Chinese AI startup DeepSeek has found itself in the spotlight for two very different reasons. On Monday, the company revealed plans to temporarily limit new user registrations following a cyberattack, mere hours after its AI assistant became the top-rated free app on Apple’s U.S. App Store. The sudden surge in downloads tested the firm’s infrastructure, leading to a wave of login errors and website outages—its longest downtime in three months.
What can’t you sign up DeepSeek?
Behind the scenes, the spotlight remains on the DeepSeek-V3 model powering this viral AI assistant. DeepSeek positions V3 as a lean and resource-efficient competitor to incumbent industry heavyweights, claiming it rivals closed-source models in performance at a fraction of the cost. While the startup has reported using Nvidia’s H800 chips—viewed as less cutting-edge than those Washington hopes to keep out of China—it insists its total training expense stayed under $6 million. This assertion has prompted skepticism among U.S. tech executives, who question the real impact of export controls and wonder how a relatively unknown Hangzhou-based team managed to achieve high-tier AI results so rapidly.
Someone stop DeepSeek: Meet Janus-Pro-7B, another free AI model
Observers also note the ripple effects of DeepSeek’s swift rise on tech markets. Shares of U.S. chipmaker Nvidia and other major technology stocks plummeted Monday, reflecting broader uncertainty about global AI competition and the near-daily advances in model development. By contrast, DeepSeek’s success suggests that China’s AI ecosystem could be maturing faster than many expected—even amidst ongoing trade and export restrictions.
Little is known about DeepSeek beyond its 2023 founding and low-profile origins, but its abrupt upswing highlights a possible shift in the industry’s balance of power. With user interest at fever pitch and questions swirling around the efficacy of export controls, all eyes are now on how DeepSeek will handle its mounting security and infrastructure challenges
