My Thoughts on Deepseek as an AI Entrepreneur & Hedge Fund Manager

Every now and then, a moment comes along that forces everyone to stop and rethink what they thought they knew. DeepSeek—a relatively unknown AI startup—just delivered one of those moments.

Imagine this: last week, a company most people had never heard of claimed to have built a cutting-edge AI model for just $5.6 million – and they made it open source. Meanwhile, industry giants like Meta, Google, and OpenAI spend hundreds of millions—sometimes billions—on similar efforts. The implications are massive, and investors felt it immediately.

Within hours, Nvidia’s stock price plunged 16.86%, wiping out over $600 billion in market value. Other AI-heavy stocks followed suit. The reaction? Panic, speculation, and a flood of questions:

• Have we been overpaying for AI development?
• Is brute-force scaling the wrong path?
• What does this mean for the balance of power in the AI world?

What’s happening here isn’t just a new startup making waves. It’s a fundamental shift in how AI is built—and it could change the competitive landscape for years to come.

What makes DeepSeek different?

To me, the main point is that, until recently, AI companies have been treating LLMs like an arms race—the more GPUs, the better. That’s part of why LLMs have become so expensive. Meanwhile, DeepSeek took an efficiency-first approach that reduced computational waste and more efficiently used GPU resources.

This type of disruption will become increasingly frequent across all industries. Exponential technologies are accelerating innovation cycles and opening unexpected development paths. While these discontinuous innovations may seem like isolated surprises, they’re becoming a fundamental part of the competitive landscape. The key isn’t to react to each disruption individually - it’s to develop better strategies for operating in this rapidly evolving environment.

In this case, Deepseek’s breakthrough wasn’t about doing more of what its competitors were good at. Instead, they thought of something else they could do to well to compete. Their breakthrough wasn’t about power ... it was about perspective. It is a good reminder and blueprint to study.

Some think the key message is that businesses no longer need to stock up on Nvidia’s best GPUs to have an advantage. Even if that is true, it begs the question: Will the demand for high-end GPUs decline if AI processing needs start shrinking? Asked another way, if other AI companies also use efficiency-first architecture, will Nvidia’s dominance weaken?

From my perspective, the answer is that Nvidia will continue to grow. I’m excited about a world where the companies that can afford to spend the most money don’t necessarily win. Increased accessibility to high-performing AI models means more entrepreneurs can join the space and create new businesses with novel approaches. Ultimately, many more businesses will build AI into their infrastructure, and chip sales will continue to expand.

AI isn’t just about computing power anymore – it’s about creativity and utility, too. Constraints have always bred innovation, and as companies begin to do more with less, I think you’ll see a shift in the balance of power.

What about America and data privacy?

Despite my excitement about the AI boom, there are valid privacy fears. When I discussed TikTok a few weeks ago, it was clear that the U.S. was taking a stronger stand on protecting its citizens from foreign interference. DeepSeek presents challenges.

To start, it does censor content deemed sensitive by the Chinese government – that also means that particular historical and political topics are met with ambiguous or sanitized responses.

It collects all of your user inputs, any personal information you share, and technical details like your keystroke patterns. They do explicitly tell you this in their Terms of Service. The data is stored in China, raising concerns about legal oversight and privacy laws.

You likely have very little recourse if your data is accessed by or shared with external entities.

My Thoughts

First, DeepSeek is undoubtedly exciting. Even if they’ve embellished their story, the result is impressive. However, it’s also not the end of Nvidia or American-based AI companies. 

Knowing when to sprint and when to jog is an important skill.

AI is shaping our digital world, and DeepSeek is a massive force function in the industry. It is ultimately a good thing for AI (yes, even the Chinese competition).

However, you should probably ask yourself if the benefits of using Deepseek outweigh the risks.

When I coach people on using generative AI, I’ve always urged caution about sharing your sensitive IP with publicly hosted GPTs. Even with OpenAI or Meta, there are clear risks.

It’s also clear that China isn’t the only entity trying to track you or better use your data. As a reminder, if you don’t know how someone is making money off a product … you’re the product. They’re commoditizing you and your data.

That being said, It’s important to be proactive, and it seems to become even more important every week. The landscape is shifting so quickly.

Even so, you have time.

Being proactive also means being smart. As a user of the tools, you don’t need to switch to DeepSeek. Even as a non-tech entrepreneur, now is the time to explore the AI uses that excite you, not to force big leaps. 

If you’re like me and are an entrepreneur in the space, you should certainly take the lessons from DeepSeek and think about how to apply these ideas in your ecosystem.

But my #1 lesson for AI adoption is the same as my #1 lesson for building any habit: Start small - and start in a way that gives you energy.

Humans are good at seeing big changes on the horizon but notoriously bad at predicting what those changes will be.

The goal isn’t to be so fast that you beat everyone else … just to be fast enough not to be left behind.

Onwards!