ByteDance Raises 2026 AI Infrastructure Budget 25% to $29.4 Billion Amid Soaring Chip Costs

ByteDance Doubles Down on AI Infrastructure

ByteDance, the Chinese parent company of TikTok, has raised its planned 2026 AI infrastructure budget by 25%, taking the total to more than 200 billion yuan, or roughly $29.4 billion. The figure is up from a preliminary plan of 160 billion yuan that was discussed late in 2025, and reflects both the company's expanding AI ambitions and a brutal rise in memory chip prices that is reshaping budgets across the global tech industry.

A Memory Chip Crisis Hits Big Tech

The revision is not happening in isolation. TrendForce has reported DRAM contract prices rising roughly 95% quarter over quarter in Q1 2026, with a further 58% to 63% increase projected for Q2. Microsoft has attributed $25 billion of its record $190 billion capex budget to higher memory and component costs, while Meta has raised its full-year capex range to $125 billion to $145 billion. Collectively, the largest US tech companies have announced around $725 billion in AI capital expenditure for 2026, up 77% from the prior year, sending shares in memory specialists like SK Hynix to record highs.

Domestic Chips Take Centre Stage

A notable element of ByteDance's revised budget is a proportionally larger allocation to domestic AI chips, a strategic shift aimed at reducing exposure to US export controls and aligning with Beijing's drive for semiconductor self-sufficiency. Reuters has reported that ByteDance is developing its own AI inference chip, working with Samsung for manufacturing, with a target of at least 100,000 units this year and plans to scale to 350,000. Its in-house chip design division now employs around 1,000 people and has produced a processor that reportedly rivals the efficiency of Nvidia's China-market H20 at a lower cost.

Offshore Compute and Geopolitical Calculus

At the same time, ByteDance has pursued offshore arrangements to access cutting-edge foreign silicon. The company has assembled more than $2.5 billion of Nvidia Blackwell systems in Malaysia through a South-East Asian cloud partner, deploying around 36,000 B200 chips for AI research outside China. The dual approach, investing in domestic silicon while building offshore capacity for restricted foreign chips, underlines how US export rules are reshaping the AI supply chain. With memory costs climbing and geopolitical tensions persisting, ByteDance's willingness to spend close to $30 billion signals a bet that AI infrastructure will stay central to its competitive position in short-form video, e-commerce, and enterprise cloud.