ByteDance, the Beijing-based parent company of TikTok and Douyin, is deep into building its own artificial intelligence (AI) chips.
According to reports from Reuters and The Information, the company is developing custom central processing units (CPUs) designed for AI inference, the process through which AI models generate responses after they have been trained.
ByteDance’s push to develop its own AI chips started two years ago, with the pace to scale that effort picking up in 2026.
A Chip Already in Progress
ByteDance’s new inference-focused chip draws inspiration from Groq’s language processing units (LPUs), which are purpose-built for running trained models rather than the more computationally intensive training phase.
The company is also working with Chinese startup InnoStar Semiconductor on memory technology for the project. This matters because advanced AI chips often depend on expensive and limited high bandwidth memory. If ByteDance can use another memory approach, it may reduce its exposure to global HBM supply problems and export restrictions.
Why ByteDance Is Doing This Now
The short answer is cost and access. ByteDance currently sources CPUs from Intel and AMD, and both companies have raised prices significantly, with quarter-over-quarter increases ranging from 10% to as much as 35% in recent months, according to two sources familiar with the matter. That pricing pressure accelerated the Chinese company’s push for in-house alternatives.
The longer answer involves U.S. export controls. ByteDance spent over $2 billion buying more than 200,000 Nvidia H20 AI GPUs this year alone, with many of those chips not yet delivered. The H20 is itself a downgraded chip, one that Nvidia specifically designed for the Chinese market after the U.S. tightened export restrictions on its more powerful hardware.
A $23 Billion AI Infrastructure Bet
ByteDance plans to spend $23 billion in capital expenditure in 2026, with more than half of that going toward acquiring advanced semiconductors to develop AI models and applications. And even while building its own chips, the company is still purchasing heavily from outside vendors.
ByteDance plans to spend about 100 billion yuan, or roughly $14 billion, on AI chips from Nvidia in 2026, a significant increase from roughly 85 billion yuan in 2025, although that plan depends on whether the U.S. permits Nvidia to sell its H200 GPUs in China.
What This Means for the Broader Semiconductor Market
Global hyperscalers including Google, Amazon, and Microsoft are also developing their own custom CPUs to reduce costs and tailor performance to their specific AI workloads. ByteDance is joining that trend, but with an added layer of urgency that Western tech companies do not face, which is that its access to the most advanced chips can be restricted at any point by U.S. policy.
There’s also the trend of large AI companies no longer wanting to rely fully on outside chip vendors. ByteDance now appears to be moving in the same direction as Google, Amazon, and Meta, but with the added pressure of U.S. export controls and China’s push for semiconductor independence.
While ByteDance is unlikely to replace Nvidia quickly, it remains a company that runs some of the world’s most-used AI applications, including the Doubao chatbot and its popular TikTok app, while scrambling to control the hardware that makes all of it run. Ultimately, ByteDance is making clear it intends to compete on every level in the AI arms race.
