Tencent

IThe business

Tencent runs Weixin, the messaging app 1,439 million people open every month, and it sells to those people in four ways: games and subscriptions, advertising, payments and loans, and cloud services.⁷ In the twelve months to June 2026 revenue was RMB 788.5B and IFRS operating profit was RMB 258.5B, a margin of 32.8%.⁶⁹ Since 2016 revenue has grown 19.4% a year and operating profit before other gains 19.3% a year.C (calculated)¹⁴¹⁷ Mainland China was 88% of 2025 revenue.C (calculated)⁹

Tencent allocates no costs below gross profit to its four segments, so the chart shows gross profit. Value-added services, which holds games and social networks, is the largest and the most profitable by gross margin. Marketing Services is a fifth of revenue and of gross profit. FinTech and Business Services is second and carries the lowest margin of the three big ones.C (calculated)⁶⁹

How it makes money

• Games. RMB 241.5B in 2025: Domestic RMB 164.2B, up 18%, and International RMB 77.4B, up 33%.R (reported) Domestic growth came from Delta Force, VALORANT and evergreen titles such as Honour of Kings. In the June quarter Domestic grew 17% to RMB 47.3B and International fell 0.8% to RMB 18.6B, or rose 4% in constant currency.⁹⁷
• Social networks. RMB 127.7B in 2025, up 5%, from Video Accounts live streaming, music subscriptions and in-game items.R (reported)⁹
• Marketing Services. RMB 145.0B in 2025, up 19%, and RMB 159.1B in the twelve months to June 2026.C (calculated) The ads run inside Video Accounts, Weixin Search, Mini Programs and Tencent’s mobile ad network. Tencent credits AI ad targeting and closed-loop ads, where the user clicks through to a Mini Shop or Mini Game inside Weixin.⁹⁷
• FinTech and Business Services. RMB 229.4B in 2025, up 8%: commercial payments, wealth management, consumer loans, cloud and the technology fees on Mini Shop sales.R (reported) Cloud revenue growth “accelerated to low-twenties %” in the June quarter. Tencent does not report cloud revenue.⁹⁸

What changed in 2026

Tencent spent little on capital for a company its size until 2024: on the company’s definition, capital spending was 3% to 4% of revenue in 2022 and 2023, then 11.6% in 2024 and 10.5% in 2025.C (calculated)^1211109 The AI build is the larger step. Capital spending was RMB 84.7B in the first half of 2026, up 82% on RMB 46.6B, and RMB 52.8B in the June quarter alone, up 176% and 25.8% of the quarter’s revenue.C (calculated) That first half exceeds any full year in the annual reports from 2016 to 2025; the largest was RMB 79.2B in 2025.C (calculated)⁶⁷⁹ Alibaba’s June-quarter capital spending was RMB 67.7B, 25.2% of revenue; its release does not say whether prepayments are inside, so the comparison is loose.C (calculated)³⁶

Capital spending appears in three definitions on this page. My valuation uses the third.

Free cash flow was negative RMB 13.8B in the June quarter, which Caixin described as the first since 2019.⁷³³ Net cash fell from RMB 146.9B in March to RMB 58.2B, mostly the RMB 41.6B annual dividend and RMB 14.7B of buybacks; free cash flow accounts for RMB 13.8B.C (calculated)⁷⁸ Operating cash flow carried RMB 51.4B of prepayments for compute; without them free cash flow would have been RMB 37.6B.C (calculated)⁷ Tencent says its new AI products cost RMB 18B in 2025 and that it expects to “more than double” them in 2026. In the first half the products cost RMB 19.3B net of any revenue, the gap between operating profit excluding them and non-IFRS operating profit; that is a different measure from the 2025 cost figure.C (calculated)¹⁹⁷²⁰

The cash flow statement shows only part of the commitment. Note 35 of the interim report records contracted commitments of RMB 372.9B at 30 June 2026, up from RMB 75.9B at 31 December 2025.C (calculated) Capital commitments for assets other than buildings and investees went from RMB 3.4B to RMB 74.3B. Other commitments, mainly technical infrastructure operations costs, game licensing, media content and other technical services, went from RMB 55.2B to RMB 280.1B. Non-current prepayments for capital assets and services went from RMB 14.1B to RMB 81.1B. The report gives no maturities.⁶⁹

Tencent’s slide says the build has “clear downside protection”, because the spending is in infrastructure it can rent out through Tencent Cloud. That depends on outside demand for rented compute at a good price, and Tencent says it is “still compute constrained”.⁸

Contracted commitments rose to 4.9 times their December level in six months, to 47% of a year’s revenue.C (calculated)

IIThe moat

I think Tencent’s moat is a network effect. Weixin connects its users with each other, and payments, Mini Programs, Mini Shops, Video Accounts and advertising sit on top of it. A merchant goes where the users are, and a user stays where the merchants and friends are. Tencent’s own list adds IP, low take rates, regulatory requirements and private data.⁸ Each user is worth more to Tencent, though the number of users has nearly stopped growing; AI has not yet tested the moat.

Weixin and WeChat monthly users grew 11% a year in 2017 and in 2018 and 2% in the year to June 2026.¹⁷¹⁶⁷

Marketing Services revenue per user is the best Weixin-specific measure I can build. Online Advertising, the earlier segment name, brought in RMB 30 per year-end Weixin user in 2016; Marketing Services brought in RMB 102 in 2025 and RMB 111 in the twelve months to June 2026, 3.4 times in nine years or 14.5% a year, though the segment’s scope changed.C (calculated)^17151296 Group revenue per user rose 3.1 times, from RMB 171 to RMB 530, but it includes international games, cloud and fintech that Weixin does not earn, so I do not use it.C (calculated)¹⁷⁹

Marketing Services grew 22% in the June quarter on 2% more users. Tencent attributes it to better ad targeting and closed-loop ads, with an ad load it says is below the industry’s; the reports give no split between price and the number of ads shown.⁷⁸ Fee-based subscriptions averaged 259 million a day in the June quarter, down from 264 million a year earlier. Tencent discloses no Mini Program or Mini Shop sales, WeChat Pay volume, time spent or retention for Weixin, so lock-in has to be read from the segments.⁷⁸

In 2022, the bad year, revenue fell 1% and advertising fell 7% (Online Advertising, the earlier segment name). Core operating profit fell 11% to RMB 102.8B, and free cash flow was still RMB 89.3B.C (calculated)¹²¹¹ In 2021, after new limits on minors, their time on domestic games fell 88% in the fourth quarter, and revenue still grew 16% that year.C (calculated)¹³ The moat holds against competitors better than it holds against regulators, and the Section 1260H listing and the Structure Contracts are in the risks below.

Weixin’s users grew 2% in the last year; Marketing Services revenue grew 22% on them.

IIIThe competitive advantage

Tencent’s advantage over its rivals is that it already owns the user. A rival in advertising or games has to win the user first. The June quarter of 2026 puts numbers on the gap.

Baidu’s chief executive says its online marketing business remains under pressure.⁴⁰ Each peer’s number is from its own release. ByteDance, which owns Douyin and Doubao, is private and the largest rival for attention and ad budgets; I cannot size it from filings, so the ad comparison covers the three companies that give a figure.

Where Tencent leads

The lead is distribution. Advertisers reach users across Weixin and pay for conversions inside it, and Tencent credits the closed loop for the growth.⁷ In games, Tencent’s revenue grew 11% in the June quarter against 9.7% at NetEase, whose figure includes overseas; China’s games industry grew 7.7% in 2025 by an association’s sales measure, which is not Tencent’s revenue measure.C (calculated)⁷³⁷⁴¹ Payments is the second lead: Tencent’s slide puts it first in China by monthly and daily users, citing QuestMobile, and I have no independent share figure.⁸

Where it trails

Cloud and consumer AI trail. Alibaba’s AI cloud grew 45% against Tencent’s low twenties, and Tencent says it is compute constrained.³⁶⁸ Yuanbao is not among the three Chinese assistants named in the May 2026 QuestMobile ranking.⁴² International games are flat.⁷ Tencent is spending at Alibaba’s rate and is behind it in the markets that spending serves.

Tencent grew advertising 22% while Alibaba (1% like for like), PDD and Kuaishou grew theirs 4.4% or less.

IVManagement

Ma Huateng, 54, is one of Tencent’s core founders, has been employed by the group since 1999, and is chairman and chief executive. Martin Lau has been President since 2006 and John Lo chief financial officer since 2012.⁹ Under them revenue grew 19.4% a year from 2016 to 2025, and non-IFRS profit attributable to shareholders grew 21.4% a year.C (calculated)¹⁷⁹

The board has eight directors: one executive, two non-executives who represent Prosus and Naspers, and five independents. Three of the five independents have served more than nine years: Li Dong Sheng and Ian Stone since April 2004 and Yang Siu Shun since July 2016.C (calculated)⁹⁴⁶ The annual report names no successor to the chief executive. It says the board ensures succession is planned, and it records that combining chairman and chief executive is at variance with the Hong Kong governance code.⁹ Prosus’s holding fell from 33.17% in 2017 to 22.80% at the end of 2025. Prosus’s own page shows 2,044.0 million shares, 22.5% of the issued shares, on 16 October 2026, after it sold 0.9% of Tencent’s shares for US$6.5B between April 2025 and March 2026.C (calculated)^1795051 Ma’s holding rose from 8.63% to 8.82%. Prosus’s sales are an overhang the filings do not size.

Incentives

Ma’s 2025 pay was RMB 48.5M: salary of RMB 7.6M, a bonus of RMB 39.4M, fees and benefits of RMB 1.5M and no shares.⁹ His 804.9 million shares are worth about HK$309B at the price, so a 1% move in the share price changes his wealth by about HK$3.1B, 54 times his annual pay.C (calculated) The share price is his incentive.

The rest is looser. Tencent discloses no formula for the chief executive’s bonus, and its annual report shows no advisory vote on pay. The 12 senior managers below the chief executive were paid RMB 4.46B in 2025, RMB 3.92B of it in stock. All options and awards involving new shares granted to employees in 2025 carried no performance targets.⁹ Stock pay was RMB 31.9B in 2025, including a one-off of about RMB 4B, and RMB 27.5B in the twelve months to June 2026, or 3.5% of revenue.C (calculated)⁹⁶ Nothing in the disclosed pay measures return on capital.

What does that pay them to do? Raise the share price over years. No disclosed pay measure touches return on capital, and the awards below Ma carry no targets. So I read the record: the JD.com and Meituan distributions are the best part of it, the 2025 and 2026 buybacks are 21% to 26% under water, and the AI spending has no result yet.⁹⁶

Capital allocation

The record in one table, with the AI build last because it has no result yet.

Buybacks before 2025 are worth more than they cost; the 2025 and 2026 purchases are worth 21% to 26% less.C (calculated)^12111096 After 30 June the company bought about HK$0.1B a day between 23 September and 15 October at HK$414 to HK$442, and in March it said it would likely repurchase a lower value of shares in 2026 to fund AI and raise dividends.²⁵¹⁹ Issued shares fell 5.4% from December 2021 to October 2026, to 9,092.6 million.C (calculated)²⁴²⁵¹³

The JD.com and Meituan distributions are the best part of the record: Tencent held the stakes for years and handed the shares to its owners. From 2016 to 2025 Tencent also paid RMB 203.6B for stakes in associates and RMB 78.7B for business combinations.C (calculated)^{17161514131211109} Investment marks move reported profit and equity, not cash flow. Listed holdings fell from RMB 672.7B to RMB 487.2B in six months, and the first-half loss in other comprehensive income was RMB 81.5B, of which RMB 50.8B was fair-value changes and RMB 30.6B currency translation.C (calculated)¹⁹⁷⁶

In its FY2025 presentation Tencent described its new AI spending as:

similar to CAPEX or investments in affiliates in nature

I agree, and its record is short.

VFinancials

Profit is growing and cash is not. IFRS operating profit was RMB 258.5B in the twelve months to June 2026, up 7% on 2025, and free cash flow was RMB 136.1B, down 26% from RMB 183.4B.C (calculated)⁹⁶

I define owner free cash flow as operating cash flow less capital spending, less stock pay. Capital spending includes purchases of property and equipment and prepayments for it, media content and game licenses, and lease principal. This free cash flow matches the company’s own within 0.6% for 2024 and 2025 (RMB 155.3B and RMB 182.6B), and I take stock pay off it because Tencent pays its people in shares and then spends cash repurchasing them.^91061823

Owner free cash flow was 13.8% of revenue in the latest twelve months, the lowest since 2022, and that is after the RMB 51.4B of compute prepayments.C (calculated)⁷

Gross margin rose to 57.8% in the June quarter, and operating margin held at 33%. Non-IFRS operating margin, which excludes stock pay, slipped to 36.9% from 37.5% a year earlier. Excluding the new AI products it rose to 42% from 39%.⁷⁸ The core business is getting more profitable and the new products absorb the gain.

Accounting to watch

1. Prepayments. Non-current prepayments for capital assets and services rose RMB 67.1B in six months.C (calculated) The company does not say whether they become equipment or prepaid rent; rent reaches operating profit as the service is used.⁶
2. Payables. Accounts payable rose 45% in six months, from RMB 121.1B to RMB 175.6B, while operating cash flow was RMB 154.1B against RMB 151.3B a year earlier.⁷⁶ That support reverses when suppliers are paid.
3. Non-IFRS profit. It was RMB 271.6B in the twelve months to June, 15% above the IFRS figure of RMB 235.5B, mainly share-based pay, investee gains and losses, and amortization of acquired intangibles.C (calculated)⁷⁶
4. Asset lives. Equipment is depreciated over 2 to 10 years.⁹ Depreciation was RMB 9.6B in the June quarter against RMB 6.1B a year earlier.⁶ My value rests on cash, so lives barely move it: four years changes my base by +HK$2 and seven years by -HK$4.E (estimate) Reported operating profit moves far more.

The RMB 280.1B of “other commitments” is in neither the balance sheet nor the cash flow; I price it in the valuation section.

VIValuation

How this goes badly wrong

Before the numbers, the ways I lose money, most likely first.

1. The build does not pay. Capital spending stays near the 28% of revenue of the last twelve months and the new products add little. My stress run, with spending at 58% of revenue for two years, gives HK$214 a share.E (estimate).⁶⁹
2. Advertising stalls. It grew 22% on users growing 2%. In 2022 it fell 7%, and a repeat while the build runs leaves nothing to pay for it.¹²¹¹
3. The structure fails. The companies held under Structure Contracts with four individuals carried 57% of 2025 revenue on about RMB 27B of net assets, and the annual report calls the legal position uncertain.⁹ I cannot price that, so I added 5 points to my margin of safety.

What an owner can take out

Profit attributable to shareholders was RMB 235.5B in the twelve months to June 2026. An owner could take out RMB 108.6B, after RMB 27.5B of stock pay and all the capital spending, or 46% of reported profit. In 2025 it was 67%C (calculated).⁶⁹ At the price that is a yield of 3.7%, or 5.1% on 2025, against an earnings yield of 7.8%.

What I can’t judge

I can’t tell whether the June quarter’s RMB 51.4B of compute prepayments were a catch-up or the new pace, or what the AI products will earn. So I show a range. Spending at 28% of revenue for fifteen years gives HK$295 to HK$338 a share, and my 17% gives HK$508. The price sits between.E (estimate).⁶⁹

What the price implies

At HK$383.60 Tencent’s market value is HK$3,488B, or US$444.7B, on the 9,092.6 million issued shares, which include 84.3 million held by the award-scheme trust; my per-share values use the same count.C (calculated)⁵⁹²⁵⁵⁶ I value the Hong Kong share, counter 00700. The over-the-counter ADR, TCEHY, closed at US$49.41 on 23 October, or HK$387.6 at 7.8436; each ADR is one ordinary share.R (reported)⁵⁸ In US dollars my bear, base and margin-of-safety values are about US$34, US$65 and US$45, against US$48.90 for HK$383.60.C (calculated)⁴⁵

Take out net cash and investments at my haircuts (table below), HK$727B or HK$80 a share, and the operating business is priced at HK$2,761B. That is 11.3 times its after-tax core operating profit of RMB 207.3B, at the 19% cash tax rate I use, or 9.2 times before tax, the basis of the peer table below; it is 7.1 times EBITDA.C (calculated)⁶⁷ On the whole company the price is 12.8 times trailing diluted IFRS earnings per share and 11.1 times non-IFRS earnings. The IFRS earnings yield of 7.8% is above the 5.00% ten-year Treasury; owner free cash flow yields 3.7% on the market value, or 5.1% on 2025’s figure.C (calculated)⁴⁷ On the peer table’s basis, net cash only and before tax, the market pays 11.4 times Tencent’s operating profit against 20.7 times for Meta and 26.3 times for Alphabet.

Meta’s operating income was 37% lower in the September quarter than a year earlier, so its multiple on current profit is above the table’s 20.7 times.C (calculated)⁵² Alibaba is not in the table: I could not build a clean twelve-month operating figure for it.

Run my model backwards. Holding the 10.0% discount rate and 3.0% growth after year 15, the price needs owner free cash flow to grow 6.9% a year for fifteen years from the last twelve months’ RMB 108.6B, to about RMB 299B in year 15.C (calculated) Over the last nine years it grew 15.2% a year, and my base case gets 9.3%.¹⁷⁹

The start matters: the last twelve months carry RMB 51.4B of compute prepayments in one quarter, and the company disclosed none for 2025. From 2025’s RMB 151.5B the price needs 3.5% a year for fifteen years (3.7% for ten); from the last twelve months, 6.9% (8.2%).C (calculated)⁹⁶ I do not add the prepayments back: they are cash paid, and RMB 81.1B of non-current prepayments and RMB 280.1B of other commitments are still open. Payables also rose RMB 54.5B in the first half, which flatters operating cash flow.⁷⁶

Capital gives the cleaner test. Hold my base-case revenue and margins and solve for the all-in capital spending the price needs from year 6 on: 26.4% of revenue, against 17% in my base case, 15.9% in 2025 and 28.0% in the last twelve months. Hold the margin before depreciation at today’s 41.5% instead and the price needs 23.1%.C (calculated)⁶⁹

The price needs all-in capital spending to stay at 26% of revenue after year five; my base case has 17%.C (calculated)

If the 28% pace lasted fifteen years, the model gives HK$295 a share on today’s 41.5% margin before depreciation and HK$338 on my base margins.E (estimate)⁶⁹ The price of HK$383.60 sits between those and my base value, so it pays for part of the build to pass. The question is capital intensity after year five: the present value of the spending I assume above 17% of revenue in years 1 to 4 is HK$23 a share, 6% of the price.E (estimate)

The discount rate

I use 10.0%, a dollar hurdle rather than a derived number. The capital asset pricing model cannot supply one: Tencent’s 60-month beta against the S&P 500 is 0.17, since the stock moves with China.C (calculated)³⁵⁴⁹ Two builds sit around it. The 5.00% ten-year Treasury yield plus Damodaran’s 4.20% implied equity risk premium, at a beta of 1, gives 9.2%, or 9.5% on the 5.29% risk-free rate he used for that premium. Adding Tencent’s spread to today’s Treasury gives 10.1%: its ten-year dollar notes priced at 6.10%, 0.9 points over the 5.21% Treasury yield of 13 October, the date of the Reuters report. That build stacks an equity premium on a rate that already carries Tencent’s credit spread, so I treat it as the high end.C (calculated)^47482632

I discount renminbi cash flows at that dollar rate and convert at today’s HK$1.1759 per renminbi, which assumes the renminbi does not rise. Tencent’s ten-year renminbi notes cost 2.40% in October and its dollar notes 6.10%.²⁶ If the renminbi rose each year by the 3.7-point gap between those two notes, the equivalent rate would be 6.3% and my base HK$1,062; at the 2.40% note yield plus the 4.20% premium, 6.6%, it would be HK$974.E (estimate) I do not use them: a dollar investor’s required return is a dollar number, spot is my best forecast for the renminbi, and the grid shows 7.5% and 8.0% for a reader who wants a lower rate. A 5% weaker renminbi cuts my base by 5%.E (estimate)

One point on the discount rate moves my base by 12% to 16%. The price equals my base at a discount rate of 12.4%.E (estimate)⁴⁷⁴⁸

Bear, base and bull

The model runs fifteen years, from July 2026 to June 2041, in renminbi. Revenue is Marketing Services plus everything else. The margin before depreciation is after stock pay, and I estimate it. Depreciation comes from the capital spent: the equipment on the balance sheet at the June quarter’s rate, construction in progress and prepayments written off over five years from mid-year, and new spending over five years. Cash tax is 19% of operating profit, the ratio of income tax paid to core operating profit in the twelve months to June 2026; the 17.1% book rate in 2025 includes untaxed investment gains, and cash tax that year was 18.3%.C (calculated)⁶⁹ Each year’s owner free cash flow is margin before depreciation, less tax, less all-in cash capital spending. The same form gives RMB 109B for the twelve months to June 2026 against RMB 109B calculated from the cash flow statement, where that figure carries RMB 51.4B of compute prepayments and gains from higher payables that roughly offset, and RMB 145B for 2025 against RMB 152B.C (calculated)⁶⁹

I add net cash and investments at the haircuts in the table, count 9.0926 billion issued shares, and convert at one rate. Lease liabilities are not deducted, because lease principal is already inside all-in capital spending. Stock pay is charged as a cost while the shares behind awards already granted sit in the count, an overlap under 1%.⁶⁵²⁵

The listed haircut covers tax on the sale of stakes held at a low cost, the discount on large blocks, and a fall in Hong Kong tech since 30 June; in July 2026 Tencent also sold some shares in a listed associate, which the interim report does not size.⁶ The unlisted haircut is larger because carrying values are the company’s own. The non-controlling interests are at book value, and their market value is higher but not disclosed.⁶⁷³⁴ At the company’s marks with no haircut the base is HK$537; with the investments worth nothing it is HK$424.E (estimate)

My base case takes the business as it is and the build as it comes. Revenue grows 6.8% a year for ten years, to RMB 1,523B, and Marketing Services grows 20% in year 1 and fades to 5%, so it rises from 20% to 27% of revenue. Year-10 Marketing Services of RMB 413B is RMB 235 per monthly Weixin user if users grow 2% a year, and RMB 287 if they do not, against RMB 111 now: 7.8% a year per user, against 14.5% from 2016 to 2025. I found no verified market-size figure for China’s advertising, so per-user history is my only anchor.E (estimate)⁶⁹¹⁷ The margin before depreciation dips to 40.5% for two years as the new AI products cost more, then climbs to 45%, below the 47% or so that the existing business earned in the June quarter.C (calculated)⁷⁶ All-in capital spending is 28% of revenue in year 1, the last twelve months’ level with the compute prepayments, and falls to 17% by year 5. That is 31% below the first half’s 40.4% and about half the June quarter’s 57.6%; I assume the second quarter’s prepayments were a catch-up that the commitments already cover in part, a judgment the stress run below tests.E (estimate)

The base case spends RMB 1,131B in the first five years, 3.0 times the RMB 372.9B committed. To the extent the commitments are capital or licenses (RMB 92.7B of capital commitments, plus game licensing and content inside other commitments) they are inside that spending, not on top of it; infrastructure operating costs and technical services run through the margin I assume. Each RMB 1 of all-in spending in years 1 to 10 adds RMB 0.32 of revenue in my base case, against RMB 0.95 from 2016 to 2025; the cases table has the bear and bull, and the operating profit each RMB 1 adds.C (calculated)⁶⁹¹⁷

The bear case is an arms race without pricing power. The build runs at 36% of revenue in year 1, between the first half’s 27.6% without the compute prepayments and 40.4% with them, and settles at 21%. In year 2 the worst year on record repeats: in 2022 advertising fell 7%, everything else was flat, and the margin before depreciation fell 1.3 points.C (calculated)¹²¹¹ The bear’s margin starts at 38.5% and falls 2 points.E (estimate) A separate stress run, outside the weights, keeps the bear but puts all-in spending at the June quarter’s 58% of revenue for two years, then 40%, 30% and 21%: it gives HK$214 a share, below the bear, so the bear is not the floor.E (estimate) Tencent says it can rent the compute out, which would soften both.

My bull case is the one where AI pays inside Weixin and Tencent Cloud. Marketing Services grows 22% and fades to 6%, the rest of the business grows 9.5% and fades to 5.5%, the margin before depreciation reaches 49%, and capital spending falls to 15% of revenue. I found no consensus estimates I could verify, so none is used. The bear and bull move all inputs together. Weighted 25%, 50% and 25%, the cases give HK$503 a share against a price of HK$383.60.E (estimate)⁵⁹

Margins and capital intensity, the two things the AI build changes, move my value as much as revenue does. Each 2 points of long-run capital spending is worth about HK$27 a share.E (estimate)⁶⁹ The commitments are small beside that: if the RMB 225.0B rise in other commitments were dead cost with no revenue, paid over years 1 to 3, my base would fall by HK$20 a share, or HK$26 with the rise in capital commitments.E (estimate)⁶⁹

I discount each year’s cash flow at year end and date the value from 30 June, the last balance sheet, as the Alphabet page does. Mid-year timing and dating the values to Q4 2026 would lift every case by about 7%: bear HK$278, base HK$542, bull HK$787.E (estimate) I keep year end for the stated values and show both below. The model gives no credit for the RMB 81.1B of prepayments if they are prepaid services, or for value in cloud or AI beyond the revenue and margin I give them.⁶

The margin of safety

• PriceHK$383.60Close, HKD counter 00700, Q4 2026 [59]
• Base valueHK$508My estimate E (estimate)
• Margin of safety belowHK$35530% under my base value C (calculated)

My rule is 25% below base for a business with net cash and no customer above a tenth of revenue, 30% when one cycle or two customers drive results, and 35% when fixed charges such as leases sit ahead of the shares. Tencent has net cash and no large customer, so the rule gives 25%. I add 5 points, worth HK$25 a share, for risks the cash flows do not price. Tencent holds its Chinese licenses through Structure Contracts with four individuals, and the companies under them carried about RMB 430B of 2025 revenue, 57% of the group, on net assets of about RMB 27B; the annual report calls the legal position uncertain. The US Department of Defense also listed Tencent under Section 1260H in January 2025 and again in June 2026, but the stated effect is on US defense contracting and not on Tencent’s revenue, so it weighs less.C (calculated)⁹²⁹³⁰ The 5 points are a judgment specific to this page. So I set the line at 30% below base, HK$355, and show the others:

The price is 8% above the HK$355 line. On mid-year timing dated Q4 2026 it is 1% above the HK$380 line and 6% below the 25% line of HK$407. The 25% tier assumes net cash, which fell to RMB 58.2B against RMB 453.0B of debt; the 30 September balance, with no dividend in it, will show the underlying rate.C (calculated)⁷⁸

The 52-week closing high was HK$657.00 on 12 November 2025 (CNBC daily bars; the exchange file for that date was not available), and the stock has fallen 42% since. It fell 6.8% on the day after the 2025 results, when Tencent said it would repurchase fewer shares, and 4.5% on the day after the June-quarter results.C (calculated)³⁵¹⁹⁷ The last 11%, from 30 September to the Q4 2026 close, has no results announcement behind it; the press blames AI-spending worries and a de-rating of Hong Kong tech.³⁴ The shares closed at HK$377.60 on 29 October, the lowest close in the 52 weeks, and rose 1.6% on 30 October to HK$383.60.⁶⁰⁵⁹

The next best use of the money

My base returns 12.4% a year over ten years from this price, counting owner cash paid out. Meta, the closest comparable here, returns 6.8% a year on my base, as its page shows. The business and the price are two questions. The business earns a 56.7% gross margin on 1,439 million users. The price is 24% below my base but 8% above my margin-of-safety line, and what it pays for depends on capital I cannot yet know.⁷

My base needs three things together: advertising growth fading from 20% to 5%, capital falling from 28% to 17% of revenue by year 5, and the margin before depreciation climbing to 45%. At 60%, 50% and 60% those multiply to 18%, about one in five or sixE (estimate). That is my judgment. The price needs only capital under 26%.

Where this could be in 5 and 10 years

For my base to hold at five years, revenue has to reach RMB 1,191B by June 2031, with owner free cash flow of RMB 270B, 23% of revenue C (calculated). Capital spending has to be back to 17% of revenue. At ten years it needs revenue of RMB 1,523B and owner cash of RMB 344B, still 23% E (estimate). Value climbs from HK$508 as the cash comes nearer.

The cases differ on growth, margin and capital. In the bear, spending is 36% of revenue in year 1 and 21% for good, and year-10 owner cash is RMB 163B. In the bull, spending settles at 15% and owner cash reaches RMB 498B, 27% of revenue. The discount rate is 10.0% in all three. At 12.0% my base is HK$399 today. The price is above my bear value, so the bear returns 5.3% a year, under 10.0%.

The most likely way my base is wrong is capital spending. If it stays at 21% of revenue from year 6 and not 17%, the base is HK$611 at five years and HK$780 at ten, about 11% to 12% lower E (estimate). Today’s price needs 26% from year 6.

These are estimates of value, not forecasts of the share price. The market price can sit above or below value for years.

VIIThe case for the price

What most people miss

The headline profit line hides how the old business is doing. In the June quarter non-IFRS operating profit, Tencent’s adjusted measure, rose 9%. Without the new AI products it rose 19%, to RMB 86.1B, and that margin went from 39% to 42%.⁷ The new products cost RMB 10.5B in the quarter, and the split is Tencent’s own. C (calculated)⁷ Martin Lau, the president, also told the August call that compute ordered a few months earlier could be resold “at more than 30% profit”.⁶³ That is his claim, and no filing shows it. If he is right, the money prepaid for compute is not stuck.

The bull case

The argument is that this spending is a bulge and the market treats it as permanent. On the 12 August call Lau called the AI-native spending “a lump sum” to be spent this year and next. James Mitchell, the chief strategy officer, said Tencent “could recover the depreciation almost immediately” by renting the compute out, though Tencent is putting much of it into its own models and apps.⁶³ The broker views here are secondhand: I read press reports, not the notes. BigGo Finance reported on 13 August that Gary Yu’s team at Morgan Stanley cut their target to HK$550 and expect earnings largely flat from the second half of 2026 through 2027, but see spare compute rented through Tencent Cloud as protection.⁶⁶ Minichart’s excerpt of a 21 October CGS International note says it kept HK$635, using an 11.2% discount rate, higher than my 10.0%, and expects heavy AI spending to squeeze margins now and pay back later.⁶⁵ Tiger Brokers’ feed reported Alex Liu at Bank of America arguing on 13 July, before the June-quarter results, that the market underrates Tencent in AI agents, where WorkBuddy leads China by users.⁶⁴

Against reality

The gap from the price to my base is HK$124. One assumption alone can close it. The price needs all-in capital spending of 26% of revenue for good, my base has 17%, and each 2 points is worth about HK$27 a share. E (estimate)⁶⁹ All-in spending was 28.0% in the last twelve months. So the market is pricing roughly today’s build forever.

If Marketing Services grows near 20% this year and the rest near 8%, both fading as I model, and all-in spending falls to 17% of revenue by year five, HK$508 is possible. Revenue is on that path today. Marketing Services grew 22% against the 20% I need, and everything else grew 8.4% against 8%. C (calculated)⁷ The margin before depreciation, 41.5%, is above the 40.5% I assume for the next two years and 2 points short of the 43.5% I need by year five. C (calculated)⁶¹⁸

Spending is the far-off part. It has to fall 11 points in about four years, and no reported broker forecast has it falling in renminbi. Bank of America, reported in July, has RMB 185B, RMB 225B and RMB 250B for 2026 to 2028.⁶⁴ CGS has RMB 200B, then RMB 230B.⁶⁵ Morgan Stanley has RMB 200B in both years.⁶⁶ On the company’s narrower definition, as I read them, the 2027 figures are 22% to 25% of my model’s revenue, against 10.5% of revenue in 2025. E (estimate)⁹ Two more heavy years would not settle it: the spending above 17% in my years one to four is worth HK$23 a share. E (estimate) What decides it is the level after year five, and no forecast I found goes that far. My base also equals the price at a 12.4% discount rate, and no filing shows when a market stops asking for that. E (estimate)

What they have to hit

Reports are on Tencent’s investor page (opens in a new tab) and HKEXnews. Years are from my model, which starts in July 2026.

Where I come down

The 19% moved me, because it supports the margin in my base. The lump sum did not, because commitments reached 4.9 times their December level in six months and Bank of America, as reported in July, has spending still rising in 2028. C (calculated)⁶⁶⁴ I would raise my values if all-in spending settled below 17%, which adds about HK$13 a share a point, and if rented compute showed up in reported returns. E (estimate)

VIIIRisks, and what would change my view

The risk Tencent’s own numbers point to is the one I weight most: a build that costs more and earns less than the company says. The company calls it downside-protected, and its commitments are above.⁸⁶

1. The AI build. Capital spending was 25.8% of revenue in the June quarter on the company’s definition and 57.6% all in, free cash flow was negative, and debt rose 17% in six months to RMB 453.0B.C (calculated)⁷⁶ The new AI products cost RMB 19.3B net in the first half and Tencent expects to more than double them in 2026.¹⁹⁷ My bear case prices a build at 36% of revenue in year 1 and 21% for good; the stress run at 58% for two years gives HK$214.
2. Funding. Tencent sold US$4.25B and CNY 41B of notes in 2026, in June and October. The ten-year dollar coupon rose from 5.00% to 6.10%, and Reuters reported orders above US$7.4B for the October dollar notes and CNY 105B for the renminbi notes, about US$23B together.²⁶²⁷³²
3. Shareholder returns. Buybacks fell from HK$112.0B in 2024 and HK$80.0B in 2025 to HK$24.4B in the first half of 2026, and the company has said it will repurchase less; dividends are rising, at a 1.4% yield.¹⁰⁹⁶¹⁹
4. Competition in AI. Doubao, Qwen and DeepSeek had 382 million, 167 million and 130 million monthly users in May 2026; Yuanbao’s figure is not in the report I could read.⁴² Weixin’s agent Xiaowei is in a small prototype test, and ByteDance is private.⁷
5. China rules. Games, advertising, data and minors are regulated. In 2021 minors’ time on domestic games fell 88% in one quarter after new limits, and in December 2023 the regulator published draft online-game rules.¹³⁴³ It approved 206 domestic titles in September 2026.⁴⁴ The Structure Contracts, which cover the RMB 430B of revenue above, carry what the annual report calls substantial uncertainties.⁹
6. The US list. Tencent is on the Section 1260H list of Chinese military companies. It says the listing is a mistake, relates only to US defense procurement and does not stop anyone else from dealing with it or its securities. A law-firm note dates the contracting restrictions at 30 June 2026 for direct and 30 June 2027 for indirect dealings with the Department of Defense.²⁸³⁰³¹
7. Investments and Prosus. Investments of RMB 875.1B are 30% of the market value at the price, and listed holdings fell 28% in six months.C (calculated) Prosus has cut its holding from 33.17% to 22.5% since 2017, and any further sale is an overhang.⁷¹⁷⁹⁵⁰
8. Governance. One founder is chairman and chief executive, the annual report names no successor, and awards carry no performance targets.⁹

What would move me

The thresholds are my judgments. The first test comes on 18 November 2026, when Tencent reports third-quarter results.⁴⁶⁴⁵ I will read four lines first: capital spending, free cash flow before prepayments, net cash and prepayments on the balance sheet. The commitments note will not be there: last year’s third-quarter announcement had none. Tencent publishes it in the interim and annual reports, so I expect the next figure with the 2026 annual report, around April 2027.⁶⁷⁹