EXPLAINER: What to know about DeepSeek — ChatGPT’s Chinese rival

DeepSeek is a large language model (LLM) developed in China, similar to OpenAI’s ChatGPT.

The chatbot recently made headlines when it surpassed ChatGPT as the most downloaded free app on Apple’s app store.

But beyond its growing popularity, DeepSeek has also sparked global discussions about its capabilities, cost-efficiency, and potential geopolitical implications.

WHO CREATED DEEPSEEK?

DeepSeek was founded in 2023 by Liang Wenfeng, a 40-year-old billionaire, and is headquartered in Hangzhou, Zhejiang, China.

Unlike many artificial intelligence (AI) startups backed by the government or some corporate funding, DeepSeek says it is independent and is solely funded by High-Flyer, an $8 billion hedge fund also founded by Wenfeng.

HOW DOES DEEPSEEK COMPARE TO CHATGPT?

DeepSeek is built on an advanced reasoning model similar to OpenAI’s o1 model, which was released in September 2024. This means that, like OpenAI’s model, DeepSeek does not just generate responses; it “thinks” before answering and produces a structured chain of thought to tackle complex queries.

Access to OpenAI’s o1 model comes at a premium of $200 per month. However, DeepSeek provides similar functions for free.

The Chinese chatbot has also been praised for its efficient design, as most AI systems need a lot of computing power, while DeepSeek uses memory more wisely, keeping costs low while still delivering a brilliant performance.

HOW DID DEEPSEEK TRAIN ITS AI DESPITE SANCTIONS?

The US imposed strict export controls in 2022, limiting China’s access to advanced AI chips like Nvidia’s high-end semiconductors, H100, which are important for training large-scale AI models.

Yet, DeepSeek claimed it successfully trained its model using Nvidia’s H800 chips for just $6 million, which is about a fraction of the billions spent by US companies like OpenAI.

DeepSeek’s claim has raised questions about whether the company truly relied on lower-capacity hardware or secretly obtained banned H100 chips, with several leaders in the tech community expressing doubts about DeepSeek’s cost-efficiency claims.

Alexandr Wang, the chief executive officer (CEO) of Scale AI, suggested that DeepSeek may have secretly acquired more advanced H100 chips than it admitted.

In response, Elon Musk, owner of X, agreed with Wang’s skepticism by simply posting “Obviously” on his platform.

Ted Mortonson, a tech strategist at Baird, also questioned whether DeepSeek truly built such an advanced model using only the reduced-capacity H800 chips.

WHAT’S DEEPSEEK’S IMPACT ON GLOBAL MARKETS?

DeepSeek’s high-performance ratings rattled global tech stocks on Monday, triggering a sharp selloff.

Nvidia’s stock dropped by 18 percent, erasing over $589 billion from the company’s market valuation, while the S&P 500 fell by 2.3 percent and the Nasdaq 100 plummeted by 3.6 percent.

The market turbulence left analysts wondering if DeepSeek’s success signals a shift in AI dominance from the US to China.

DOES DEEPSEEK CENSOR CERTAIN TOPICS?

DeepSeek appears to avoid politically sensitive subjects.

Users reported that when they asked about topics like the Tiananmen Square protests or Taiwan’s independence, the AI either refused to respond or gave a vague reply, such as “Sorry, that’s beyond my current scope. Let’s talk about something else”.

TheCable conducted a test and got the same results, with DeepSeek sidestepping the questions.

US RESPONSE AND NATIONAL SECURITY CONCERNS

DeepSeek’s rise has triggered national security concerns in the United States, mirroring debates surrounding TikTok.

Donald Trump, the US president, acknowledged China’s AI advancements but warned that American tech companies should view DeepSeek’s success as a “wake-up call”.

John Moolenaar, chairman of the house select committee on the Chinese communist party, called for stronger restrictions to protect the US technological edge.

“The U.S. cannot allow Chinese Communist Party models such as DeepSeek to risk our national security and leverage our technology to advance their AI ambitions,” he said.

In response to China’s rapid AI advancements, Trump recently announced a $500 billion investment in AI data centres in America.

The initiative called the Stargate Project is a joint venture between Oracle, Japan’s SoftBank, and OpenAI.

It is aimed at securing America as a global leader in AI innovation and creating thousands of jobs as well as bolstering national security.

DEEPSEEK AS A MILESTONE IN CHINA’S AI STRATEGY

DeepSeek’s emergence aligns with China’s long-term AI ambitions.

In 2017, China’s State Council launched the “New Generation Artificial Intelligence Development Plan,” setting a goal to establish the country as a global AI leader by 2030.

DeepSeek’s rapid rise underscores China’s progress toward that vision, proving itself as a serious competitor on the global AI stage.

Even Sam Altman, OpenAI’s CEO, has acknowledged DeepSeek’s cost-efficient performance, calling it an impressive model. However, he remains confident that OpenAI will continue to lead with more advanced AI systems.

DeepSeek’s disruptive approach offers a blueprint for AI development that prioritises efficiency over sheer computational power, challenging American companies to rethink their strategies.

In an interview, DeepSeek’s CEO said the company’s mission “isn’t for quick profits but advancing the technological frontier to drive ecosystem growth”.