Chinese startup DeepSeek's AI Assistant skyrocketed to the top of the Apple App Store chart while sinking the stock market.

Getty Images

Simone Del Rosario Business correspondent

Business

What is DeepSeek? Chinese AI startup dethrones ChatGPT on App Store

Monday

Listen

Simone Del Rosario Business correspondent

By Simone Del Rosario (Business correspondent), Brent Jabbour (Senior Producer), Emma Stoltzfus (Video Editor)

Chinese artificial intelligence startup DeepSeek’s AI Assistant dethroned ChatGPT to take the No. 1 spot on Apple App Store’s free app chart. The AI disruptor also rocked the stock market Monday, Jan. 27, while facing “large-scale malicious attacks” that forced it to limit users.

Bias Distribution Powered by Ground News™

DeepSeek’s rise sunk Nvidia and other tech companies’ stock prices. Nvidia’s share price fell nearly 17% on Monday while the tech-heavy NASDAQ dropped 3%.

Download the SAN app today to stay up-to-date with Unbiased. Straight Facts™.

Point phone camera here

“I’m going by the papers they’re publishing, and it’s very impressive,” Microsoft CEO Satya Nadella told CNBC at the World Economic Forum in Davos, Switzerland. “So I think we should take the development out of China very, very seriously.”

What is DeepSeek?

Hedge fund manager Liang Wenfeng founded DeepSeek in 2023. The nascent AI lab gained a ton of attention this month after releasing its R1 model to rival OpenAI’s GPT o1 model. It’s already nearly as good as OpenAI, according to the Artificial Analysis Quality Index.

“What we found is that Deep Seek, which is the leading Chinese AI lab, their model is actually the top performing, or roughly on par with the best American model,” Scale AI CEO Alexandr Wang told CNBC last week.

DeepSeek’s R1 is open-source, meaning any AI developer can use it, modify it and enhance it. The company said it can learn and improve without human involvement. Users say it excels at chain-of-thought reasoning, a process involving breaking down a complicated task into logical steps.

Why did tech stocks dive?

U.S. tech companies have been pouring billions of dollars into AI development, but DeepSeek is reportedly breaking the internet on a budget.

According to the DeepSeek-V3 technical report released in December 2024, it took two months and less than $6 million to train this model using Nvidia’s H800 chips, which are modified to be exported to China.

However, not everyone believes exactly what they are reading in the technical report. Wang, whose company provides training data to AI companies, said DeepSeek’s parent company has around 50,000 Nvidia H100 chips, which are subject to export controls imposed by the U.S.

Back in 2022, the U.S. placed export controls on Nvidia to stop the company from selling its A100 and H100 chips to Chinese companies. These chips are used to develop AI models.

OpenAI CEO Sam Altman also questioned DeepSeek’s breakthroughs the day after the research paper was released.

it is (relatively) easy to copy something that you know works.

it is extremely hard to do something new, risky, and difficult when you don't know if it will work.

individual researchers rightly get a lot of glory for that when they do it! it's the coolest thing in the world.
— Sam Altman (@sama) December 27, 2024

“It is (relatively) easy to copy something that you know works,” he wrote in a post on X. “It is extremely hard to do something new, risky, and difficult when you don’t know if it will work. Individual researchers rightly get a lot of glory for that when they do it! It’s the coolest thing in the world.”

It’s unclear what type of future DeepSeek will have with export controls in place. But President Donald Trump is working with tech companies to boost AI in the U.S. He recently announced the $500 billion Stargate Initiative, a private sector deal with OpenAI, Softbank and Oracle. It’s expected to be the largest AI infrastructure project in history.

Tags: Artificial Intelligence, Chatbot, ChatGPT, China, DeepSeek, Donald Trump, Liang Wenfeng, Microsoft, Nasdaq, NVIDIA, Satya Nadella, Stargate, Stock Market

Simone Del Rosario:

It’s been a wild day for Chinese AI startup DeepSeek. On Monday, its AI Assistant went from dethroning ChatGPT on top of the Apple App Store chart to facing “large-scale malicious attacks” that forced it to limit users.

The AI disruptor is also rocking the stock market. Nvidia’s share price was down 15% midday Monday while the tech-heavy NASDAQ dropped 3%.

So what’s the story behind this Chinese newcomer?

SATYA NADELLA:

I’m going by the papers they’re publishing, and it’s very impressive. I mean, to see the deep seek new model, it’s super impressive.

So I think we should take the development out of China very, very seriously.

Simone Del Rosario:

Hedge fund manager Liang Wenfeng founded DeepSeek in 2023. The scrappy AI lab gained a ton of attention this month after releasing its R1 model to rival OpenAI’s o1 model. It already barely trails OpenAI, according to the Artificial Analysis Quality Index.

ALEXANDR WANG:

What we found is that Deep Seek, which is the leading Chinese AI Lab, their model is actually the top performing, or roughly on par with the best American models

Simone Del Rosario:

So what is setting R1 apart? For one, it’s open-source, meaning any AI developer can use it, modify it, and enhance it.

Second, DeepSeek says it can learn and improve on its own without human involvement.

Users say it excels at chain-of-thought reasoning, where you break down a complicated task into logical steps.

Why is it sinking tech stocks? U.S. tech companies have been pouring billions into AI development, but are they overspending?

According to the DeepSeek-V3 technical report released last month (Dec. 26), it took just two months and less than $6 million to train this model using Nvidia’s H800 chips, which are modified to be exported to China.

But not everyone believes the report.

ALEXANDR WANG: the Chinese labs, they have more h1 hundreds, than than people think, you know, the
Andrew Ross sorkin: and these are the highest powered Nvidia chips that they were not supposed to have
ALEXANDR WANG: Yes, my understanding is that is that deepseek has about 50,000 h1 hundreds, which they can’t talk about, obviously, because it is against the export controls that the United States has put in place. And I think it is true that, you know, I think they have more chips than other people expect, but also go on a go forward basis, they are going to be limited by the chip controls and the export controls that we have in place.

Simone Del Rosario:

Back in 2022, the U.S. placed export controls on Nvidia to stop the company from selling its A100 and H100 chips to Chinese companies. These chips are used to develop AI models.

I’m trying not to get too technical here. So the bottom line is that the H100 is a better, more sophisticated chip than the H800. The H100 is not allowed to go to China, yet Alexandr Wang says DeepSeek has them.

A day after DeepSeek released its research paper, OpenAI’s Sam Altman seemed to throw cold water on its breakthroughs.

At the time, he posted to X:

“It is (relatively) easy to copy something that you know works. it is extremely hard to do something new, risky, and difficult when you don’t know if it will work. individual researchers rightly get a lot of glory for that when they do it! it’s the coolest thing in the world.”

It’s unclear what type of future DeepSeek will have with export controls in place. What we know is its founder wants to create human-level AI.

President Donald Trump is working with tech companies to boost AI in the U.S. He recently announced the $500 billion Stargate Initiative, a private sector deal with OpenAI, Softbank and Oracle. It’s expected to be the largest AI infrastructure project in history, but AI history is moving fast.

For Straight Arrow News, I’m Simone Del Rosario.

To keep up with the many changes in the world of AI, download the Straight Arrow News app and enable notifications.