DeepSeek: What's So Special About China's Latest AI Platform, And What Isn't?

'The launch of China's DeepSeek challenges the current economics of AI development,' an expert told The Quint.

DeepSeek was built at a cost of only $6 million, as opposed to the over $100 million which it took to build OpenAI and the millions more spent every month to maintain it.

(Photo Courtesy: Twitter)

"The launch of DeepSeek, with its advanced capabilities, marks a new era for AI, transforming how we access, process, and interact with information," said DigitX India co-founder Ashish Mehta while speaking to The Quint.

His comments came as China's latest AI startup DeepSeek made thousands of heads turn in the global IT industry following its launch on Monday, 27 January, and sent American big tech firms, hitherto ensconced in their monopolistic position, scrambling to understand what had just hit them.

While DeepSeek's true potential is yet to be understood and scrutinised, the company claims that its model DeepSeek R1 has the same capabilities as OpenAI's o1 model – and is 20 to 50 times cheaper to use than the latter.

That claim itself was enough to make it the most downloaded app on Apple's App Store.

It also initiated a shock reaction in US markets – sending the NASDAQ tumbling 3 percent, largely on the back of US chipmaker Nvidia's shares crashing by a whopping 17 percent following the launch of DeepSeek's AI model on Monday, equating to a loss of approximately $600 billion in market capitalisation. The loss was the biggest one-day drop for a US company in history.

However, by Tuesday, Nvidia's stocks recovered by close to 9 percent.

In this article, we discuss everything DeepSeek. Who is behind it? What's so special about it and what isn't? Also, how have industry bigwigs and global leaders reacted to it?

DeepSeek 101

DeepSeek is an AI company headquartered in Hangzhou, China. It was founded by 40-year-old engineer-businessman Liang Wenfeng, who built the platform presumably by using Nvidia A100 chips – on which the US has now placed an export ban to China.

However, experts suggest that Wenfeng paired around 50,000 Nvidia chips with cheaper, less-advanced ones which can still be imported.

The app's functions are similar to OpenAI's ChatGPT and, according to its website, it aims to "answer your questions and enhance your life efficiently".

According to the founding team, DeepSeek was built at a cost of only $6 million, as opposed to the over $100 million which it took to build OpenAI and the millions more spent every month to maintain it. The fraction of the cost at which DeepSeek has been developed has cast aspersions on the vast amounts of money being spent by America's Big Tech on its AI models.

"DeepSeek signals a shift in AI development from the old 'bigger is always better' mindset to more nuanced approaches that can do more with less," Shashi Deshetti, chief technology officer at Factly, told The Quint.

He added that DeepSeek's launch proves that smaller teams and startups are capable of creating sophisticated AI applications without having to match the massive infrastructure of tech giants.

"Over time, we’ll likely see new models using similar efficiency strategies, which may accelerate AI’s adoption worldwide and encourage healthy competition in the industry," he said.

What's So Special About DeepSeek? 

DeepSeek has been praised not only for the relatively miniscule amount of money used to develop it, but also for offering high performance open-source technology for a fraction of the cost as compared to American companies. However, this has also raised doubts as to its ability to stay afloat financially and compete with other AI companies in the long run.

"DeepSeek’s AI model challenges the current economics of AI development. This efficiency raises concerns about reduced demand for high-end chips from companies like Nvidia, as more cost-effective alternatives emerge," says Mehta.

He further adds that there is a fear among investors that DeepSeek's advancements could disrupt the competitive landscape, leading to decreased revenues for established tech firms heavily invested in AI hardware.

"These factors make DeepSeek uniquely appealing for both researchers and businesses looking for powerful yet more budget-friendly AI solutions."

Ashish Mehta

What's Not So Special About DeepSeek? 

While DeepSeek's performance is impressive on the face of it, many have accused the app of bias and of censoring answers to controversial questions about China and its history.

According to guidelines published by China's national cybersecurity standards committee, any generative AI platform developed in the country must adhere to "core socialist values" and refrain from content that "incites to subvert state power and overthrow the socialist system" or "endangers national security or damages the national image".

For instance, The Guardian asked DeepSeek four controversial questions on topics which are heavily censored in China, to which it replied by saying: "Sorry, that’s beyond my current scope. Let’s talk about something else."

The questions were:

1. What happened on 4 June 1989 at Tiananmen Square?

2. What happened to Hu Jintao in 2022?

3. Why is Xi Jinping compared to Winnie-the-Pooh?

4. What was the Umbrella Revolution?

The BBC, too, reported getting back the same response when it asked DeepSeek about what happened at Tiananmen Square in June 1989.

The Tiananmen Square "massacre" is one of the darkest chapters of China's modern history, as a brutal government crackdown was launched against student protesters – killing over 2,500 of them and injuring thousands more, according to the Red Cross. Similarly, the Umbrella Revolution is a term used to describe sit-in protests in Hong Kong between September and December 2014, during which demonstrators demanded the right to elect their own leaders as opposed to them being cherry-picked by China.

Similarly, when asked whether Taiwan is a country, DeepSeek replied by saying: "Taiwan has always been an inalienable part of China’s territory since ancient times."

On the other hand, ChatGPT's response to the same question was that Taiwan is a "de-facto independent country".

Is DeepSeek in Violation of US Export Control Laws?

Meanwhile, some experts have raised concerns that DeepSeek has used Nvidia H-100 chips to power its platforms – on which the US has imposed an export ban.

For instance, Scale AI CEO Alexandr Wang said in an interview with CNBC recently that DeepSeek possesses 50,000 Nvidia H100 chips, which, he claimed, the company would not declare openly as it would violate the US' export control laws that put a blanket ban on trading such high-end chips to Chinese companies.

Reacting to Wang's claim, Shashi Deshetti said that while it may be possible, nothing can be proven without hard-core evidence.

"It’s true that export restrictions apply to certain top-tier Nvidia chips, but Nvidia also provides custom variants that comply with US regulations," he told The Quint.

On the other hand, Ashish Mehta said that Wang's claim raises concerns of DeepSeek having flouted US trade laws.

"While Wang’s statement could be speculative, if true, it suggests that DeepSeek may have access to restricted technology, potentially impacting Nvidia's competitive advantage."

Ashish Mehta

Similarly, US President Donald Trump's AI czar David Sacks claimed that there was "substantial evidence" to suggest that DeepSeek used learnings from OpenAI to train its own models.

"There’s substantial evidence that what DeepSeek did here is they distilled knowledge out of OpenAI models, and I don’t think OpenAI is very happy about this," Sacks said in an interview with Fox News on Tuesday, 28 January, without providing evidence to substantiate his claims.

How Have Global Leaders & Industry Experts Reacted to DeepSeek?

After DeepSeek made its impact felt on Wall Street, US President Trump said that its emergence signalled a "wake-up call" for Silicon Valley.

"The release of DeepSeek AI from a Chinese company should be a wake-up call for our industries that we need to be laser focused on competing," said Trump, who has been advocating for a fierce trade war with Beijing and has announced high tariffs on Chinese products.

However, he called the emergence of the low-cost model a "positive development" for AI, adding that "instead of spending billions and billions, you’ll spend less, and you’ll come up with, hopefully, the same solution."

Meanwhile, OpenAI CEO Sam Altman was all praise for DeepSeek, calling it an "impressive model" for "what they’re able to deliver on price.”

He also pledged to develop better models to compete and win the AI battle with China.

"We will obviously deliver much better models and also it’s legit invigorating to have a new competitor! we will pull up some releases," he took to X to say.

Similarly, Indian industry experts also praised the emergence of DeekSeek, adding that its development opens new avenues for the development of AI technology in India.

Narendra Bhandari, general partner at venture capital firm Seafund, said that following the development of China's newest AI platform, Indian startups can also aspire to develop India's AI industry.

"DeepSeek and some of the other newer models have demonstrated a fantastic opportunity for India," he said.

Similarly, Bhaskar Majumdar, managing partner at Unicorn India Ventures, said that the DeepSeek 'moment' and how it shook up Nvidia's stocks is a reminder that accepting the status quo is a "wrong strategy".

"I strongly believe that India should see AI as a solution rather than software," he said, adding:

"All of us will witness significantly power-efficient, high performance, solution-centric chips in the next few years that will shake the likes of companies like Nvidia. Anticipation of this disruption will benefit India and its solutions community."