HomeNews

DeepSeek Caps Free Chat Quotas Amid China Compute Crunch

Saurabh Anand
Saurabh Anand
•
Nov 17, 2026 10:53 AM
•
0
 min read
Select Emergent as your Preferred news source
DeepSeek Caps Free Chat Quotas Amid China Compute Crunch

💡 TL;DR

  • DeepSeek now limits free users to daily conversation quotas and caps individual prompts at 40,000 Chinese characters.
  • Conversations inactive for over seven days automatically expire due to insufficient computing capacity, the company states.
  • Eight service incidents since October signal mounting pressure from US chip-export controls on advanced Nvidia GPUs.

DeepSeek quietly rolled out hard usage caps on its free chat service over the weekend, limiting users to daily conversation quotas and capping individual prompts at roughly 40,000 Chinese characters. The restrictions, officially launched on November 17, 2026, mark the most visible sign yet that US chip-export controls are constraining China's leading AI labs. The company told users the changes stem from insufficient computing capacity and has not disclosed when normal service will resume.

What Changed for Free Users

Users visiting DeepSeek's web platform or mobile app now encounter three hard limits. Individual prompts cannot exceed approximately 40,000 Chinese characters, conversations idle for more than seven days automatically expire and cannot be resumed, and a daily usage quota blocks further queries once a user hits the threshold. According to TechNode, the company displayed a notice Sunday stating that long conversations inactive for more than a week would no longer be accessible because of compute constraints.

Neither the exact daily quota nor the character cap has been formally announced by DeepSeek. Users discovered the limits through trial and error, and the thresholds appear to vary by account type. The restrictions apply to both the free consumer chat interface and select API tiers, though paid enterprise customers reportedly retain higher or unlimited quotas.

Service Incidents Mount Since October

DeepSeek's status page logged eight separate incidents affecting its chat service or V4 Pro and V4.1 Flash APIs since October 2026. The incidents ranged from brief outages to degraded response times, and the company attributed several to capacity shortages. Weekly flagship-model requests reportedly reached 59.2 trillion tokens during the November 9-15 window, a figure that underscores the scale of demand the lab is managing under constrained hardware allocations.

The timing aligns with tightened US export rules on advanced Nvidia GPUs to China, which took full effect in mid-2026. DeepSeek and rival labs have publicly acknowledged the chip restrictions force them to maximize utilization of existing clusters and delay infrastructure expansion. The company has not said whether it plans to lift the caps or impose stricter limits if demand continues to climb.

Why This Matters for the China AI Ecosystem

DeepSeek's throttling is a bellwether for the broader China AI sector, which relies heavily on Nvidia's H100 and A100 chips for training and inference. The lab's V4 Pro and V4.1 Flash models compete directly with OpenAI and Anthropic offerings in multilingual reasoning and long-context tasks, and free-tier access had been a key differentiator. By capping usage, DeepSeek signals that even well-funded Chinese labs cannot sustain unrestricted free inference at frontier scale under current hardware constraints.

Industry observers expect other China-based AI companies to follow suit. Baidu, Alibaba Cloud, and SenseTime have already introduced tiered pricing and usage caps on their own chat products, and DeepSeek's move validates the shift toward metered access. The changes also accelerate the push for domestic chip alternatives, though analysts note that Chinese GPU startups remain years behind Nvidia in performance and ecosystem maturity.

What This Means

DeepSeek's new quotas and character caps transform its free chat service from an open research preview into a metered utility, mirroring the trajectory of Western AI labs that have steadily moved users toward paid tiers. The restrictions underscore how US chip-export policy is reshaping the competitive landscape, forcing Chinese labs to ration compute and prioritize paying customers over free users. For developers and consumers relying on DeepSeek's APIs, the incident is a reminder that access to frontier AI in China now carries hard supply-side constraints, with no clear timeline for relief.

About the writer

Saurabh Anand Rai is the Head of Product at Emergent, where he leads product strategy and innovation for AI-powered tools that help people build software faster.

Start Building
on Emergent today
Try Emergent