Welcome to the WC, wherein you’re trapped in my mind for eight to ten minutes weekly.
This week, you’ve got two stories to digest:
Table of Contents
Why you should care about DeepSeek
Last week, a seismic shift in artificial intelligence caught the tech world off guard. A Chinese startup called DeepSeek demonstrated an AI model that matches or exceeds the capabilities of industry leaders—but at a fraction of the cost. This development has profound implications for the entire AI value chain and how we think about investing in technology’s future.
And while it’s got obvious flaws…
Hey guys, are you sure you want to give @deepseek_ai all your personal info and ask it sensitive questions?
— Wyatt Cavalier (@itiswyatt) January 27, 2025
Seems like maybe it's not 100% legit. pic.twitter.com/Rupz4aXhES
…the model itself is excellent.
What does it all mean?
What’s DeepSeek?
DeepSeek, based in Hangzhou, has accomplished what many thought impossible: building a state-of-the-art AI model for approximately $6 million. To put this in perspective, Meta plans to spend $65 billion on AI infrastructure this year alone. The company’s R1 model shows performance comparable to leading models from OpenAI and Anthropic but achieves this through innovative architecture and efficient training methods rather than brute computational force.
Their secret sauce combines several technical innovations: – Mixture-of-Experts (MoE) architecture – Advanced knowledge distillation techniques – Innovative load balancing strategies – Efficient training methodologies that bypass traditional supervised fine-tuning.
You don’t need to know any of that. However, you do need to know that American model makers will copy the approach, and it will become the new standard.
The Market Reaction
When news of DeepSeek’s breakthrough hit, tech stocks experienced their worst single-day decline since 2023. Nvidia, the poster child of the AI boom, saw nearly $600 billion in market value evaporate. Other casualties included:
- Cloud infrastructure providers
- Semiconductor manufacturers
- Data center REITs
- AI infrastructure companies
Why such a dramatic reaction? Because DeepSeek’s approach challenges fundamental assumptions about AI economics. The industry has been building massive data centers and designing specialized chips based on the belief that advanced AI requires enormous computing power. DeepSeek suggests there might be a more thoughtful way.
The Jevons Paradox and AI
This is where economic theory offers an interesting perspective. The Jevons Paradox, named after economist William Stanley Jevons, states that when something becomes more efficient and, therefore, cheaper, we typically end up using more of it, not less.
Jevons observed this with coal: as steam engines became more efficient, coal consumption increased because new uses were found for the technology.
Applied to AI, this suggests that while individual model costs might plummet, overall AI usage could explode.
We’re likely to see:
- More applications across industries
- Deeper integration into existing products
- New use cases previously considered uneconomical
- Broader adoption by smaller businesses
This is great for everyone who isn’t a commodity AI infrastructure provider or servicer.
Winners and Losers in the New AI Economy
Clear Losers
Pure Infrastructure Players
- Generic cloud providers
- Basic hardware manufacturers
- Commodity compute providers
- Companies betting solely on computational advantage
Traditional Model Providers
- Companies without specialization
- Pure play model developers
- Basic API providers
- Undifferentiated cloud services
Emerging Winners
Interface Layer Champions
- Companies owning customer relationships
- Enterprise software providers
- Industry-specific solution providers
- UX/UI innovators making AI accessible
Vertical Integration Specialists
Companies combining deep industry expertise, proprietary data, custom implementation, and compliance frameworks. Particularly in regulated industries like healthcare and finance
Data Advantage Players
- Companies with a unique dataset
- Industry-specific data aggregators
- Real-time data providers
- Businesses with strong data network effects
Application Layer Innovators
- Novel AI application developers
- Workflow automation providers
- Custom enterprise solutions
- Industry-specific AI tools
How to invest given the implications of DeepSeek
So where should we put all our money before AI renders currency useless?
Near-term Opportunities
- AI optimization software providers
- Integration platforms
- Monitoring and governance solutions
- Security and compliance tools
Medium-term Plays
- Enterprise software companies with strong AI integration potential
- Industry-specific solution providers
- Companies with unique data advantages
- Firms with strong customer relationships and implementation capabilities
Long-term Bets
- Companies building new categories of AI-native applications
- Businesses creating strong network effects through AI
- Firms developing proprietary AI optimization techniques
- Organizations with defensible vertical integration strategies
Look for companies with strong moats through specialization, unique data advantages, deep customer relationships, scalable value capture mechanisms, and industry-specific expertise
The DeepSeek moment marks a transition from the “bigger is better” era of AI to one where efficiency and application matter more than raw compute power. This shift doesn’t spell doom for AI investment – it just means the value will accrue differently than many expected.
The winners won’t be those who build the biggest AI models but those who can turn increasingly affordable AI into valuable solutions for specific problems.
A badass book
China’s introduction of DeepSeek is a shot across the bow in the (currently cold) war between the USA and China.
I’ve got a not-yet-well-formed theory that this conflict will generate demand in the USA for cultural items created by Chinese dissidents.
With that in mind, consider buying a first edition copy of You Must Take Part in Revolution, co-authored by two rising stars in the Chinese diaspora:
Badiucao ” is a Chinese Australian artist, activist, and political provocateur. One of the most popular and prolific creatives from China, he confronts various social and political issues in his work, often using satire to tackle censorship, authoritarianism, and capitalism.”
Melissa Chan “is an Emmy-nominated Hong Kong and Taiwanese American foreign correspondent based between Los Angeles and Berlin. She was previously posted in China until she became the first journalist in more than a decade to be expelled by the Chinese authorities in 2012.”
While you’re at it, check out:
Badiucao’s artwork. It’s disturbing but stunning
The Economist’s podcast The Prince. It dives into Chinese leader Xi Jinping’s life and gave me a better understanding of both the man and the country.
That’s all for this week; I hope you enjoyed it.
Cheers,
Wyatt
Disclosures
- Our pals at CapitalPad brought you this issue.
- I own no cow gallstones. Yet.
