Key Points:
- ByteDance is developing its own custom central processing units (CPUs) to support its rapidly expanding artificial intelligence servers.
- The push for proprietary silicon aims to bypass severe global CPU shortages and double-digit price hikes by Intel and AMD.
- The custom CPUs will optimize “inference” workloads, supporting the rollout of autonomous AI agent platforms like “Coze.”
- The initiative follows reports on the company’s “SeedChip” project, which aims to produce up to 350,000 in-house AI processors.
Chinese technology giant ByteDance is developing its own custom central processing units (CPUs) to support its rapidly expanding artificial intelligence infrastructure. In a report published on Thursday, May 28, 2026, industry sources revealed that the parent company of short-video platform TikTok took this step to bypass surging chip prices and prolonged supply shortages. The move marks a major milestone in ByteDance’s multi-billion-dollar push to establish total control over its global data center hardware stack.
The push for proprietary silicon highlights a rapid, industry-wide shift from initial AI model training to the active “inference” phase. During inference, AI models run in production to perform complex, autonomous “agentic” tasks. These autonomous workloads demand significantly more processing power from CPUs, which must work in tandem with the graphics processing units (GPUs) that have dominated the initial AI boom. This shift has created a severe global shortage of high-performance server CPUs over the past several months.
Currently, ByteDance sources its server CPUs from industry leaders Intel and Advanced Micro Devices (AMD). However, both manufacturers have raised their prices significantly in recent months, with quarter-over-quarter price hikes ranging from 10% to 35% due to sustained demand and rising material costs. AMD Chief Executive Officer Lisa Su recently warned that the global CPU market remains tight, with demand consistently outpacing forecasts and supply constraints expected to persist.
Beyond price hikes, long delivery lead times have further hampered ByteDance’s infrastructure expansion. Earlier this year, Intel warned its Chinese enterprise customers that delivery lead times for advanced server CPUs had stretched to up to six months. By developing its own custom processors, ByteDance aims to insulate its servers from these supply bottlenecks, ensuring it can scale its computing infrastructure without relying on external partners’ production schedules.
ByteDance plans to deploy its proprietary CPUs directly within its own servers and data centers to support internal operations and high-growth services. Most notably, the custom silicon will power a massive, global rollout of the company’s autonomous agent-based products, including its popular “Coze” platform. Coze allows developers to build sophisticated AI bots and automated workflows that require highly optimized, low-latency CPU processing to manage thousands of simultaneous user requests.
The CPU initiative is part of a broader, highly ambitious semiconductor strategy. In February 2026, reports revealed that ByteDance was developing its own custom AI inference chip under the internal codename “SeedChip.” The company has held discussions with Samsung Electronics to manufacture these prototype chips using Samsung’s advanced foundry services. The company plans to produce at least 100,000 units of this custom inference processor this year, with a progressive ramp-up to as many as 350,000 units.
An enormous financial budget backs this massive hardware push. To support its global AI ambitions, ByteDance plans to spend over 160 billion yuan (approximately $22 billion) on AI-related procurement and R&D in 2026 alone. While more than half of this budget remains allocated to purchasing Nvidia’s high-end H200 graphics processors and upcoming Blackwell systems, the rest will fund its in-house semiconductor efforts, including its 5-nanometer custom ASIC project developed in partnership with U.S. chip designer Broadcom.
ByteDance’s push for in-house silicon also serves a vital geopolitical purpose. Under strict U.S. export controls, Chinese tech companies cannot purchase the world’s most advanced AI and supercomputing processors. This regulatory barrier has forced major Chinese tech firms such as Alibaba, Baidu, and Tencent to develop custom processors to protect their technological sovereignty. By building its own CPUs and ASICs, ByteDance is ensuring that its global platforms can continue to operate and evolve without facing sudden regulatory shutdowns.
As the global artificial intelligence boom enters its highly specialized, inference-driven phase, the tech companies that control both software and hardware will likely emerge as the ultimate winners. By joining global hyperscale peers like Google, Amazon, and Microsoft in developing custom CPUs, ByteDance is transforming from a platform company into a vertically integrated hardware powerhouse. If successful, this strategic pivot will allow the company to bypass the supply constraints of Intel and AMD, driving down its long-term operational costs and securing its digital sovereignty for the next decade.
