What’s the Buzz About "DeepSeek"?

DeepSeek AI Breakthrough Shakes Nasdaq: Tech Stocks Tumble by 3%

On January 27 (local U.S. time), the Nasdaq Composite Index experienced a dramatic 3% decline, driven by growing investor focus on the revolutionary advancements of Chinese AI startup DeepSeek.

During an on-site interview at the Nasdaq Stock Exchange, Vice Chairman McCoy remarked that "DeepSeek represents a pivotal component in the next revolution of artificial intelligence."

The U.S. stock market opened with sharp losses, with the tech sector hit particularly hard. Market analysts attributed the downturn to DeepSeek's latest AI breakthroughs, which have challenged the previously unshakable dominance of the U.S. technology industry.

Leading the decline, NVIDIA, a U.S. chip giant, saw its stock plummet by approximately 17% on the same day. Broadcom shares fell 17%, AMD dropped 6%, and Microsoft was down 2%. Furthermore, industries adjacent to artificial intelligence, such as power suppliers, were not spared. United States energy company stocks tumbled 21%, while Vistra's shares nosedived by 29%.

DeepSeek’s innovation has not only disrupted AI development paradigms but also sent shockwaves across global markets, sparking conversations about the shifting balance of technological leadership.

DeepSeek Surpasses ChatGPT: The Most Downloaded Free App in the U.S. Apple App Store

 In a stunning development, DeepSeek has replaced its competitor, OpenAI’s ChatGPT, to become the most downloaded free application in the U.S. Apple App Store. Reports indicate that DeepSeek’s AI model outperforms OpenAI in multiple benchmarks, all while being developed at a cost of less than $6 million.

 DeepSeek, officially known as Hangzhou DeepSeek AI Foundation Technology Research Co., Ltd., was founded on July 17, 2023. This innovative tech company focuses on developing advanced large language models (LLMs) and related cutting-edge technologies.

 In Silicon Valley, DeepSeek has long been regarded as a "mysterious force from the East," earning widespread attention and recognition.

DeepSeek Disrupts AI Again: Open-Sources Revolutionary R1 Model

At the end of 2024, DeepSeek unveiled its next-generation large language model, V3, and made it open-source, marking a milestone in AI accessibility. Benchmark results demonstrated that V3 outperformed several mainstream open-source models, offering significant cost advantages.

On the opening day of the 2025 World Economic Forum Annual Meeting, DeepSeek once again captured global attention by releasing its latest open-source model, R1. According to the company, the R1 model represents a major technical breakthrough, achieving emergent reasoning capabilities purely through deep learning methods. This innovation allows the AI to excel in tasks such as mathematics, coding, and natural language reasoning, with performance comparable to the finalized version of OpenAI's o1 model.

In addition to its technological advancements, R1 continues DeepSeek’s tradition of delivering exceptional cost efficiency, further solidifying its reputation as a leader in the AI field.

DeepSeek’s Models Spark Doubts About U.S. Tech Dominance

Market analysts have described the launch of DeepSeek’s models as a “seismic shock,” raising concerns about the competitiveness of the U.S. technology sector. The breakthrough has prompted investors to question the perceived leadership, substantial investments, and profitability of American tech giants.

Annex Financial’s Chief Economist, Jacobson, commented: "What truly surprises people is the position of companies like NVIDIA. They are considered to have an almost monopolistic hold on the AI ecosystem, particularly in chip production, and are surrounded by what’s perceived as an extraordinarily strong 'moat.' But perhaps this moat is not as impenetrable as we once believed."

DeepSeek Unleashes New AI Models: Janus-Pro and JanusFlow

Following its groundbreaking R1 model that shook Silicon Valley and rattled Wall Street, DeepSeek continued to make waves with another bold move late at night. On January 27, around 10 p.m., DeepSeek’s founder, Liang Wenfeng—an entrepreneur born in the '80s—announced on the social platform X that a new version was on its way.

Just over three hours later, the AI open-source community HuggingFace revealed the release of DeepSeek’s latest series of multi-modal models: Janus-Pro and JanusFlow, featuring parameter sizes ranging from 1 billion to 7 billion.

This latest release places a strong emphasis on text-to-image generation, a capability poised to redefine creative workflows and enhance content creation. By expanding the potential applications of AI across industries, DeepSeek continues to solidify its reputation as a global trailblazer in innovation.

DeepSeek Limits New Registrations Amid Large-Scale Cyberattacks

On January 28, the service status page of DeepSeek's official website revealed that the platform had been subjected to large-scale malicious cyberattacks. To ensure uninterrupted service, DeepSeek has temporarily restricted new registrations to phone numbers with a +86 country code. However, existing users can continue to log in and access services without disruption.

This proactive measure reflects DeepSeek’s commitment to safeguarding its platform while maintaining reliable service for its user base during a challenging period. The company is actively addressing the situation and reinforcing its cybersecurity infrastructure to prevent further incidents.

DeepSeek’s swift response highlights the growing importance of security in the AI space as its influence continues to expand globally.