A Chinese tech startup called DeepSeek has sent shockwaves through US stock markets and reignited debate over economic and geopolitical competition between the United States and China. So, What is DeepSeek, and why did it shake US tech stocks?
The company’s AI chatbot soared to the top of Apple’s free app downloads on Monday, drawing curiosity as a potential rival to ChatGPT.
One of the main concerns among US tech observers is the claim that DeepSeek has rapidly caught up with American firms at a fraction of the cost. If true, it could cast doubt on the massive spending US tech giants say they require for developing advanced generative AI. However, hype and misconceptions surrounding DeepSeek’s progress have also contributed to the market’s anxious reaction.
A rapid rise in AI
Founded in 2023 in Hangzhou, DeepSeek entered the spotlight by releasing an advanced large language model supposedly on par with offerings from major US companies like OpenAI. Reports suggest it requires fewer resources, thanks to more efficient training processes. The real jolt came when DeepSeek published research on its new “R1” model, demonstrating sophisticated “reasoning” capabilities—such as revising its approach to solving maths problems—at dramatically lower costs. This prompted Bernstein analyst Stacy Rasgon to note that the model’s economics “freaked people out
Political undertones
The release of DeepSeek’s findings coincided with Donald Trump’s inauguration, fuelling concerns about the US potentially over-regulating AI, thereby surrendering its edge to China. Venture capitalist Marc Andreessen called DeepSeek’s development a “Sputnik moment,” alluding to how the Soviet Union’s 1957 satellite launch sparked a frenetic space race. Meanwhile, China’s own leaders, including Premier Li Qiang, have publicly urged entrepreneurs and tech firms to intensify domestic innovation—a stance exemplified by DeepSeek’s CEO, Liang Wenfeng, who was the sole AI figure at a recent meeting with the premier.
Collision with US export controls
DeepSeek’s emergence also calls into question the effectiveness of US export restrictions on advanced semiconductors. Despite efforts to curb sales of cutting-edge AI chips to China, DeepSeek confirms it used around 2,000 Nvidia H800 chips—compliant with US regulations—to train its models. In a sign of market jitters, Nvidia’s share price plummeted by 17 percent after DeepSeek’s chatbot launch, although Nvidia welcomed DeepSeek’s innovation and forecast increased future demand for its chips.
Challenges and outlook
Alongside its success, DeepSeek has battled large-scale cyber attacks, forcing temporary registration limits, as well as sporadic outages caused by sudden popularity. Nonetheless, the startup’s rapid ascent highlights China’s growing ambitions in AI and underscores intense competition between Beijing and Washington. With significant investments already committed by American tech giants, DeepSeek’s progress has reignited the debate over whether US companies can maintain their leadership or risk being overtaken by a leaner—and increasingly sophisticated—Chinese rival.
The Chinese company just shattered OpenAI’s $18 billion advantage:
They proved you can build ChatGPT-level AI for 100x cheaper.
And it’s about to trigger the biggest market crash since the dot-com bubble.
Here’s why the Chinese are keeping this a secret: For the past 5 years, the AI industry has been built on one core assumption: Building powerful AI requires massive resources. Huge data centers. Thousands of engineers. Billions in funding.
That assumption just got shattered: OpenAI has raised $18 billion.
They’ve built some of the world’s most advanced data centers, hired the brightest minds, and spent years perfecting their models. All based on the belief that this was the only way forward.
But something just changed everything: A new AI model has matched their performance at a fraction of the cost. Not just marginally cheaper.
We’re talking about an order of magnitude reduction in what it takes to build world-class AI. This isn’t just disruption – it’s an extinction event. China knows this, that’s why DeepSeek just restricted registration to mainland China mobile numbers only.
To understand why, we need to look back at the dot-com crash of 2000. History doesn’t repeat itself…it rhymes. Companies spent hundreds of billions laying fiber optic cables under oceans. This is remarkably similar to what happened with GPUs in the 2000s.
They built enough infrastructure to handle 50 years of internet traffic growth. Everyone was convinced this was the future: But they made a crucial miscalculation: The infrastructure far outpaced actual demand. Companies like Webvan built billion-dollar warehouses, assuming everyone would instantly start buying groceries online. They were right about the trend, but catastrophically wrong about timing.
That is a universal law of disruptive technology – eventually, supply exceeds demand and the cycle resets: When reality hit, it was brutal:
• Stock prices collapsed
• Companies went bankrupt
• Trillions in value vanished
Today’s AI boom is following the exact same pattern, but the numbers are even more shocking: Sequoia predicts the AI industry needs $600 billion in revenue to justify current investments. We’re at just 10% of that today. This isn’t just a gap – it’s a chasm between expectation and reality. And now comes the catalyst that brings it all down: This breakthrough in efficient AI development proves something stunning: Most of the massive infrastructure being built right now is unnecessary.
Think about the implications:
• Billions in data centers? Overkill
• Massive hardware investments? Wasteful
• Current AI valuations? Pure fantasy
Here’s why this changes everything: The entire economic model of the AI industry is built on scarcity: “Only a few companies can afford to build powerful AI.”
That scarcity just vanished. And when artificial scarcity disappears, bubbles pop: Just like the internet in 2000, AI is about to enter its “trough of disillusionment.” Not because the technology doesn’t work.
But because we’ve built an entire industry on assumptions that are no longer true. History tells us what happens next:
After the dot-com crash, something fascinating happened: While the media declared the internet “dead,” the real revolution began quietly. Amazon wasn’t building unnecessary infrastructure.
They were solving real problems for customers: The same transformation is coming to AI.
Web4 will emerge from the ashes, built on fundamentally different principles:
• Efficient, affordable AI models
• Decentralized infrastructure
• Natural human interfaces
• Value from data quality, not brute force compute
This reshapes the entire landscape: The winners won’t be who you expect. Just like Amazon understood retail better than understood pets… The next AI giants will be built by people who understand real human needs. Three massive shifts are already beginning:
1. Edge Computing Over Cloud:
AI moves from massive data centers to local devices, bringing privacy and speed
2. Data Quality Over Size:
Better data beats bigger models
3. Practical Uses Over Demos:
Solving real problems beats impressive parlor tricks
The smart money sees this coming:
We’re witnessing a pivotal moment in tech history.
The AI bubble will burst.
But just like the internet after 2000, what comes next will be far more revolutionary than anyone expects.
