ByteDance Develops Custom CPUs to Ease AI Infrastructure Bottleneck
ByteDance is building its own central processing units to support its AI expansion as chip shortages and price increases squeeze its infrastructure plans, three people familiar with the project said.
The Beijing-based company, which owns TikTok, is targeting internal deployment of proprietary CPUs in its own servers and data centers. It plans to use the chips to support agent-based products including its Coze platform, the first source said. ByteDance has approached external partners to help with design work and secure manufacturing capacity at foundries.
The project remains early stage. ByteDance did not respond to a request for comment.
Two Design Paths
ByteDance is pursuing two architecture tracks: one based on Arm and another on the open-source RISC-V instruction set. Developing two designs simultaneously is standard practice among large tech companies, allowing them to test options before committing to expensive manufacturing runs.
This dual approach reflects uncertainty about which architecture will best suit the company's long-term data center needs.
Pressure From Supply Constraints
The move comes as CPU availability tightens across the industry. Intel warned Chinese customers in February of server CPU delivery lead times reaching six months. The chipmaker said in April that demand from AI firms was so strong in the first quarter that it sold chips it had previously written off as unsalable.
AMD CEO Lisa Su warned last week that the global CPU market is "tight," with demand outpacing forecasts and supply constraints expected to persist.
ByteDance sources CPUs from Intel and AMD. Both vendors have raised prices significantly, with quarter-over-quarter increases ranging from 10% to 35% in recent months, two sources said. Intel cited sustained demand, increased component and material costs, and evolving market dynamics for its price updates. AMD did not respond to a request for comment.
Broader Industry Shift
ByteDance joins a growing list of major tech companies designing custom CPUs. Google, Amazon, and Microsoft are all developing proprietary processors to reduce costs and optimize performance for their specific workloads.
The shift reflects a change in how AI systems are deployed. The industry is moving from training models toward inference-running deployed models to perform agent-based tasks that demand more from CPUs, working alongside Nvidia's graphics chips.
This inference surge has created the CPU shortage now constraining expansion across the sector. Nvidia itself is moving beyond GPUs into CPUs. CEO Jensen Huang said the company hopes its new "Vera" processors will access a $200 billion market. Nvidia unveiled a new central processor and AI system built on technology from Groq, a chip startup specializing in inference, in March.
Intel and AMD have emerged as leading challengers to Nvidia's dominance in AI chips, partly because CPU demand has become acute.
For IT and development professionals, understanding custom silicon strategies is becoming essential. Read more on AI for IT & Development and Generative AI and LLM to understand how infrastructure decisions shape AI deployment.
Your membership also unlocks:
