After initially encouraging heavy use of generative AI tools, Chinese tech firms are now rationing AI tokens—the units needed to process AI tasks—imposing monthly and annual quotas to rein in significant cost increases driven by widespread adoption and intensive AI workloads.
- ByteDance, Alibaba, Baidu, and Tencent set new AI token usage limits for employees
- Quotas reflect rising costs from complex AI workflows and soaring demand
- Companies balance cost control with ensuring employee access to AI resources
What happened
Major Chinese technology companies have begun implementing token quotas to restrict employee consumption of AI computational power. Earlier in the AI adoption phase, workers were encouraged to use AI tools extensively, with high token usage viewed as a marker of productivity. Now, firms like ByteDance, Alibaba, Baidu, and Tencent are placing ceilings on token allowances, reimbursing partially or requiring employees to track usage carefully.
For example, ByteDance requires employees using closed-source AI models to bear part of the external tool costs, reimbursing up to US$1,000 annually for technical staff but limiting non-technical reimbursements to about US$300. Alibaba allocates monthly credits for internal platforms, capping external AI tool reimbursements at US$200. Baidu offers a base allowance of roughly US$223 monthly, with possible top-ups. Tencent shifted from fixed annual tokens to pooled departmental resources managed by supervisors.
Why it matters
These new token management policies reflect a broader challenge within China’s tech sector: balancing rapid AI adoption with sustainable operational costs. The increasing complexity of AI workflows is driving token consumption dramatically higher. Research from Goldman Sachs projects enterprise and consumer AI token use to increase 24-fold from 2026 to 2030, reaching unprecedented volume levels.
Despite anticipated reductions in per-token processing costs by 60 to 70 percent annually, affordable computation invites increased usage, potentially compounding expenses. By rationing tokens, companies aim to prevent wasteful overuse—sometimes called “tokenmaxxing”—and to ensure spending aligns with genuine workload demands rather than usage inflation.
What to watch next
The evolution of token management strategies at Chinese tech firms will be key to controlling AI-driven expenditures without stifling innovation or productivity. Observers should monitor adjustments in quota levels, reimbursement policies, and how companies handle token allocation requests amid growing AI adoption in coding, content generation, and complex autonomous workflows.
Another critical factor will be whether firms develop internal AI platforms and tools to better optimize token use and performance efficiency. Market players’ responses to tokens as a policy tool also could influence global operational practices for managing the cost of generative AI technology at scale.