ByteDance

IWhat it does

ByteDance is a Beijing-founded technology group that is not listed on any stock exchange. You cannot buy it on a public market. Its shares change hands only in company buybacks and private secondary sales, so any price you see is a negotiated number, not a quote²¹⁷. This page is research, not investment advice.

The company was founded in 2012 by Zhang Yiming and Liang Rubo¹⁷. Liang took over as chief executive from Zhang in 2021¹⁹. Reuters reported in November 2021, citing one person, that Zhang also stepped down as chairman and that Liang took that role too²³. I found no later report that says otherwise. The parent company is incorporated in the Cayman Islands, a common route for Chinese tech groups that raise money from overseas investors²².

It runs three families of products. The first is short video: Douyin in China (launched in late 2016) and TikTok everywhere else (launched in 2017)¹⁹. The second is a set of other apps, including CapCut (a video editor) and Lemon8 (a lifestyle sharing app), which appear in the US joint-venture documents⁵. The third is artificial intelligence. Doubao, a chatbot app, is the best known of these. Sacra, a research firm, puts Doubao at 382 million active users, ahead of Alibaba’s Qwen at 167 million and DeepSeek at 129 million, though user definitions differ between trackers and I could not verify these counts against a primary source⁴.

What ties the products together is a recommendation system: software that decides what video to show next, based on what each person watches. That system learns from enormous amounts of viewing behaviour. It is what keeps people watching, and watching is what advertisers and shop owners pay for.

IIHow it makes money

Most of the money comes from advertising. Sacra estimates advertising at about 60% of revenue and live broadcasting, where viewers tip streamers with virtual gifts, at about 26%⁴. The rest is a mix of e-commerce commissions, games and cloud and AI services. I could not find a reliable split of that remainder, so I do not give one.

ByteDance does not publish audited accounts. Every revenue number on this page is therefore either press-reported (leaked or sourced to anonymous people) or an outside estimate. For 2025 the figures disagree by a wide margin. A Nikkei-sourced summary says $165B². Sacra estimates $186B, about 20% above its $155B for 2024⁴. The Information, as relayed by a news site, says about $200B, up 29%⁸. I treat the true figure as somewhere in a $165B to $200B range and use H1 2026 numbers, which are more recent, as my anchor.

The anchor is this. The Information reported that H1 2026 revenue rose about 30% to about $120B, while net profit slipped by a single-digit percentage to about $20B, weighed down by spending on AI computing⁸. That second-hand report is the best recent profit data point I found. Profit for 2025 is also disputed: Bloomberg projected about $50B, and Sacra reports a fall of more than 70%⁴. Sacra’s own text is not consistent here, because it also says about $40B had been earned in the first three quarters of 2025⁴. I do not rely on either full-year profit figure.

Geography matters. Sacra estimates that overseas revenue was about $39B in 2024, roughly 25% of the total, and rose above 30% in 2025, and that TikTok US alone was about $27B in 2024⁴. Those are estimates, and since 22 January 2026 the US business sits inside a joint venture, which changes who books what (see the ownership section).

Revenue and profit figures that disagree, and one anchor that is more recent

IIIThe market and the competition

ByteDance competes for attention and for the advertising money that follows attention. Its biggest rival in dollar terms is Meta, which owns Facebook and Instagram. Meta reported revenue of $64.6B in the quarter to 30 September 2026 and $181.7B for the first nine months, with nine-month net income of $52.5B (company filing)²⁰. On Meta’s trailing twelve months, outside data show revenue of $241.6B and net income of $75.3B¹⁵. That shows ByteDance is a similar size on revenue but, on the press-reported profit above, earns less on it.

In China, Douyin competes with Kuaishou and with Tencent, whose WeChat Channels video product sits inside a messaging app that most Chinese internet users open daily. Tencent’s own results show how big the home market is: second-quarter 2026 revenue of RMB 204.8B (up 11%) and marketing services revenue, mostly advertising, of RMB 43.6B (up 22%)²¹. I did not open sources with Kuaishou’s numbers, so I do not quote them.

Outside China, TikTok competes with Instagram Reels and YouTube Shorts. These are copies of the TikTok format, run by companies that already own the audience and the ad relationships. That is the main way TikTok’s lead could shrink without anyone banning anything.

In AI, ByteDance competes with Alibaba, DeepSeek and others inside China, and with OpenAI, Google and Anthropic abroad. Doubao is a large consumer app by user count, but I found no reliable figure for what it earns, and the user numbers vary by tracker⁴. ByteDance is spending heavily on computing. Sacra describes AI capital spending of about RMB 160B (about $23B) and says a 2026 total could reach $70B, but it does not make clear which year the RMB 160B figure belongs to, so I treat the spending as large and uncertain⁴. Tencent, for comparison, reported capital spending of RMB 52.8B in one quarter, up 176%²¹. The whole sector is spending at a rate that pays back only if AI products earn real money.

IVThe moat and how it could be disrupted

I rate the moat strong, not fortress. The advantage is real. A short-video app improves with more viewers because the recommendation system learns faster, creators go where viewers are, and advertisers go where both are. ByteDance has run that loop at global scale for years, and it has done it in two separate markets (Douyin and TikTok) with one core technology.

But the moat is thinner than it looks in three ways. First, switching costs for viewers are low. A user can open another video app in seconds, and Meta and Google have copied the format. Second, a large part of the moat is legal permission to operate, and that is the part that has been under attack. The US forced a divestiture of TikTok’s American business, and the EU is investigating TikTok’s design under its Digital Services Act (the DSA, the EU’s online platform law)¹³⁷. Third, in AI there is no clear lead yet. Doubao’s user count is large but the models are close to those of rivals, and I found no evidence of pricing power.

How it could break: a regulator forces changes to the core feed, for example limits on endless scrolling, that lower watch time. A rival product takes younger users. A second country copies the US structure and takes the local business out of ByteDance’s hands. Or the AI spending does not earn a return and squeezes profit for years. I cannot judge how likely each is, and a durable moat would need most of them to fail.

VThe numbers we can see

Because there are no public accounts, the numbers we can see are prices paid for shares. They come in three kinds, and they do not agree.

The first kind is the company’s own employee buyback, where ByteDance buys back options and shares from staff at a price it sets. In March 2025 the price was $189.90, implying about $312B to $315B. In August 2025 it was $200.41, implying more than $330B⁴. A report relayed by KuCoin says a new round in October 2026 priced at $241.35 per share, pre-tax, with former staff paid the same¹⁸. That report is second-hand and gives no total value. If the share count were unchanged, $241.35 would imply roughly $400B. That is my arithmetic (330 x 241.35 / 200.41), not a reported figure.

The second kind is outside secondary sales, where an existing investor sells to a new buyer. A block in November 2025 was reported at $480B⁴. Reuters reported on 25 February 2026 that General Atlantic, an investor since 2017, was selling shares at about $550B, 66% above the buyback valuation, to give its own funds cash; the stake size and buyer were not disclosed¹. SCMP, as relayed by a data site, put the asking valuation at over $600B in April¹⁷.

The third kind is gray-market and marketplace quotes. A Nikkei-sourced summary said shares reached about $900B in secondary trading in June 2026, with talk of $1 trillion²¹⁰. The same summaries note these are thin quotes, not completed deals, and one data site says the number came from the gray market¹⁰. As of 2 November 2026 a private-share marketplace showed a last price of $349.01, which StockAnalysis converts to an implied $723.9B, but the page does not explain how it counts shares and the figure does not line up with the buyback share counts¹⁷.

A reported $29.6B loan facility to ByteDance (September 2026) was cited alongside a $600B valuation, but the article names no source or lender, so I give it no weight⁹.

Valuation marks for ByteDance, oldest to newest. None is a public price.

VIWho runs it and who owns it

Liang Rubo is chief executive. Reuters reported in 2021 that Zhang Yiming, the founder, had stepped down as chairman²³. Reuters reported in March 2024, citing sources, that Zhang holds over 50% of the voting rights¹⁹. Sacra gives the split of shares as about 60% for international investors, 20% for founders and Chinese investors and 20% for employees⁴. Named backers include Sequoia, SoftBank and KKR⁴, and Susquehanna and General Atlantic¹⁹. If the voting report still holds, a holder of ordinary shares has no say in how the company is run. Chinese state-owned entities also bought a 1% stake in a main Chinese subsidiary in 2021, a so-called golden share, and have a board seat there²²²⁴.

What can an owner actually take out? I found no reported dividend. The practical routes are the employee buyback (open to staff), occasional investor stake sales (as with General Atlantic) and, in future, a listing that the company says is not planned¹²¹⁸. A minority outside holder who is not in one of those routes holds a claim that is hard to sell.

The big structural event of 2026 is TikTok’s US business. Under a 2024 US law, TikTok had to be sold by ByteDance or banned. On 25 September 2025 the White House issued an executive order describing a deal: a US joint venture, majority US-owned, with ByteDance below 20%, one board seat of seven for ByteDance, and the algorithm retrained on US data in an Oracle-run cloud¹⁴. The deal was reported signed in December 2025 and closed on 22 January 2026³⁵.

The final shape, as reported: Oracle, Silver Lake and MGX (an Abu Dhabi state-backed fund) hold 15% each. ByteDance holds 19.9%, just under the 20% line in the law. Other investors include the Dell family office, Susquehanna’s Vastmere, Alpha Wave, Dragoneer-linked Merritt Way, General Atlantic’s Via Nova and Xavier Niel’s family office⁵. Adam Presser, formerly TikTok’s head of operations and trust and safety, is chief executive of the joint venture, and Shou Chew sits on its board⁵. The joint venture handles data, security and the algorithm. TikTok’s global entities in the US keep commercial work such as e-commerce, advertising and marketing⁵. Reuters, citing the Wall Street Journal, reported in March 2026 that the US administration was set to receive a fee of roughly $10B from the investors in the deal. I treat that as press-reported and unconfirmed⁶.

Whether this is a clean separation is disputed. TechTimes reports that ByteDance kept ownership of the recommendation algorithm and licenses it to the joint venture for a fee, and that ByteDance keeps intellectual property over the architecture and training method¹¹. House China committee chair John Moolenaar asked on 23 January 2026 whether a ByteDance-licensed algorithm keeps Beijing from influencing it¹². A March 2026 investor lawsuit in the D.C. Circuit, still pending according to TechTimes, argues the structure breaks the law’s ban on operating relationships with ByteDance¹¹. A congressional hearing with the joint venture’s chief security officer was scheduled for 15 September 2026; I could not confirm what was said¹¹.

For ByteDance’s own value, the outcome is a trade. It lost control of the US business but kept 19.9% of it, and I found no figure for the venture’s value. The valuation of that stake is something I cannot judge.

VIIValuation: is the last price supported?

Equity value in US$ billions. Not a share price: a private company has none. Latest mark: secondary stake sale by an existing investor (press-reported); not a primary round, so neither pre- nor post-money.

What the last valuation assumes

At $550B, ByteDance would trade at about 2.2 times my 2026 revenue estimate of $250B, against 6.55 times for Meta and 4.0 times for Tencent¹⁵¹⁶. On earnings, $550B is about 14 times H1 2026 profit annualised to $40B (my arithmetic on a press-reported figure⁸), and $724B is about 18 times, close to Meta’s forward 18.4. A mark of $900B would be about 22 times, above Meta. So the lower marks assume profit stabilises and grows, while the higher marks assume profit recovers toward Meta-like margins and nothing goes badly wrong with the US structure or EU rules.

Bear, base and bull

• Bear. 2026 revenue $240B, growing 8% a year to 2030 (about $327B). Net margin 12%, so 2030 net profit about $39B. Exit price-to-earnings of 10, below Tencent’s forward 11.4¹⁶. Discount rate 12%. Dilution 2% a year. Equity value about $230B. This fits a case where EU and US rules cut growth and AI spending squeezes margin.
• Base. 2026 revenue $250B, built from H1 2026 of about $120B up about 30%⁸, with H2 expected stronger. Growth 14% a year to 2030 (about $422B). Net margin 17%, so net profit about $72B, against about $20B in H1 2026⁸ and Meta’s TTM margin near 31%¹⁵. Exit price-to-earnings 13, between Tencent’s forward 11.4¹⁶ and Meta’s forward 18.4¹⁵. Discount rate 11%, my estimate for an unlisted, high-uncertainty Chinese group. Dilution 1.5% a year. Equity value about $580B.
• Bull. 2026 revenue $250B, growing 18% a year to 2030 (about $485B). Net margin 22%, so net profit about $107B, still below Meta’s current margin¹⁵. Exit price-to-earnings 16, nearer Meta’s 18.4 forward. Discount rate 10%. Dilution 1% a year. Equity value about $1.12 trillion.

Comparables used

Where I come down

My base of $580B is near the latest sale mark of $550B, below the $600B ask and the implied $724B, and well below the $900B gray-market quotes. So the February price looks about fair on my numbers, and the later marks look high. To justify $724B or more you would have to believe revenue grows near 18% a year, profit margins climb back above 20%, and the US and EU issues cost little. The market could still be right. Buyers in thin secondary trades may know things I cannot see, such as real internal profit trends, and TikTok has repeatedly proved resilient against forecasts of decline. Scarcity also pushes prices up: few investors can own a stake in a company of this size, and the buyer pool is not large. The employee buyback price, which implies about $400B on my arithmetic, is the company’s own lower mark and sits well under most outside quotes¹⁸. Remember too that I use an arbitrary 4-year horizon and that the base is sensitive: at a price-to-earnings of 16 instead of 13 it rises by about 23%.

What I cannot judge

I cannot judge ByteDance’s true revenue or profit, because it publishes no audited accounts and 2025 revenue estimates range from $165B to $200B²⁴⁸. I cannot judge the value of its 19.9% of the US joint venture, the real size of its AI spending, Doubao’s earnings, or the outcome of the EU and US legal matters. Because of that, treat the valuation range of $230B to $1.12 trillion as the honest answer, with the base case only a midpoint.

VIIIThe IPO

No IPO plan reported. A Nikkei-sourced summary says Liang Rubo told an all-hands meeting that an IPO is "not on the table at this time"²¹⁰. That is a time-limited statement, not a ruling out. The company is reported to be buying back shares from employees and early investors instead²¹⁸. I found no filing, exchange, ticker or banks. A listing would face a hard problem, since a Chinese-controlled group with a US business structure under political scrutiny is complicated for any exchange.

What to watch

• Any change in the "not on the table" language, or a confidential filing report
• The next employee buyback price and whether it moves toward outside quotes¹⁸
• A completed General Atlantic or other large secondary sale at a disclosed size¹
• The D.C. Circuit lawsuit and any further congressional action on the US joint venture¹¹¹²
• Final decisions in the EU DSA cases, including any fine or required design change¹³⁷
• A formal ByteDance financial disclosure, which would replace the press estimates on this page

IXWhat could go wrong

Think of what is most likely to destroy value. The first candidate is the legal and political settlement over TikTok. If the D.C. Circuit lawsuit succeeded or Congress reopened the structure, the US business could be forced into a harder separation¹¹¹². A second candidate is regulators changing the product. On 6 February 2026 the European Commission preliminarily found TikTok in breach of the DSA over addictive design (infinite scroll, autoplay, push notifications and its recommender system) and said TikTok "needs to change the basic design of its service"¹³. On 24 July 2026 it added preliminary charges over children’s account privacy, with a possible fine of up to 6% of global annual turnover⁷. These are preliminary: no final fine has been imposed as of what I found. Turnover is group-wide, so 6% of $200B would be about $12B on a stylised calculation of mine, though actual fines in this area have been far smaller in past cases, and I did not verify any amount.

The third is the AI bill. Spending on computing is heavy while profit is already sliding⁸. If Doubao and AI services do not bring in real revenue, the margin stays lower than my base case. The fourth is the advertising and creator economy: ads depend on attention, and attention is fickle. The fifth is Chinese rules and geopolitics, including export controls on chips and a Chinese state role in the company (the golden share)²².

What has to go right at once for my base value to hold? Roughly four things: the US structure survives legal challenge; the EU outcome does not force a big drop in watch time; AI spending earns a return so net margin reaches about 17%; and revenue still grows about 14% a year. My rough odds are 90%, 90%, 65% and 65%. Multiplied, that is about 34%. These odds are my own guesses, not data, and they overlap, so the product is only a reminder that several things must work together.

Cost of an owner’s mistake is also uneven: a buyer of a secondary stake pays today’s price in dollars, has no vote, little information and a narrow exit. The next best alternative for the same dollar might be Meta, which trades on public accounts at a forward price-to-earnings ratio of 18.4 and trailing 20.7¹⁵, or Tencent at 11.4 forward and 13.6 trailing¹⁶. Neither carries the same legal risk over a banned app, and both can be sold on any trading day.

Last, the data itself. Much of what this page leans on is second-hand: a summary of a Nikkei story, a news site relaying The Information, and data sites. I could not open Bloomberg, which blocked access, and I did not find ByteDance’s own investor materials. Please treat each number as a range.