ByteDance builds custom CPUs on Arm and RISC-V to power AI infrastructure amid chip shortages

ByteDance Developing Custom CPU Chips to Break Free from Supply Constraints

TikTok parent ByteDance is developing its own custom CPUs to power its expanding AI infrastructure, joining a growing cohort of tech giants designing in-house silicon to escape escalating costs and supply bottlenecks2

2

. The Beijing-based company is pursuing two parallel chip architecture tracks—one based on Arm and RISC-V architectures—as it evaluates which design best suits its long-term data center requirements3

3

. This dual-track approach allows ByteDance to test options before committing to costly, large-scale manufacturing runs, a common hedge among technology giants2

2

.

The project remains at an early stage, with ByteDance approaching several external partners to assist with design work and secure manufacturing capacity at foundries2

2

. The company plans to deploy its proprietary CPU in its own servers and AI data centers to support internal operations, particularly as it prepares a massive rollout of agent-based products including its Coze platform5

5

. ByteDance also operates China's AI chatbot app Doubao and maintains a portfolio of AI models that will benefit from this custom silicon1

1

.

Rising Chip Prices and Supply Shortages Drive In-House Silicon Push

The decision to support AI rollout with custom processors stems from harsh economic realities. ByteDance currently sources CPUs from Intel and AMD, but both suppliers have raised prices significantly, with quarter-over-quarter increases ranging from 10% to as much as 35% in recent months2

2

. Intel has warned Chinese customers of server CPU delivery lead times of up to six months, while AMD CEO Lisa Su cautioned that the global CPU market is "tight," with demand outpacing forecasts and supply constraints expected to persist5

5

.

These rising chip prices and supply shortages have prompted ByteDance to accelerate its push for in-house alternatives2

2

. Intel reported that demand for its CPU from AI firms was so strong in the first quarter that it sold even chips it had originally written off5

5

. With ByteDance's 2026 AI infrastructure budget reportedly growing 25% to around 200 billion yuan ($29.4 billion), the procurement gap from these price increases has become material at group-level economics3

3

.

Industry Shift Toward AI Model Inference Creates CPU Demand

ByteDance's move to reduce reliance on US chipmakers underscores the industry's rapid shift toward "inference," where AI models are deployed to perform agentic tasks that demand more from CPUs, working in tandem with graphics chips made by Nvidia that have dominated the AI boom2

2

. The company's new chip is inspired by Groq's "language processing units," optimized for inference tasks—running AI models instead of training them—as inference-heavy agentic AI quickly becomes the new normal1

1

.

This shift has created a shortage of CPUs in recent months, and global hyperscalers including Alphabet's Google, Amazon, and Microsoft are also developing their own custom CPUs to reduce costs and tailor performance to their specific workloads5

5

. The move has also helped major CPU makers Intel and AMD emerge as leading challengers to Nvidia's AI dominance, though Nvidia is expanding beyond GPUs into the CPU market with its new "Vera" central processors, targeting access to a new $200 billion market2

2

.

Geopolitical Tensions and Strategic Autonomy Shape Chip Strategy

The custom CPU initiative takes place against a backdrop of geopolitical tensions that influence ByteDance's chip strategy. China's government has banned the purchase of Nvidia H200 Blackwell chips after the Trump administration backtracked on technological export controls1

1

. The RISC-V architecture is increasingly favored inside China because it sidesteps the licensing-and-export-control exposure that comes with Arm's UK-headquartered, SoftBank-owned intellectual property3

3

. Chinese policymakers have explicitly endorsed RISC-V as a strategic-autonomy alternative, with Beijing hardening its broader chip-sovereignty posture through 20263

3

.

ByteDance has been instructed by Beijing's National Development and Reform Commission to reject US-origin capital in funding rounds without clearance, and travel restrictions on senior AI talent that expanded across the private sector apply to ByteDance staff3

3

. This isn't ByteDance's first venture into custom silicon—the company started designing its own SeedChip AI accelerator with TSMC back in 2024, expected to tape out and be mass-produced this year1

1

. Additionally, ByteDance is partnering with Chinese startup InnoStar Semiconductor for memory technology, potentially eliminating the need to acquire rare and expensive HBM chips from Samsung1

1

. The expectation is that ByteDance will use hybrid architectures for its servers initially, but over time could significantly reduce dependence on US chipmakers1

1

.