The country that masters AI will dominate the world economically, politically and militarily. Since ChatGPT dropped, the U.S. seemed untouchable. Most of us Americans assumed we were a couple of years ahead of China in terms of AI, but the game has changed — and fast.
The latest version of DeepSeek AI, an open-source model out of China, is so good, it tanked U.S. tech stock prices (Nvidia lost $593 billion in value!), shot to No. 1 in the Apple App Store overnight and now has the entire world wondering, “If this is what China is showing us, what’s next?”
Move over, OpenAI
DeepSeek was founded in May 2023 in Zhejiang, China. Its first models were nothing to write home about; the latest release, DeepSeek-V3, is another story.
It was developed in just 55 days, trained on 671 billion parameters and performs as well as (or better than) Meta’s Llama, OpenAI’s GPT‑4o and Anthropic’s Claude Sonnet 3.5 in math, coding and reasoning. Let that sink in. It took China just two months to beat the American giants.
The money is where it gets really interesting. OpenAI spent $5 billion on its model in just one year. Google shelled out $50 billion on AI development in 2024. Microsoft has invested $13 billion into AI partnerships.
What about DeepSeek? They spent $5.6 million. It’s a cheap Chinese knockoff.
How’d they do it?
China put together a group of young, ambitious, super-smart engineers and researchers who worked under strict limitations. The official story is they couldn’t use Nvidia’s top-tier H100 chips because of U.S. export restrictions. Instead, they worked with less powerful H800 chips.
Rumors suggest China started with over 10,000 super-powered H100 Nvidia AI chips purchased before the Biden administration’s sanctions kicked in. There are also whispers they stole OpenAI’s code as the foundation for DeepSeek‑V3.
But here’s the thing: Even if they took someone else’s code, it doesn’t matter anymore. DeepSeek runs efficiently on far fewer chips, uses less electricity and is cheaper to operate than its American counterparts.
The real game-changer is right here
Most U.S. companies treat AI like a cash cow. For advanced features or running complex tasks, you pay in tokens. With ChatGPT, 1 million tokens cost $4.40. On DeepSeek? Just $0.10.
Powerful AI just became accessible to poorer nations, businesses and individuals at a fraction of the cost. That could mean incredible things for health care, education, agriculture, transportation, finance and the environment.
But it also puts cutting-edge AI in the hands of North Korea, which could use it to develop new cyberattacks. What about state-sponsored hackers looking for creative ways to steal our intellectual property? Drug cartels and scammers could tap into it to create smarter, harder-to-detect schemes. Pandora’s box is open.
Curious about trying DeepSeek AI?
I was, and I’ll tell you, it really is impressive. It’s fast, it’s smart, and its answers beat ChatGPT in nearly every prompt I’ve tried.
Yesterday, DeepSeek got hit by a cyberattack, and new signups were limited to Chinese phone numbers. (I hope it was the U.S. government, really.)
Once it opens back up, you’ll need to be careful if you decide to make an account. This is how I use it:
- Create a separate email address just for using DeepSeek. Do not sign in using your Google account or any other.
- Use it on a separate device, like an old phone or computer you don’t use for anything else.
- Treat everything like public information. Don’t enter anything personal or sensitive to your business.
Of course, China isn’t just exporting the technology; DeepSeek’s views are all state-approved. That means, according to their AI, there’s no info on the 1989 Tiananmen Square incident, and it won’t tell you about internment camps, protests in Hong Kong or anything else the Chinese government doesn’t want you to know.
Anytime you’re using a tool developed in China, assume your data is collected, stored and shipped right back to its government.
