ByteDance's H1 Net Profit Drops to $20 Billion as AI Spending Weighs on Earnings
nashnova research
ByteDance's first-half 2026 net profit dipped to roughly $20 billion even as revenue rose about 30% to $120 billion — the gap traces directly to a massive AI spending ramp, marking the moment the company once called a cash machine began trading today's profits for tomorrow's AI stack.
Revenue up, profit down — where did the money go?
First-half revenue hit roughly $120 billion, up about 30% year-on-year — slightly faster than the prior two years.
Yet net profit slipped by a single-digit percentage to around $20 billion.
This means → costs grew faster than revenue; the main drag is AI investment — compute procurement, in-house chip development, and large-model training.
In plain terms = ByteDance isn't losing customers; it's pouring every extra dollar straight into AI.
How far has TikTok's international business come?
Overseas revenue (mostly TikTok) now exceeds 30% of total revenue, up from 30% for full-year 2025 and 25% in 2024.
TikTok is accelerating U.S. e-commerce expansion, courting established brands for TikTok Shop.
In January, ByteDance sold 80% of TikTok's U.S. data-security unit to a U.S.-controlled joint venture but retained the revenue-generating operations.
This means → TikTok shed its biggest political liability while keeping the ad and e-commerce engine that actually makes money.
Stacked against Meta, is ByteDance undervalued?
Meta posted 30% revenue growth to $117 billion over the same period — roughly matching ByteDance's $120 billion.
But Meta's market cap sits at about $1.7 trillion; ByteDance's secondary-market valuation is roughly $630 billion (per CapLight data).
This means → at comparable revenue, ByteDance is valued at less than 40% of Meta.
This reflects a liquidity discount for a private company and the market's pricing of geopolitical risk around Chinese tech.
The AI arms race — where is ByteDance spending?
The company recently borrowed roughly $300 billion from banks — its largest-ever loan — with part earmarked for AI investment.
It is developing a custom inference chip — a chip designed to run trained AI models, not to train them.
Its Doubao app is China's most popular mobile AI assistant; its Seedance video-generation model ranks among the global leaders.
In plain terms = ByteDance is betting across all three layers — chips, models, and apps — wagering on full-stack AI autonomy.
Why is Zhang Yiming refusing to copy homework?
Founder Zhang Yiming stated at a July internal meeting that ByteDance will not use "distillation" — training its own models on outputs from U.S. frontier models, essentially using someone else's answers as a textbook.
He accepts that this may mean falling behind domestic rivals in the short term.
ByteDance is the only major Chinese model maker keeping most of its models closed-source; enterprise clients access them through its own cloud platform.
This reflects a bet not on near-term rankings but on a long-term technical moat and cloud-ecosystem lock-in.
When will this spending round show returns?
Seedance's video model is being commercialized, aiming to drive more revenue into the AI and cloud business.
In June the company launched "Doubao Work," a workplace AI assistant for professionals, with paid subscription tiers.
But ByteDance's own cloud platform still trails Alibaba Cloud, Asia's largest cloud provider, in market share.
This means → whether AI spending converts into visible cloud-revenue gains over the next few quarters is the key test of the investment thesis behind this entire ramp.
市场有风险，内容仅供研究参考，不构成投资建议。