Will DeepSeek AI disrupt OpenAI’s dominance?

Getting your Trinity Audio player ready...

Many people are talking about DeepSeek, the Chinese artificial intelligence (AI) company behind DeepThink R1, the latest ChatGPT competitor. DeepThink R1 is being called just as capable as OpenAI’s GPT-01. However, DeepSeek’s model is open-source, charges users a fraction of what OpenAI bills, and reportedly costs significantly less than OpenAI to get its model up and running.

🚀 DeepSeek-R1 is here!

⚡ Performance on par with OpenAI-o1
📖 Fully open-source model & technical report
🏆 MIT licensed: Distill & commercialize freely!

🌐 Website & API are live now! Try DeepThink at https://t.co/v1TFy7LHNy today!

🐋 1/n pic.twitter.com/7BlpWAPu6y

— DeepSeek (@deepseek_ai) January 20, 2025

To put this into perspective, the DeepSeek team claims they only spent $5.58 million building and training their model. Compare that to OpenAI, which has spent billions to train its models so far. These two elements—cost and open access—are why DeepSeek seems to be at the center of every tech conversation taking place right now and why it’s having a ripple effect on the United States tech market.

DeepSeek’s launch leads to market shake-up

The DeepSeek frenzy started Friday, January 24, and spread like wildfire. By late Sunday, DeepSeek had overtaken ChatGPT in Apple’s (NASDAQ: AAPL) App Store rankings for free apps.

But DeepSeek didn’t just generate initial buzz; it created a domino effect that is currently ongoing. On Monday morning, the S&P 500 (NASDAQ: SPX) opened down 2%, and the Nasdaq opened down 3%. Why? Because the launch of DeepSeek raised questions about the U.S. tech industry. If a foreign company could build something as good as OpenAI’s GPT-o1 for a tiny fraction of the cost, it suggests that American tech companies are overvalued and inefficient.

At the heart of this issue is the difference in costs. OpenAI raised $5 billion in 2024 but spent $8.5 billion training and operating its AI models that same year. DeepSeek, on the other hand, says they got the job done for $5.58 million. That means OpenAI spent many times more than DeepSeek did to get the same job done effectively.

Beyond that, DeepSeek claims they’re running their entire operation on just 10,000 Nvidia (NASDAQ: NVDA) A100 graphics processing units (GPUs), compared to OpenAI, which has access to hundreds of thousands of the superior Nvidia H100 GPUs. This massive difference in costs and spending makes it hard not to wonder if U.S. companies are overspending and raising money at inflated valuations.

Why developers are flocking to DeepSeek

One reason that DeepSeek picked up traction so quickly is because it is open-source. This is much different from the walled garden that is OpenAI, where pretty much all of its advanced, unrestricted features and tools live behind a paywall. This is why DeepSeek is beginning to emerge as a developer favorite; on DeepSeek, developers can access the source code, customize it, fine-tune it, adapt it to their needs, and even run it locally—all without dealing with restrictive licensing or high costs.

On top of that, DeepSeek’s application programming interface (API) pricing is far cheaper than OpenAI’s, which means that developers, especially those on a budget, can build highly capable systems without burning through as much cash as they had to when working through OpenAI’s API.

Critics question DeepSeek’s claims

But not everyone is buying into the DeepSeek hype. Scale AI CEO Alexandr Wang, for example, has said it’s unlikely that DeepSeek spent so little. He thinks they probably have at least 50,000 Nvidia H100 GPUs, meaning their infrastructure costs would be closer to those of smaller U.S. tech companies.

From last week: Scale AI CEO Alexandr Wang says that DeepSeek and other Chinese AI labs have more Nvidia chips than they can talk about because of current US export controls pic.twitter.com/tPzmdEvvik

— Morning Brew ☕️ (@MorningBrew) January 27, 2025

There’s also speculation that DeepSeek is downplaying its costs to dodge U.S. export controls on advanced computing hardware, which is another reason they have an incentive to hide their operation’s true size and cost. Elon Musk and other industry leaders have backed Wang’s theory, which raises more questions about how much of DeepSeek’s story we can actually believe.

DeepSeek’s impact

I don’t think this is as big of a deal as the headlines make it out to be. Yes, DeepSeek’s launch raises important questions about the U.S. tech industry—like whether companies are overvalued or overcharging for their products (the answer to both questions is yes). It also serves as a clear signal in the market that the U.S. isn’t the only dominant nation in building AI systems and that U.S.-based companies will need to evolve and become more efficient to stay competitive.

But for the average user, I don’t think DeepSeek is a game-changer. Most people use generative AI like a glorified search engine, and for that, ChatGPT or any other existing model works just fine. These users probably won’t see an urgent need or a reason to switch to a new model like DeepSeek because the old models work just fine for their primary use case.

On the other hand, developers are a different story. They’ve been very vocal about how much cheaper DeepSeek is, and for a party that is building an app or service and paying per API call, the cheapest, most capable system will be the winner—which is why this loud minority can’t stop talking about DeepSeek.

DeepSeek and the global AI race

The bigger picture here is the global race for AI dominance. DeepSeek’s launch reminds us how close the race between the U.S. and China really is. The U.S. is often seen as the leader in AI, but DeepSeek proves that China is a serious contender. The implications are huge as both countries push to outdo each other in AI development.

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

However, several months ago, Sam Altman, the CEO of OpenAI, made a good point that continues to be true about “new” companies and products: It’s easier to copy what a first mover does than to be the first mover. There’s truth to that—DeepSeek didn’t have to invent everything from scratch; they could build on what companies like OpenAI had already done.

But that doesn’t mean the first mover always wins. Look at Internet browsers. The ones that are still around today weren’t necessarily the first—they were the ones that made their products more accessible and easier to use. DeepSeek may be positioning itself to do the same by offering an open-source, affordable alternative to the walled gardens that are U.S. AI companies.

In order for artificial intelligence (AI) to work right within the law and thrive in the face of growing challenges, it needs to integrate an enterprise blockchain system that ensures data input quality and ownership—allowing it to keep data safe while also guaranteeing the immutability of data. Check out CoinGeek’s coverage on this emerging tech to learn more why Enterprise blockchain will be the backbone of AI.

Watch: Demonstrating the potential of blockchain’s fusion with AI