ByteDance, the Beijing-based parent company of TikTok and the Doubao AI assistant, is developing its own central processing units to support its growing AI infrastructure, according to an exclusive Reuters report published on May 28. Three people familiar with the matter confirmed the project, which is being driven by a combination of surging CPU prices, prolonged supply shortages, and a fundamental shift in AI workloads toward inference, the phase of AI deployment where models are actually used rather than trained.
The project remains at an early stage and ByteDance did not respond to Reuters’ request for comment. But the strategic logic is clear: as AI moves from training to inference, the computational demands on CPUs, which handle the coordination and orchestration tasks that surround GPU computation, are rising sharply, and the market is not keeping pace with demand.
The CPU Shortage Driving the Decision
The immediate trigger for ByteDance’s custom chip push is a CPU market that has become acutely constrained. ByteDance currently sources its server CPUs from Intel and AMD, and both companies have raised prices significantly in recent months. Quarter-over-quarter price increases have ranged from 10% to 35%, according to two of the Reuters sources, a cost trajectory that makes the economics of custom silicon increasingly attractive, even accounting for the substantial upfront investment required.
Intel has warned Chinese customers of server CPU delivery lead times of up to six months, a figure Reuters reported in February. Intel said last month that demand from AI firms was so strong in the first quarter that it sold chips it had originally written off as excess inventory. AMD CEO Lisa Su warned last week that the global CPU market is “tight,” with demand outpacing forecasts and supply constraints expected to persist. Intel confirmed to Reuters that it had updated prices on some products to reflect sustained demand, increased component and material costs, and evolving market dynamics.
The shortage is not unique to ByteDance. The broader shift toward AI inference is creating CPU demand across the industry. Alphabet’s Google, Amazon, and Microsoft are all developing custom CPUs to reduce costs and tailor performance to their specific workloads. Nvidia is expanding beyond GPUs into the CPU market, with CEO Jensen Huang projecting that the company’s new “Vera” central processors will give Nvidia access to a new $200 billion market.
Arm and RISC-V: A Two-Track Strategy
ByteDance is pursuing two chip architecture tracks simultaneously. One is based on Arm, the chip architecture owned by SoftBank that powers the majority of the world’s smartphones and an increasing share of data center processors. The other is based on RISC-V, the open-source instruction set architecture that has attracted significant interest from Chinese technology companies as a path to chip design that does not depend on US-controlled intellectual property.
Developing two designs in parallel is a common hedge for technology companies at this stage of chip development. It allows ByteDance to test both architectures before committing to the costly, large-scale manufacturing run that would be required for mass deployment. The company has approached several external partners to assist with design work and to help secure manufacturing capacity at foundries — a step that indicates the project has moved beyond internal research into active development.
The target deployment is ByteDance’s own servers and data centers, supporting internal operations as the company prepares a massive rollout of agent-based products including its Coze platform. As EastFrontier has reported, ByteDance doubled down on AI infrastructure with a $5 billion data center push in 2025, and Qualcomm struck an AI chip deal with ByteDance for data centers. The custom CPU project adds a third layer to ByteDance’s chip strategy: alongside GPU procurement and third-party AI chip partnerships, the company is now building its own general-purpose processing capability.
The RISC-V track is particularly notable in the context of the US-China technology competition. As EastFrontier has reported, China’s domestic chipmakers have seized 41% of the local AI market, and the drive to reduce dependence on US-controlled chip architectures is a consistent theme across China’s technology sector. A ByteDance CPU built on RISC-V could potentially be deployable without any licensing exposure to US intellectual property, a strategic hedge that goes beyond cost optimization into technology sovereignty.