The Most Expensive AI Projects in History Just Got Embarrassed by Something That Costs Less Than a Tech CEO's Bathroom Remodel

The world woke up Monday morning to a new epoch—call it the DeepSeek Era of Chinese artificial intelligence. Or not. Here’s everything you should know about DeepSeek-R1.

Investors went to sleep Friday night still believing it was a good idea for American tech companies to commit billions of dollars to refining artificial intelligence. But before the opening bell on Monday, they were thinking less is more when it comes to sinking cash into AI.

What changed was the introduction of DeepSeek-R1, a Chinese large language model that rivals privately held OpenAI’s ChatGPT. DeepSeek doesn’t just mimic ChatGPT and other models—it’s better in some ways and not as good in others.

For instance, it’s less precise than the American offerings but uses dramatically less power and was produced more quickly and at a much lower cost. The AI from the two countries is just different but roughly equal, the tech community says.

Yet in third-party tests of accuracy, DeepSeek’s model outperformed Llama 3.1 from Meta (META), privately held OpenAI’s GPT-4o and privately held Anthropic’s Claude Sonnet 3.5, according to a CNBC report.

Such tests aside, DeepSeek qualified as an overnight sensation. The app quickly became so overcrowded that anyone arriving after the first wave couldn’t get past the landing page for an opportunity try it out. There was a sense America might lose its position of AI dominance and that the world would never be the same.

No warning?

The West was apparently blindsided by the new AI from China. Just a week ago, President Trump joined with the CEOs of OpenAI, Softbank (SFTBY) and Oracle (ORCL) to announce the Stargate Project, which would invest $500 billion to transform AI.

"I think this will be the most important project of this era," OpenAI CEO Sam Altman said at the event announcing the project.

Still, Altman and the other Stargate CEOs have been far from alone in pouring staggering amounts of cash into AI research. The Magnificent Seven tech companies—Alphabet (GOOG), Amazon (AMZN), Apple (AAPL), Microsoft (MSFT), Nvidia (NVDA), Tesla (TSLA) and Meta—have collectively earmarked hundreds of billions of dollars for AI investment.

If the Seven run into sufficiently thorny problems because of their penchant for AI spending, nearly everyone in the markets could suffer. That’s because the Seven had a combined market capitalization of $18 trillion as of Jan. 6, about a third of the value of the S&P 500.

What’s more, other smaller companies produce products and provide services intertwined with the offerings of the Seven. Semtech (SMTC), for example, just staged a turnaround attributable at least in part to its partnership with Nvidia. In another example, Broadcom (AVGO) found a niche by supplying chips to Nvidia.

But the American tech community had become at least a bit complacent about maintaining the nation’s leadership in AI. It may have happened partly because the Biden administration restricted Nvidia and other chip makers from sending their most-advanced AI-related computer chips to China and other countries unfriendly the United States. As recently as Jan. 13, the outgoing president was piling on more restrictions.

But DeepSeek developed its large language model without the benefit of the most-advanced chips, according to most reports. Plus, it came up with its ChatGPT rival on a budget of as little as $6 million, somewhere around as little as 3% of what OpenAI invested in its model.

The rearview mirror

Yet in hindsight, America should have noticed what was brewing on the other side of the world, some observers contend. Last June, experts in the West were warning that China was lagging behind the U.S. in AI by only four months.

Even before DeepSeek suddenly became a factor, American tech moguls were already talking about the need to bring down the cost of developing and distributing AI and praising the potential for innovation that can occur when developers collaborate in an open-source manner.

Besides, anyone could see American AI’s unquenchable thirst for electrical power and reliance on increasingly massive and more numerous data centers didn’t seem sustainable—let alone desirable.

And until the last few days, American tech experts tended to brush off DeepSeek as a startup—if they thought about it at all. True, it was officially launched in 2023, but the seed for the company was planted long before then.

Founder and AI nerd Liang Wenfeng was still a student when he began trading stocks in the aftermath of the 2007-2008 financial crisis. By 2015, he co-founded High Flyer, a hedge fund that’s now one of China’s largest and that provided much of the DeepSeek bankroll.

Besides amassing cash, Liang stockpiled 10,000 Nvidia A-100 chips before the U.S. outlawed sharing the technology, according to a report from 36Kr, a tech-focused Chinese media company and data provider. If that accumulation of high-tech chips occurred, it remains below the radar of most American commentators.

Anyway, the black swan should have been white. But now that investors, tech leaders and perhaps the Trump administration are adjusting to a new reality, what do they foresee for the near future?

New expectations

DeepSeek’s low-cost approach to developing its model indicates throwing money at innovation isn’t the only valid strategy. Analysts will probably question mega investments tomorrow evening when they interview Meta and Microsoft executives after their quarterly earnings announcements.

Another innovation-related concern is that DeepSeek developers have probably developed skills their American counterparts lack. In China, they’ve had to make do with fewer resources and thus learned or invented different takes on computing.

Then there’s the question of how the U.S. government will react to DeepSeek, a concern raised by JJ Kinahan, CEO of IG North America, the parent company of the tastytrade brokerage and publisher of the Luckbox newsletter.

“Is this going to be another TikTok situation where a Chinese company is collecting all this data on people?” Kinahan asked rhetorically during a television interview. “This is something that is going to bother the government significantly.”

So, DeepSeek could become a national security issue, but it’s also competitor to American tech companies. Silicon Valley might have a lot to gain from sanctions or an outright ban of the Chinese AI.

Then, too, some American companies have something to gain from working with DeepSeek, according to Thomas Westwater, a tastylive financial writer and analyst. He provided Apple as an example.

“Traders view Apple as a potential beneficiary of the new AI developments, given the company aims to integrate AI into its iPhone models instead of directly developing the technology,” Westwater said. “This could make DeepSeek or similar AI models easier to run on users’ mobile devices and potentially eliminate the need to connect to servers to run AI programs.”

A new era?

Unlike most AI models, DeepSeek is available as open-source software, allowing outside developers access to its inner workings and thus making it easier to modify to fit a company’s needs.

For the U.S., China and the rest of the world, a new era may have begun last week when a 22-page data summary introducing DeepSeek became available. Whether it’s really a new era and whether it will be named for the Chinese software company remains to be seen.

Ed McKinley is Luckbox editor-in-chief.

For live daily programming, market news and commentary, visit tastylive or the YouTube channels tastylive (for options traders), and tastyliveTrending for stocks, futures, forex & macro.

Trade with a better broker, open a tastytrade account today. tastylive, Inc. and tastytrade, Inc. are separate but affiliated companies.