for itself — capacity is fully utilized by in-house products and external leasing.
The competitive field in China is a three-way race. Alibaba is building a compute backbone and hands startups capacity in exchange for equity; ByteDance bets on in-house development and enormous traffic; Tencent acts as a smart portfolio investor — minority stakes in the best labs (the “six tigers” of AI), plus its own model, plus a unique trump card: distribution through WeChat, which nobody else has.
13. Risks — what a prudent investor must consider
Risk | Substance | Mitigation |
|---|---|---|
Regulatory (China) | The state already halted growth in 2021–2022 (game licenses, antitrust measures) | Today the climate is the opposite: Tencent is a national AI champion; but the cycle could repeat |
Geopolitics and chips | US sanctions restrict access to advanced GPUs | Stakes in Chinese chipmakers (Enflame, Moore Threads); in-house optimization of models for available hardware |
Lagging AI monetization | Capacity has been bought, but revenue from new products may lag | Management’s safety net: idle capacity is leased out at an immediate profit |
Weak consumption in China | Weighs on payments and FinTech (growth of only +9%) | The other two drivers (advertising, cloud) do not depend on the consumer |
Portfolio volatility | Market drawdowns hit equity (−RMB 49 bn in the half-year) and generate alarming headlines | No impact on operating profit; over the long term the portfolio appreciates |
Portfolio concentration | Pinduoduo — about 40% of the value of listed associates | Diversification: hundreds of positions across five industry clusters |
Currency | Revenue in RMB, shares in HKD, part of the portfolio in USD | A cheap Hong Kong dollar makes the purchase more attractive for a holder of RMB-denominated revenue |
The main risk we CANNOT rule out is political: China remains a country where a single decree can change the rules of the game. That is exactly why we demand a large margin of safety in the price — and exactly why it exists now.
14. Valuation: what the company is worth and at what price to buy
Building up value piece by piece (sum-of-the-parts method)
Component | Valuation | How it was derived |
|---|---|---|
Operating business | RMB 3,795 – 4,140 bn | Adjusted annual operating profit of ~RMB 345 bn × a fair multiple of 11–12 (today the market assigns only 7.3) |
Investment portfolio | approx. RMB 700 bn | RMB 875 bn of fair/carrying value with a 20% holding-company discount |
Net cash | RMB 58 bn | Cash and deposits minus all debt |
Total equity value | RMB 4,550 – 4,900 bn | Sum of the three lines |
Per share (9,029 mn shares) | RMB 504 – 542 | Divided by the number of shares |
In Hong Kong dollars (rate ~0.87) | HKD 580 – 630 | Fair value per share |
Three scenarios over a 12–24 month horizon
Scenario | Target price | Upside from HKD 480 | Conditions |
|---|---|---|---|
Conservative | HKD 440–470 | approx. 0% | No multiple expansion, AI products delayed, portfolio at a 30% discount. Even here, dividends and buybacks deliver +2–3% a year |
Base | HKD 580–630 | +20–30% | Earnings growth of 9–11% a year, partial market recognition of the portfolio’s value, cloud and advertising growing as planned |
Optimistic | HKD 700–720 | +45–50% | AI products turn profitable, IPOs of portfolio AI companies (Moore Threads, Moonshot), a multiple of 14 |
A Buffett-style recommendation
“It’s far better to buy a wonderful company at a fair price than a fair company at a wonderful price.” Here we have a wonderful company at a price below fair value. The buy zone is up to HKD 500 per share; the aggressive accumulation zone is below HKD 450, where the margin of safety to the base-case valuation exceeds 30%. Expected total return in the base case: 20–30% price appreciation over 12–24 months plus 2–3% a year from dividends and buybacks. Holding horizon: three years or more — the main catalysts (profitability of AI products, IPOs of portfolio companies, recovery of free cash flow) play out in 2027–2028. Sell above HKD 700 or if a harsh regulatory cycle returns.
What to expect along the way: volatility is inevitable — quarterly reports will be full of paper losses on the portfolio, free cash flow will stay under pressure until peak capex ends, and Chinese equities swing widely. A prudent investor uses these swings to add to the position, not to panic: the business underneath is growing 10% a year with record margins.
Disclaimer. This document has been prepared for analytical and educational purposes on the basis of Tencent Holdings Limited’s public financial reporting for the first half of 2026, its annual reports for 2016–2025 and statements by management. It is not an individual investment recommendation, and its author is not a licensed investment adviser. Investing in equities involves the risk of loss of capital; past performance does not guarantee future results. All figures are rounded; discrepancies within rounding may occur.
Want to see what these conclusions are built on?
The full financial analysis of Tencent: equity, assets, liabilities, revenue and profit, cash flows and ratios — line by line through the financial statements.
Go to the full report →