Terug naar Nieuws
IndustrieAI Understanding-briefing

DeepSeek legt dagelijkse chatlimieten op te midden van de AI-chiptekort in China

Het Chinese AI-lab DeepSeek heeft dagelijkse conversatiebeperkingen ingevoerd en de invoerlengte voor zijn gratis chatbot verkort, met verwijzing naar een tekort aan rekenkracht terwijl de Amerikaanse exportcontroles de toegang tot geavanceerde Nvidia-chips beperken.

4 min readRead the linked source
Source-provided image accompanying DeepSeek imposes daily chat limits amid China’s AI chip shortage
BronreferentieBron opgenomen
Uitgever
t.co
Brontype
Gekoppelde bron: de status van de primaire bron is niet vastgesteld.
ContextBegrijp dit in 60 seconden

Wat is er gebeurd

DeepSeek, China’s leading AI developer, has begun enforcing daily conversation quotas, shortening maximum input length to roughly 40,000 characters, and ending support for inactive chats older than seven days. The changes were reported by multiple users through Chinese financial media on Monday, though DeepSeek has not issued an official statement. Earlier in May, the company also limited how often users could regenerate or edit responses, a move interpreted as a response to limited computing capacity.

According to user reports circulated by Chinese financial outlets, DeepSeek’s consumer chatbot now displays notices that long‑running conversations inactive for more than seven days will be terminated. The service also caps daily usage, and users hitting the limit receive a “service congested” message.

The maximum length of a single input has been reduced to about 40,000 characters, down from previously unrestricted limits. No paid tier exists that would allow users to bypass these constraints, marking a shift from prior reliance on slowdowns and busy‑server messages during peak demand.

The daily quota is a departure from DeepSeek’s earlier practice of managing load through performance throttling rather than explicit caps. The company has not publicly commented on the changes, leaving the duration and scope of the limits uncertain.

Brongegevens: t.co

Waarom het ertoe doet

The restrictions provide a rare, public glimpse into the computing power crunch facing Chinese AI firms after U.S. export controls barred direct access to Nvidia’s most advanced chips. Developers have relied on smuggling and remote access to Southeast Asian data centers, but the U.S. Commerce Department is reportedly drafting rules to further restrict chip exports and remote access to Malaysia, Indonesia and Singapore. If enacted, these measures could intensify pressure on Chinese AI budgets, boost domestic chipmakers, and slow the growth of AI services that depend on high‑end hardware. For investors, the limits signal unmet demand in China’s AI market and may affect expectations for Nvidia’s upcoming earnings.

U.S. export controls, implemented to limit China’s access to Nvidia’s latest AI chips, have forced Chinese firms to seek alternative supply routes, including illicit smuggling and remote access to data centers in Malaysia, Indonesia and Singapore. The Wall Street Journal reports that the Commerce Department is drafting rules to further restrict such remote access, potentially cutting off a critical lifeline for Chinese AI developers.

The visible throttling at DeepSeek underscores how these controls are translating into real‑world capacity constraints, affecting user experience and potentially slowing the rollout of AI‑driven products and services in China.

Investors and analysts will likely interpret the quota as a signal of unmet demand, influencing expectations for Nvidia’s upcoming earnings and for the broader market share of domestic Chinese chip manufacturers who could benefit from increased demand for locally produced AI hardware.

Interactive Mechanism

Interactief mechanisme: hoe het eigenlijk werkt

Ontdek interactief de onderliggende technologie achter deze ontwikkeling.

Model Parameter Size:8B Parameters
VRAM Required5.5 GBGPU memory footprint
Target HardwareMacBook / Single GPUDeployment tier
Privacy100% Air-GappedLocal device capability
Core takeaway: Small, quantized models (3B–8B) now run directly inside smartphones and laptops with complete data privacy, while mammoth 400B+ models remain the domain of datacenter clusters.
Interactieve conceptcheck+10 Points
AI Models Explained Quiz

Which component of an AI application is the machine-learning model itself?

Wat je nu moet bekijken

Watch for official confirmation from DeepSeek on whether the limits are temporary or permanent, and for any regulatory announcements from the U.S. Commerce Department on new export‑control rules targeting Southeast Asian chip hubs. Market reactions to Nvidia’s earnings and any shifts in Chinese domestic chip production capacity will also be key indicators of how the chip shortage evolves.

Official confirmation from DeepSeek regarding the permanence of the quotas and any plans for a paid tier or infrastructure upgrades.

Finalization and implementation timeline of the U.S. Commerce Department’s proposed export‑control rules targeting Southeast Asian chip hubs.

Nvidia’s quarterly earnings report, particularly any commentary on Chinese demand and the impact of export restrictions.

Announcements from Chinese domestic chipmakers about new AI‑focused silicon that could alleviate the shortage.

Gerelateerde gidsen en quizzen

Vond je dit nuttig?