AI Pioneer DeepSeek Restricts Free Access Amid Severe Hardware Bottlenecks and Surging Demand
KO YONG-CHUL Reporter
korocamia@naver.com | 2026-11-18 08:11:53
Chinese AI innovator DeepSeek has implemented restrictions on its complimentary tier for the very first time in its history. Industry insiders attribute this shift to escalating server infrastructure shortages, compounded by stringent U.S. export controls that restrict smooth access to cutting-edge AI processing units, alongside an unprecedented spike in user traffic.
Reports from local financial outlets indicate that DeepSeek recently deployed capacity caps across its web platforms and mobile applications under the banner of "insufficient computing resources." Under the new policy, single prompt lengths are capped at roughly 40,000 characters, and long-standing conversational threads initiated more than a week ago can no longer be resumed if left idle. Furthermore, daily message caps have been introduced, with several active users hitting their thresholds within a mere two hours of interaction.
With management offering little official communication regarding the changes, customer frustration is mounting. Regular consumers have voiced grievances online, noting that while they are willing to pay for reliable access, the platform currently lacks standard subscription tiers for everyday users—though developer-focused API services remain commercialized.
The emergency measures follow a string of infrastructure failures. Financial tracking data reveals that DeepSeek has logged dozens of service disruptions over the year, including severe outages that temporarily paralyzed web, mobile, and third-party integrated platforms.
The heavy computational strain is largely fueled by the exponential rise of AI agents powered by DeepSeek technology. Unlike traditional chat interfaces, automated agents execute multi-step workflows, query external tools, cross-examine outputs, and continuously parse extensive contextual data. Industry benchmarks suggest that a typical AI agent task consumes roughly four times the computational tokens of standard chatbot queries.
Concurrently, investigative tech reports have surfaced alleging that DeepSeek sought alternative hardware channels to bypass restrictions. Sources claim that entities linked to business executive Alice Huang—specifically cloud provider Century Litong and its parent firm—participated in DeepSeek's recent investment funding rounds. Insiders suggest the startup marketed the firm to stakeholders as a conduit for securing Nvidia silicon via secondary or grey-market channels, utilizing special purpose entities to keep specific corporate backers off public registry records.
[ⓒ Global Economic Times. 무단전재-재배포 금지]
WEEKLY HOT
- 1Navigating Geopolitical Shifts: Seoul and Beijing Must Deepen Dialogue
- 2Beyond the Garden Gate: Why South Korea’s New Special Zones Risk People for Profit
- 3Balancing Joint Operations and Tradition: South Korea Shifts Strategy for the Armed Forces Academy
- 4Controversy Erupts as South Korean Court Acquits Daegu Mayor in Martial Law Vote Obstruction Case
- 5Independence Fighter Choi Bong-jin: From the India-Myanmar Front to the Korean Liberation Army
- 6A New Frontier of Creation Unfolding in the Science City of Daejeon