ByteDance’s Global Business: TikTok, Douyin, Commerce and AI

ByteDance doesn’t just own TikTok. It runs one of the most valuable private technology companies on the planet^[1] — valued at US$550 billion as of February 2026^[2] after a General Atlantic secondary sale confirmed the benchmark (TechBuzz). Founded by Zhang Yiming in Beijing in 2012,^[3] the company generated US$155 billion in revenue for full-year 2024, growing 29% year-on-year,^[4] and about US$200 billion in 2025, according to media reports.^[5] Net profit for 2025 was reported at about US$42 billion,^[5] although other reports say profit fell sharply on an IFRS basis as AI spending rose. The portfolio spans TikTok (2.04 billion monthly active users globally), Douyin (China’s equivalent, with more than 700 million daily active users),^[6] TikTok Shop (US$64.3 billion in GMV for 2025),^[7] the Lark enterprise suite, and a rapidly expanding AI infrastructure division backed by roughly 200 billion yuan (about US$28 billion) in planned capital expenditure for 2026.^[8] That’s not a social media company. It’s a full-spectrum digital conglomerate with a commerce layer, an enterprise arm, and an AI engine — operating across every major region simultaneously.

How big is ByteDance, really?

The numbers are almost difficult to contextualise. ByteDance has moved ahead of Meta as the world’s largest social media company by quarterly revenue: in Q2 2025 it posted US$48 billion against Meta’s US$47.5 billion.^[9] The valuation trajectory tells the story of accelerating investor confidence: T. Rowe Price valued the company at US$450 billion in mid-2025,^[4] a November secondary trade priced it at US$480 billion,^[4] and the February 2026 General Atlantic transaction established a US$550 billion mark.^[10] A proposed equity sale in April 2026 put it at more than US$600 billion.^[10]

TikTok alone reaches 2.04 billion monthly active users globally, adding 180 million users since the US ownership restructuring was completed in January 2026 (DemandSage). Indonesia is TikTok’s largest market at 180.1 million users, followed by the United States at 153.1 million^[11] and Brazil at 91.7 million.^[12] Its audience grew 17% year-on-year in 2025, ahead of LinkedIn at 14% and Instagram at 13% (Hootsuite).^[13] Douyin, meanwhile, continues to dominate in China with an estimated 714 million daily active users^[6] and an ecommerce GMV of approximately ¥3.5 trillion (US$485 billion) in 2024, up 30% year-on-year.^[14]

ByteDance employs roughly 120,000 people^[3] across offices in Beijing, Shanghai, Singapore, London, Dublin, New York, and Los Angeles. The headcount has stayed relatively flat since 2023 as the company shifted focus from expansion to efficiency and AI integration.

How does ByteDance make money?

Advertising remains the core revenue engine, accounting for roughly 60% of total revenue.^[4] The model is familiar — algorithmic content feeds generate enormous attention, which ByteDance monetises through targeted advertising across TikTok, Douyin, and its other content platforms including Toutiao (news aggregation) and Xigua Video. What’s less appreciated is how effective ByteDance’s recommendation algorithm is at converting ad spend. TikTok’s average revenue per user (ARPU) continues to climb, particularly in the US and Europe where brand budgets have shifted meaningfully toward short-form video.

Ecommerce is the fastest-growing revenue category. TikTok Shop’s global GMV reached US$64.3 billion in 2025, up 94% on the previous year (Momentum Works).^[7] Douyin’s ecommerce operation adds another ¥3.5 trillion in China.^[14] ByteDance takes a commission on transactions, charges for promoted listings, and increasingly captures logistics fees as it builds out fulfilment infrastructure. The enterprise segment — primarily Lark — is smaller but strategic. Lark has signed over 1,000 enterprise customers including Xiaomi, Nio, Li Auto, and Xpeng,^[15] positioning itself as the collaboration suite of choice for Chinese tech companies expanding globally (KrASIA).

What’s the TikTok Shop strategy?

TikTok Shop is ByteDance’s most ambitious bet outside China, and the execution has been remarkably fast. The US became TikTok Shop’s largest single-country market in 2025, with US$15.1 billion in GMV — up 68% year-on-year^[7] (Momentum Works). Southeast Asia remains the anchor for scale at US$45.6 billion in regional GMV,^[16] with Indonesia, Thailand, and Vietnam as the top three markets.^[17]

The commerce model is distinct from Amazon or Shopee. TikTok Shop’s conversion funnel starts with content — short videos and livestreams — rather than search intent. Live commerce now accounts for 14% of GMV, up from 10% in 2024,^[7] with a 7.4% conversion rate that exceeds traditional ecommerce by more than threefold^[18] (Resourcera). The average order value sits at US$58^[19] and is climbing.

For 2026, TikTok Shop is expanding aggressively into Europe and Japan, markets that ByteDance views as the next US$770 billion growth engine (Nova Analytics).^[20] Brazil, which launched in 2025, generated US$400 million in GMV within months.^[21] In the US, the platform also announced plans to integrate logistics more directly into its infrastructure from March 2026 to standardise fulfilment and improve delivery traceability, though the mandate was paused in February.^[22] The strategic logic is clear: own the entire funnel from discovery to delivery, exactly as Douyin has done in China.

How is ByteDance investing in AI?

The AI spend is staggering. ByteDance initially planned US$23 billion in AI infrastructure capital expenditure for 2026, up 7% from approximately US$21 billion in 2025 (Financial Times),^[23] and later raised the budget to about 200 billion yuan (US$28 billion).^[8] Nearly half of the initial 2026 budget — about 85 billion yuan (US$11.8 billion)^[23] — is earmarked for AI chip procurement. The company plans to spend roughly US$14 billion on Nvidia chips alone,^[24] including test orders for 20,000 H200 units,^[25] assuming US export controls permit the sales (South China Morning Post).

The consumer-facing results are already visible. ByteDance’s Doubao chatbot has become China’s most downloaded AI application,^[26] outpacing rivals from Baidu and Alibaba on user engagement metrics (Caixin Global). The underlying Doubao large language model lags behind Alibaba’s Qwen and DeepSeek on independent benchmarks, but ByteDance is betting that distribution matters more than raw model performance — and with Douyin’s user base as a distribution channel, that’s a credible wager.

An additional 20 billion yuan is allocated for LLM upgrades and applications, with 5 billion yuan reserved for AI talent recruitment.^[26] The applications span content recommendation, ad targeting, commerce personalisation, and real-time translation across Lark. ByteDance’s AI strategy isn’t about building a foundation model to rival OpenAI. It’s about embedding AI into every product surface where it can improve conversion, engagement, or efficiency.

What’s happening with the US TikTok ban?

The regulatory saga that defined ByteDance’s risk profile for three years has largely resolved — though not without drama. The US Supreme Court unanimously upheld the Protecting Americans from Foreign Adversary Controlled Applications Act (PAFACAA) in January 2025,^[27] requiring ByteDance to divest TikTok’s US operations or face a ban (Supreme Court, TikTok Inc. v. Garland). TikTok briefly went dark for US users on 19 January 2025^[28] before President Trump issued executive orders extending the deadline.

The resolution came in December 2025,^[29] when ByteDance signed paperwork to divest TikTok’s US entity to a new joint venture — TikTok USDS Joint Venture LLC — comprising Oracle, Silver Lake, and Abu Dhabi’s MGX as managing investors (Center for American Progress). The transaction was completed on 22 January 2026,^[30] with the US government approving the restructured ownership. ByteDance retains a minority stake and continues to license its recommendation algorithm to the US entity, though the data infrastructure now sits under Oracle’s sovereign cloud.

The ownership change removed the single largest regulatory overhang on ByteDance’s valuation and directly catalysed the jump from US$480 billion to US$550 billion within weeks.^[2]

What are the risks?

Three deserve attention. First, geopolitical exposure hasn’t disappeared. The US divestiture resolved one jurisdiction, but ByteDance faces ongoing scrutiny in Europe, India (where TikTok remains banned since 2020),^[3] and potentially other markets where data sovereignty concerns are intensifying. The algorithm licensing arrangement with TikTok USDS will face continuous regulatory review, and any deterioration in US-China relations could reopen the question entirely.

Second, TikTok Shop’s growth is subsidised. ByteDance is prioritising volume over monetisation, absorbing logistics costs and offering seller incentives that compress margins. That’s the right strategy for market capture, but it means the commerce business isn’t yet proving it can generate sustainable profit at scale. Douyin’s ecommerce operation in China is more mature and profitable — the question is whether that model translates to markets with weaker logistics infrastructure and lower average order values.

Third, the AI spending is enormous relative to visible returns. Roughly US$28 billion in annual capex demands^[8] either breakthrough product applications or sustained revenue growth to justify the outlay. If AI-driven improvements to ad targeting and commerce conversion don’t materialise at the expected rate, that spend becomes a drag rather than an accelerant.

What’s the outlook?

ByteDance enters 2026 in a position that would’ve seemed implausible three years ago. The US ban risk is resolved. Revenue is approaching US$200 billion.^[5] Profitability is substantial, though heavy AI spending is weighing on it. And the commerce and AI verticals provide growth optionality that pure advertising businesses lack.

The strategic priorities are clear: scale TikTok Shop to profitability, particularly in Europe and Japan; maintain Douyin’s dominance in China against competition from Kuaishou and Xiaohongshu; convert AI infrastructure spending into measurable product improvements; and navigate the complex governance structure of TikTok’s new US joint venture without operational disruption.

An IPO remains the elephant in the room. ByteDance’s private valuation, put at more than US$600 billion in a proposed share sale in April 2026,^[10] makes it the most obvious pre-IPO candidate in global tech, but Zhang Yiming — who stepped down as CEO in 2021^[3] but retains significant influence — has shown no urgency to list. The secondary market provides liquidity for early investors, and going public would introduce disclosure requirements and quarterly earnings pressure that ByteDance has so far avoided.

What’s harder to replicate is the underlying engine. ByteDance built the most effective content recommendation system in the world, then layered commerce, enterprise software, and AI on top of it. The competitors are all working backwards — Amazon trying to add entertainment, Meta trying to add commerce, Google trying to hold attention. ByteDance started with attention and is converting it into everything else. Whether a valuation above US$600 billion proves conservative or stretched depends on one question: can ByteDance turn content-driven commerce into a margin structure that justifies the price? The 2025 numbers suggest it can.