The AI Race Comes Down to One Thing

China's DeepSeek startup has made big waves in the U.S...

Its artificial-intelligence ("AI") model R1 matches the power of top AI systems, like ChatGPT, at a fraction of the cost.

And that's worrying U.S. investors... China seems to be closing the AI gap faster than anticipated.

Two days after releasing R1, the S&P 500 Information Technology Index plummeted 6.6%. Though America didn't just sit back and watch...

Tech heavyweights Oracle (ORCL) founder Larry Ellison, OpenAI CEO Sam Altman, and Softbank (9984.T) founder Masayoshi Son teamed up with President Donald Trump to unveil the $500 billion Stargate Project.

The goal of Stargate is to help America beat China in the AI infrastructure race. At the same time, DeepSeek is continuing to innovate... despite hefty financial and tech constraints.

Today, we'll outline these two very different approaches to the AI race... and what the U.S. can do to come out on top.

Stargate isn't as glamorous as it sounds...

This project includes a massive investment in data centers. It also promises to create 100,000 U.S. jobs immediately.

Yet the details are sparse. And some feel that Stargate is just too ambitious...

The problem is, AI infrastructure, particularly data centers, doesn't create many jobs in the long run. Once constructed and installed, each data center only needs five to 30 people to run it.

Another big concern is funding... Elon Musk has publicly questioned whether one of the tech bigwigs – Son – can help secure the money for Stargate.

Regular readers are familiar with Son. While he has had a few lucky breaks, his Softbank Vision Fund also lost billions of dollars chasing risky or ill-timed bets... like WeWork and Uber Technologies (UBER). The point is...

Throwing money at the problem doesn't guarantee success...

DeepSeek is a classic bootstrapped startup, meaning it didn't have unlimited funding or cutting-edge technology.

In fact, the company developed its R1 model using old chips... after the U.S. blocked access to Nvidia's (NVDA) most advanced processors.

DeepSeek did all this on a shoestring budget of $6 million. Compare that with OpenAI's ChatGPT, whose latest model cost at least 10 times that amount.

These constraints forced DeepSeek engineers to become extremely resourceful. They developed leaner, more efficient AI models through sheer innovation.

Soon after releasing the R1 model, DeepSeek overtook ChatGPT in downloads on Apple's (AAPL) U.S. App Store.

China is now putting pressure on U.S. tech firms to step up.

We're not saying America is in dire straits... or that it's going to lose the AI race. What we are saying is that...

You don't need big corporations with huge budgets...

In fact, the Stargate Project might be able do more with less... by forcing U.S. companies to innovate.

Obviously, money is an important boost for many tech projects. Yet it's not essential for AI dominance.

The most successful AI companies won't necessarily be the biggest... But they will be the most creative and innovative... the ones that rise to the challenge.

Right now, the U.S. needs to focus more on innovation and less on funding. And the DeepSeek competition might be helping it do just that.

Regards,

Joel Litman
February 13, 2025