Tencent Is Spending Billions on AI and Still Getting Richer. The RMB 708 Billion Portfolio Is a Free Option.

At HK$456.20 on July 14, 2026, the market valued all of Tencent at about HK$4.1 trillion. Most of that price is the business everyone knows: WeChat, the games, the payments network. Part of it is something owners tend to forget they hold.

Tencent is also one of the largest technology investors in the world. It owns listed stakes in Pinduoduo, Meituan, Sea, Kuaishou and Bilibili, alongside private positions in Riot Games and Supercell. At the end of 2025, that portfolio was carried at RMB 708 billion. Call it HK$820 billion. It sits on the balance sheet, marked to market, repriced every day the market opens.

Add RMB 147 billion of net cash on top. Together they come to roughly a fifth of what the market pays for Tencent today. Neither one is the business that runs WeChat.

Strip both out, and that operating business is priced near 11 times its earnings. NetEase, a smaller rival carrying no comparable portfolio, trades at 12.4.

Operating business ~HK$3.1T · WeChat, games, fintech Listed portfolio RMB 708B / ~HK$820B Net cash RMB 147B / ~HK$170B

What Tencent Actually Does

Tencent Holdings (0700.HK), one of the largest technology stocks on Hong Kong's HKEX, runs the operating system of Chinese daily life. WeChat reaches 1.43 billion monthly users. In most of China, leaving the house without it is closer to leaving without a wallet than without a phone.

The business has three parts. Games and social networks are roughly half of revenue, spanning Honour of Kings at home and a foreign catalogue built on Riot and Supercell. Advertising, sold mostly inside WeChat, is about a fifth. Payments, cloud and business services make up the rest. This is not a startup and not a bet on one technology. It is a mature platform that prints cash.

The 2025 numbers were the strongest in years. Revenue rose 14% to RMB 752 billion. Profit rose 16% to RMB 225 billion. Gross margin, the share of revenue left after the direct cost of delivering it, expanded to 56% from 53%, and held at 57% in the first quarter of 2026. The expansion came from all three segments at once, not from accounting mix.

Along the way, Tencent spent fifteen years funding a generation of Chinese internet companies. Unlike most strategic investors, it mostly kept the shares. The business threw off RMB 190 billion of free cash flow last year and returned HK$121 billion to shareholders through buybacks and dividends. Folded into the same balance sheet is the RMB 708 billion portfolio, and the share price barely acknowledges it.

Why the Portfolio Counts for So Little

The market prices operating companies on the cash they generate. A portfolio of minority stakes in other businesses generates no cash the parent can spend. The dividends are small, the gains are on paper, and the holder is under no obligation to ever sell. So investors discount it, often heavily.

There is a live example of how heavily. Prosus, the Dutch holding company, owns about 23% of Tencent, a stake that dwarfs the rest of its assets. For years it has traded at a wide discount to the value of that single holding. The market does the same thing to holding structures everywhere: value that sits one layer removed from cash flow gets marked down until someone moves it.

Tencent's portfolio carries the same penalty for a reason that is at least defensible. The stakes are minority positions. Selling RMB 708 billion of them into the open market would take years and move prices against the seller. The equity-method earnings they produce are lumpy, swinging with the Hong Kong market, and the operating multiple ignores them by design. On a spreadsheet, the portfolio is a number with no obvious date attached.

Then there is the AI worry layered on top. Through the first quarter of 2026, Tencent spent RMB 37 billion building AI infrastructure. Analysts asked whether the capital was burning a hole in the balance sheet.

Net cash grew by RMB 40 billion that quarter.

The spending is producing shippable output, not just data centres. In July 2026, Tencent released Hunyuan Hy3, a 295-billion-parameter open-weight model it benchmarks against systems several times its size. A reported AI revenue line has not yet appeared, and that disclosure is one of the catalysts the bull case waits on.

That worry is about the money Tencent spends. A sharper one runs the other way: if artificial intelligence lets anyone assemble software cheaply, are the companies inside the portfolio the ones about to be undercut?

The answer turns on what the portfolio actually holds, and it is not enterprise software. The five largest listed stakes are Pinduoduo and Meituan, two marketplaces worth close to US$28 billion between them, Sea in South-East Asia, and the video platforms Kuaishou and Bilibili. Together those five carry roughly US$40 billion of the book. Not one of them is the seat-priced business tool a weekend AI project can now rebuild. A marketplace's moat is its merchants and its delivery fleet, a game's is its players and its catalogue, a video platform's is the creators already posting there. None of that is a few thousand lines of code.

Where the AI question does bite is narrower. If people start to search and shop through an assistant instead of scrolling a feed, the holdings that live on discovery and advertising, Kuaishou and Bilibili here, lose some of the attention they sell. That risk is real, and it sits on roughly a tenth of the portfolio rather than the whole of it. It also runs both ways: the same companies use AI to target ads and generate content, and Tencent ships its own model rather than renting one.

One structural fact caps the damage. The listed stakes are marked to market. If the businesses inside them lose value, the RMB 708 billion falls with them the same day, in public, with no write-down to argue over. The portfolio cannot be carried above what the market already pays for it. The thesis does not need these companies to be immune to AI. It needs only that whatever they are worth on the day Tencent hands them over reaches shareholders whole, as JD's value did.

The reasons for the discount are real, and the AI question adds one more. What none of them does is make the value vanish. They lock it in place until Tencent chooses to release it.

None of this makes the RMB 708 billion a fixed sum an owner is owed. It is a live mark that moves with the holdings, and the largest of them shows how far. Pinduoduo trades at roughly half its 2021 high: its Temu arm ran on the US de minimis exemption that Washington scrapped in 2025, and Chinese consumption has cooled on top. Because the book is marked to market, that decline already sits inside the number rather than hiding beneath it. Hold the portfolio in mind as a bonus on the operating business, one that could arrive soon through a distribution but is neither promised nor fixed, not as value the price is quietly owed.

Tencent Has Done This Before

This is not a thought experiment. The mechanism that crystallizes the portfolio has already run once, in full view, with a single corporate action.

1. 01 In December 2021, Tencent distributed its entire JD.com stake directly to its own shareholders. Roughly 457 million JD shares, worth about HK$127 billion, were handed out as a special dividend paid in shares rather than cash. Holders woke up owning JD.com directly. No sale into the market, no discount from dumping a block, no corporate tax drag on the transfer. The value moved from inside Tencent to its owners in one step. Tencent's stake in JD fell from about 17% to roughly 2%.
2. 02 The logic is repeatable, and management has leaned on it since. Tencent has trimmed other mature holdings in the years after the JD distribution. The pattern is consistent. Three conditions line up: the stake has matured, the parent no longer needs the capital, and the market gives the position no credit. When they do, the cleanest way to surface the value is to hand the shares to the owners. The 2021 distribution proved the move is real, not theoretical.
3. 03 Every condition that made 2021 attractive is present again. The portfolio stands at RMB 708 billion. The operating business funds itself, with RMB 190 billion of free cash flow and net cash that grows even through heavy AI spending. And the share price gives the portfolio almost no weight. That is the same setup that preceded the JD distribution, which is why the portfolio is the asymmetric part of this case, not the WeChat business.

What the Operating Business Is Worth

The valuation here is arithmetic, not forecast. Start from the market value, credit the cash and the portfolio at what the balance sheet already carries them, and read what is left on the operating business.

On next year's consensus earnings, the same business sits under 10 times. The headline multiple, before stripping anything out, is 13.5 times, already a 13% discount to the sector average.

The downside assumes the RMB 708 billion portfolio is worthless. The balance sheet, repriced every day, says it is not.

The upside does not depend on the market changing its mind. Prosus shows the discount can sit unrecognized for years. What it depends on is Tencent releasing the portfolio to its owners, the way it released JD.

The Risks We Are Not Downplaying

Two slower forces sit underneath all four. Prosus continues to distribute Tencent shares to its own holders at a few percent a year, adding a steady trickle of supply to the float. And the portfolio itself swings reported profit with the Hong Kong market, which makes any single quarter's earnings noisier than the operating business alone.

The Decision

Written at HK$456.20 on July 14, 2026.

This is a quality business at a fair price, carrying an investment portfolio the market values at close to nothing. The portfolio is the kicker, not the reason to own the shares. The operating company compounds at low-to-mid teens and funds its own AI build without dilution. The action that closes the gap between the portfolio's carrying value and its price is one Tencent has executed before.

This framework does not produce a buy signal or a sell signal. It produces a single question. What is that RMB 708 billion portfolio worth, when the company that owns it has handed one out before and now trades as if it never will?

Three signals would break the case. First-half 2026 profit, scheduled for August 12, that falls below the run-rate the operating multiple assumes. Gross margin slipping under 53% in any reported quarter. Or a hard regulatory turn that reopens the 2021 playbook. The May 20 annual meeting already cleared the near-term governance check, re-authorising the buyback mandate for another year.

Last year, Tencent spent HK$80 billion buying back its own shares at an average of HK$521.55. The stock last traded at HK$456.20 on July 14, more than 12% below that repurchase average.

I will revisit this if Tencent announces another in-kind distribution, or if the August 12 results break the line that keeps the operating business cheap.