The AI Shockwave

Yesterday, my sister—who is not a techie by any means and has barely used AI in her career—messaged me out of the blue:

“Tech guy, what do you think of all this noise about DeepSeek?”

That moment stuck with me. When people who don’t normally follow AI developments start asking questions, you know something seismic is happening.

Some have framed this as AI’s “Sputnik moment,” suggesting that DeepSeek’s breakthrough represents a geopolitical and technological shock that will redefine global superpower competition. But I don’t see it that way. The launch of Sputnik in 1957 was an existential wake-up call for the U.S.—a stark realization that the Soviet Union had outpaced them in space technology. It ignited a zero-sum, government-driven race where the only path forward was massive state intervention and funding.

DeepSeek’s rise is significant, but not in the way Sputnik was. It doesn’t signal a single country pulling ahead in an AI arms race. Instead, it underscores a shift in how AI progress is being made—moving away from the domain of trillion-dollar corporations toward a more accessible, decentralized model. With a fraction of OpenAI or Google’s budget, DeepSeek trained a top-tier AI model using a lean team and open-source innovations, proving that cutting-edge AI is no longer reserved for the biggest players.

This isn’t just a breakthrough in AI capability—it’s a shift in the very economics of AI.

Unlike past technological revolutions driven by government funding and state-led competition, AI is evolving along a different trajectory. It’s not a single moment in time but an accelerating wave of innovation. It doesn’t belong to one nation or entity, and while government policies play a role, the real AI revolution is happening in research labs, startups, and open-source communities worldwide.

If anything, this feels more like the rise of electricity or the dawn of the internet. Much like AI today, electricity wasn’t immediately transformative—early power grids were fragmented, expensive, and unreliable. But over time, as the technology improved, distribution expanded and use cases emerged, electricity became an essential layer of the modern world.

Similarly, the internet didn’t change everything overnight. It started as a niche tool for academia and government agencies before being democratized through the personal computer, the web browser, and mobile devices. Each layer of innovation unlocked exponential value, eventually transforming how we communicate, work, and do business.

AI is on the same trajectory. DeepSeek is just another step forward in a long, inevitable march. What makes this moment significant isn’t just that a new player has emerged—it’s that the barriers to AI innovation are falling faster than anyone expected.

The real story here isn’t about geopolitical competition or whether the U.S. can “catch up.” It’s about what happens when AI becomes radically cheaper, more accessible, and embedded in everyday life. The biggest winners won’t be the governments scrambling to control it—it will be the innovators, startups, and industries that figure out how to apply AI to solve real-world problems.

What is DeepSeek, and Why Does It Matter?

DeepSeek has pulled off something that many thought was impossible. With a lean, research-focused team, it trained its latest AI model, R1, for just $5.6 million—a fraction of the $100 million+ OpenAI reportedly spent on GPT-4. By leveraging open-source models like Meta’s Llama and optimizing training efficiency, DeepSeek is proving that AI innovation is no longer just a game of brute-force spending.

This approach has already sent ripples through the AI ecosystem. DeepSeek’s rise is forcing China’s biggest tech firms—including Alibaba, Baidu, and ByteDance—to accelerate their AI strategies. Alibaba responded by releasing Qwen 2.5-Max, a model it claims outperforms both DeepSeek and OpenAI’s GPT-4o.

Beyond China, DeepSeek’s rapid ascent has also sparked debate about AI supply chains and the impact of U.S. export controls. While some speculate that DeepSeek may have secretly used high-end GPUs despite restrictions, the bigger story is how Chinese firms are optimizing training techniques and leveraging open-source AI to stay competitive. Rather than slowing China down, sanctions may have inadvertently forced innovation in efficiency, much like how trade restrictions in other industries have historically led to self-sufficiency and breakthroughs.

But DeepSeek’s success also raises bigger questions: Is AI becoming a commodity? If a relatively unknown team can train a top-tier model for a fraction of the cost, what does that mean for the trillion-dollar valuations of AI giants? Investors are now rethinking whether massive spending is still a defensible moat or if the real opportunity lies in AI’s application layer rather than in the infrastructure arms race. The shock to the market isn’t just about a new competitor—it’s about the very foundation of how we’ve valued AI leadership up until now.

The Market Fallout: Disrupting AI’s Economics

The financial world has finally taken notice—albeit in its usual dramatic, belated fashion. Within days of DeepSeek’s breakthroughs, Nvidia’s stock plunged 17% in a single day, wiping out $589 billion in market cap—the biggest single-day loss of value for any public company in history. Other AI infrastructure heavy stocks like Broadcom (-17%), Taiwan Semiconductor Manufacturing Company (-13%) also took hits.

And yet, here’s the irony: Silicon Valley has been talking about DeepSeek for several weeks now. Mark Zuckerberg even mentioned it in a conversation with Joe Rogan weeks ago, and insiders have been discussing its implications since the first research papers surfaced late last year. But Wall Street? It only started panicking this past weekend.

One has to wonder—did investors just wake up and realize that AI models can be built more efficiently? The news didn’t change overnight, but the market psychology did. The same firms that were bidding up AI stocks to stratospheric levels are now acting like DeepSeek single-handedly ended the AI boom.

Interestingly, while AI infrastructure stocks tumbled, Meta’s stock rose about 5% despite announcing a staggering increase in AI spending. The company reported $14.84 billion in Q4 2024 capex, bringing the full-year total to $39.23 billion. For 2025, it expects to spend between $60-65 billion—nearly doubling its AI-driven infrastructure investment. Meta is making a bold bet that AI efficiencies and engagement gains will more than justify the spend.

So why is Meta rallying while Nvidia and other AI infrastructure stocks are getting hammered? The market seems to be differentiating between companies exposed to AI infrastructure costs and those that stand to benefit from AI-driven efficiency gains. Meta, as a consumer-facing company, is positioned to capitalize on AI optimizations—reducing costs while increasing engagement and ad revenue. Meanwhile, Nvidia and other AI hardware providers are more vulnerable to cost-cutting pressures as companies seek to train models more efficiently.

Another factor? Meta’s open-source AI strategy. Unlike OpenAI or Google, which operate in closed ecosystems, Meta’s Llama models have become a foundational layer for startups and researchers. This positions Meta well to benefit from an AI world where efficiency and accessibility matter just as much as raw power.

What about Nvidia? In the short term, the AI narrative has shifted from one of brute-force compute power to algorithmic efficiency, and that presents challenges for Nvidia, which has been the dominant supplier of AI hardware. But long-term, Nvidia is still well-positioned, much like Apple was in the early smartphone era.

Nvidia’s deep investments in vertical integration—designing AI inference solutions, edge computing, and specialized applications like robotics and drug discovery—ensure its continued relevance. Its GPUs are not just for training; they are critical for large-scale AI inference, a necessity as AI expands beyond text into multimodal capabilities like vision, audio, and robotics. While Nvidia’s dominance in AI hardware remains intact, the short-term market reaction reflects a shift in focus—companies are now prioritizing efficiency over brute-force compute. This doesn’t spell doom for Nvidia, but it does mean volatility. Whether the stock drops by 60% or doubles in the next 18 months, Nvidia’s role in AI’s future isn’t in question—only how the market values it in the near term.

For companies like OpenAI, the situation is more complicated. While private market valuations remain high, the rise of open-source AI is challenging the assumption that only a handful of players can build frontier models. DeepSeek’s ability to replicate OpenAI’s capabilities at a fraction of the cost has raised serious questions about the long-term value of proprietary models and the underlying firms’ business models.

Meanwhile, Apple—largely absent from the AI arms race—may be positioned to benefit from this shift. Unlike OpenAI and Google, which have spent billions on large-scale training runs, Apple has quietly been investing in efficient, on-device AI solutions, betting that AI models will eventually be commoditized and that differentiation will come from user experience and tight hardware-software integration.

The lesson here? AI’s trajectory hasn’t changed—but the market’s mood swings will keep making headlines. What actually matters is not today’s stock price, but who is positioning themselves to take advantage of the AI shift in the years ahead.

The Biggest Opportunity Lies in Applications

While the infrastructure and model layers are still evolving, one thing is clear: the real economic value in AI will be created at the application layer.

McKinsey predicts that generative AI could add $2.6 trillion to $4.4 trillion annually across industries—a staggering figure, that dwarfs the GDP of the United Kingdom. This impact won’t come from a single breakthrough but from AI embedding itself deeply into industry value chains, transforming how businesses and entire industries operate.

But unlocking this potential isn’t just about improving AI models or making GPUs more efficient—it’s about how AI is applied to solve real-world problems. This is the same pattern we saw with the internet. The biggest economic impact didn’t come from the companies laying fiber-optic cables or manufacturing routers. It came from those who built on top of the infrastructure—Amazon revolutionized commerce, Google redefined information access, and Facebook reshaped digital communication.

AI is now at a similar turning point. What was once an expensive, compute-heavy experiment is shifting into a foundational technology that will power everything from healthcare to finance to education. The next decade won’t be defined by who builds the biggest model but by who applies AI in ways that drive tangible value.

AI is at that inflection point now. The shift from an expensive, compute-heavy experiment to a foundational technology that will power everything from healthcare to finance to education is underway. The companies that understand specific industry pain points and apply AI to real-world problems will be the ones that define the next decade.

That’s exactly why I left big tech to jump into the founder grind. The AI paradigm shift isn’t just about technological progress—it’s about building things that actually improve people’s lives.

Building for the Future: Key Considerations for AI Startups

For startups and builders, this moment is filled with opportunity. Success in this new AI era will require technical adaptability, domain expertise, and a relentless focus on user needs. Founders should be thinking about:

• Understanding real-world problems—The best AI applications will solve tangible problems that accrue value to end consumers, not just showcase technical capabilities.

• Building flexible architectures—The ability to rapidly integrate different AI models (rather than relying on a single proprietary system) will be key to long-term success.

• Leveraging AI’s growing efficiency—As AI becomes cheaper and more accessible, companies that optimize for efficiency will gain a competitive edge.

• Looking beyond internal R&D—The open-source AI movement is accelerating. The smartest companies will tap into broader innovations rather than trying to build everything in-house.

The AI Revolution is Just Beginning

DeepSeek’s rise marks a turning point in AI, but it’s not an endgame—it’s a glimpse of what’s next. AI is becoming an unstoppable force, not just in tech but across healthcare, food science, materials discovery, and beyond.

As AI moves from research labs into real-world applications, the biggest opportunities will come to those who can bridge the gap between innovation and impact. The future of AI isn’t just about who builds the most powerful models—it’s about who can integrate them seamlessly into people’s lives to solve meaningful problems.

That’s exactly what I’ve been focused on. At the intersection of AI and health, we’re building something that tackles a major, real-world challenge. I’ll be sharing more soon, but for now, I’ll leave you with this: the AI revolution isn’t just for the biggest tech companies building the biggest models. It’s for those who see the opportunity to build, adapt, and solve problems that matter. Stay tuned—there’s a lot more to come.