An Eventful Week for AI Flips the Script --- DeepSeek's rise, Nvidia's decline and a huge bet on OpenAI change the game

"Artificial intelligence has been a fast-moving, high-drama industry since OpenAI launched the viral chatbot ChatGPT a little more than two years ago and awakened the world to the technology's potential.

But nobody has seen a week like the past one: DeepSeek, a disruptive new Chinese AI company, emerged seemingly out of nowhere; the world's most-valuable company lost nearly $600 billion of market value in a day; and it emerged that the developer of ChatGPT was in talks for a fundraising round that would value it at an eye-popping $300 billion.

Here are the most important things we learned from last week's events:

DeepSeek is a coup for China and a new approach to AI.

Most people heard of DeepSeek for the first time in the past week, but tech geeks had it on their radar for months. The Chinese startup -- launched by an engineer who ran a hedge fund -- released a series of AI models nearly as good as the best that Silicon Valley has to offer.

DeepSeek isn't just good. Thanks to some widely praised engineering breakthroughs, it was made at a fraction of the cost and in a fraction of the time of those made by Silicon Valley tech giants.

Those breakthroughs allowed it to be made in China, a country that many experts thought was behind the U.S. because of export restrictions on powerful AI chips. And DeepSeek gave its models away, as open-source code, which helped make them immediately popular among consumers, businesses and developers.

DeepSeek's rise made investors skeptical of Nvidia.

As buzz about DeepSeek's latest model grew and its app shot to the top of the download and performance charts, Wall Street panicked.

If artificial intelligence can be made more simply, investors figured, maybe the world won't need as many pricey chips.

Shares of Nvidia, which was the world's most valuable company just a week earlier, plunged 17% last Monday, as part of a market wipeout that erased some $1 trillion of stock value.

Nvidia, like other tech companies, said that DeepSeek's ascendance would increase overall demand for AI and boost demand for the company's products.

But Nvidia stock recovered only slightly by the end of the week, signaling that investors are still wondering whether Nvidia's power as the computational arms dealer atop the AI boom has been diminished.

Chip limits on China aren't stopping its AI ascendance.

The U.S. in recent years has tightened restrictions on shipments of chips to China, aiming to limit development of the most advanced AI tools. Nvidia responded to the controls by producing less-powerful chips for the Chinese market. DeepSeek has said it used some of those chips to develop previous AI models.

Many in Washington, D.C., are pressuring President Trump to take action to keep the U.S. ahead in an AI arms race. And members of Congress are advocating for a ban on exporting H20 chips to China.

DeepSeek may be succeeding by free-riding on OpenAI's work.

As American technologists began investigating how DeepSeek got so good, the word "distillation" started popping up. It is a technical term for an AI model that becomes intelligent by asking another model questions and learning from its answers.

DeepSeek said it distilled from an open-source model released by Meta Platforms and from one of its own.

But some technologists think it also may have distilled from other proprietary models whose terms of service forbid distillation.

OpenAI said Wednesday it is reviewing indications that DeepSeek did just that to its systems.

If that is the case, it would mean that American investment is benefiting China's new AI star.

DeepSeek didn't respond to requests for comment.

DeepSeek's strategy puts Silicon Valley's at risk.

A growing chorus in Silicon Valley is questioning whether it still makes sense for the biggest AI developers, which are collectively losing billions of dollars a year, to keep spending gobs of cash developing advanced technologies.

Now that DeepSeek has proven how successful distillation can be, investors expect many startups that could never afford to develop complex artificial-intelligence models to ape the Chinese company's approach.

That could be great for consumers who may get more AI at low or even no cost.

But it could dissuade leading AI companies from building the next cutting-edge model, out of fear that their work will quickly be distilled and benefit competitors.

SoftBank is making a huge bet on OpenAI.

OpenAI Chief Executive Sam Altman praised DeepSeek for "impressive work" in a post on X but said that his company is "excited to continue to execute on our research road map."

On Thursday, The Wall Street Journal reported that a powerful partner is standing behind Altman.

SoftBank, the Japanese tech and investment conglomerate led by Masayoshi Son, is in talks to invest $15 billion to $25 billion in OpenAI.

That money would be part of an investment round of up to $40 billion from numerous backers that would value OpenAI as high as $300 billion.

In October, OpenAI was valued at $157 billion. The increase is a loud endorsement by Son of Altman's spend-big-to-dominate strategy.

As the wild week for AI ended, Altman threw one last bomb: In an "ask me anything" session on Reddit on Friday, the CEO said his company should consider giving away its AI models, just as DeepSeek does.” [1]

Oh my Son... What did you do? You have chosen loser again.

1.  An Eventful Week for AI Flips the Script --- DeepSeek's rise, Nvidia's decline and a huge bet on OpenAI change the game. Fritz, Ben; Gryta, Thomas.  Wall Street Journal, Eastern edition; New York, N.Y.. 03 Feb 2025: B4.