If you’ve been keeping even half an eye on global news this week, you’ve probably heard about DeepSeek. The Chinese AI startup burst onto the scene and sparked a wave of panic across Wall Street, sending some of the biggest tech stocks tumbling. How did a relatively unknown company manage to rattle giants like Nvidia and Microsoft—and in such a short time? Let’s dive in.
At the start of this week, many tech investors probably wished they’d stayed in bed. Shares of Nvidia plunged 17%, wiping out around $600 billion in market value—a record single-day drop for any company in history. That alone would’ve been huge news, but it wasn’t just Nvidia; other AI-driven stocks also took a hit. The reason? DeepSeek, a young AI firm from China, claimed it developed a cutting-edge model for a fraction of the cost that big-name U.S. companies have been spending.
Immediately, investors started questioning the logic behind plowing billions of dollars into AI infrastructure if some upstart could do it for less than $6 million. The whole AI sector—from Meta and Microsoft to Oracle and Broadcom—got caught in the crossfire. And just like that, the hype around DeepSeek went global.
DeepSeek’s story traces back to 2023, when Liang Wenfeng, a hedge fund co-founder, decided to launch a dedicated AI lab under the umbrella of his fund, High-Flyer. Fast-forward a little over a year, and DeepSeek dropped a bombshell by releasing R1, its brand-new AI “reasoning” model.
Before that, the startup had been rolling out smaller models and updates, but R1 is a game-changer. Despite using less advanced chips—at least, that’s their claim—DeepSeek produced a model that rivals some of the top efforts by OpenAI and Google. To make things even more interesting, DeepSeek decided to open-source R1, meaning anyone can use or tweak it for free. That starkly contrasts with how much bigger companies charge for their AI services.
One reason is cost. DeepSeek says it built R1 with around $6 million, while U.S. players often spend billions. If that’s accurate, it’s a massive shake-up to the idea that only deep-pocketed organizations can lead in AI. Another reason is that DeepSeek’s success comes despite U.S. export controls that ban sending the most advanced chips to China. And yet, they claim to have created a top-tier AI model without them.
There are also more significant political concerns. The U.S. government just joined a $500 billion AI project called Stargate, launched by President Trump in partnership with OpenAI, SoftBank, and Oracle—an initiative meant to keep the U.S. on top of AI. The sudden emergence of DeepSeek threatens to undercut that effort, fueling discussions about whether China is catching up or even jumping ahead.
Of course, it hasn’t been all smooth sailing for DeepSeek. Right as it took the number one spot on Apple’s App Store with its AI assistant app, the company claimed it was hit by large-scale cyberattacks. That forced DeepSeek to temporarily limit new users just when everyone wanted to try it out. Soon after, word got out that a back-end database exposed some of DeepSeek’s sensitive data, including user chat histories and API keys. The company says it’s fixed the issue, but it’s still an embarrassing security slip-up for a startup now in the global spotlight.
Meanwhile, accusations have been flying. Some industry experts think DeepSeek used more advanced (and restricted) chips than it admits. Others believe the startup might have based its model on OpenAI’s work, using a method called “distillation” to glean insights from ChatGPT outputs. Both claims spark heated debate over who owns what in the open-source world and how you even begin to police that.
Whether DeepSeek’s cost claims hold up under scrutiny or not, the company has already succeeded in rattling big tech. The idea that world-class AI models can be created without astronomical budgets is both exciting and unsettling. If DeepSeek’s approach really does lower the barriers to entry, we might see a wave of new players building capable AI systems—and a shift in how major companies price and profit from their AI offerings.
One thing’s for sure: this last week proved that AI remains a wild frontier, where a single newcomer can shake industries and reshape assumptions almost overnight. So, while cooler heads prevailed throughout the rest of the week, brace yourselves for more breakthroughs and, hopefully, more innovation that benefits everyone.
Markets / Economy
- Equity markets were mostly lower on the week, with the S&P down -1.0%, the Nasdaq down -1.6%, and the small-cap Russell 2000 down -0.9%.
- The U.S. economy expanded an annualized 2.3% in Q4 2024, the slowest growth in three quarters and down from 3.1% in Q3. Forecasts were for growth of 2.6%, according to the advance estimate from the BEA.
- Personal consumption remained the main growth driver, increasing 4.2%, the most since Q1 2023 (vs. 3.7% in Q3 2024).
Stocks
- U.S. equities were in negative territory. Energy and Technology led the decline, while Communication Services and Healthcare outperformed. Value stocks led growth stocks, and small caps beat large caps.
- International equities closed lower for the week. Developed markets fared better than emerging markets.
Bonds
- The 10-year Treasury bond yield decreased six basis points to 4.57% during the week.
- Global bond markets were in positive territory this week.
- Government bonds led for the week, followed by corporate bonds and high-yield bonds.
