READ TIME: 5 MINUTES
When people found out a new Chinese AI could beat U.S. AIs for much less budget, markets freaked out.
$2 trillion wiped out.
NVIDIA alone lost $600 billion.
Sam Altman called it an “impressive model.”
DeepSeek launched R1, a game-changing AI model.
Open source and efficient.
API costs are 95% cheaper.
Built for just $6M (compared to $600M for GPT-4).
Runs at $4 per million tokens (vs. $100+ for OpenAI).
Can run on gaming GPUs instead of data center hardware.
As AI becomes cheaper, demand will only rise.
DeepSeek is not only cheaper but also uses 35% less energy.
The AI arms race has only just begun
But in times of change, great CFOs don’t panic.
They stay calm.
DeepSeek CEO Liang Wenfeng Quit $8 Billion Hedge Fund
Most hedge fund managers want more profits. But Liang Wenfeng left finance to do more. In 2008, he believed AI would change everything. That belief led him to start DeepSeek in 2023.
Liang invested in 10,000 Nvidia A100 chips despite U.S. restrictions.
He hired 200+ PhDs from top universities to compete with OpenAI.
He turned down a $10 billion buyout, saying, “DeepSeek is not for sale. It’s a mission.”
DeepSeek faced criticism over its data collection and ties to the Chinese government. China’s National Intelligence Law forces tech firms to hand over data if needed.
Despite this, DeepSeek’s open-source AI models made it a strong competitor.
Liang’s gamble paid off.
DeepSeek built AI models at 1/20th the cost of OpenAI. Its R1 model surpassed ChatGPT in iOS downloads and challenged OpenAI’s pricing.
Silicon Valley and Washington took notice.
And now Liang wants to build AGI to predict the entire financial market.
Risky bets can pay off.
For CFOs, Cheaper, Faster AI is Great News
AI used to be expensive.
Now AI costs are collapsing.
High capital costs held many companies back. Now, with affordable options like DeepSeek, more businesses can adopt AI faster.
SAP CFO Dominic Asam sees more AI models as an advantage.
More options lead to better products.
SAP’s cloud ERP services grew 34% for 12 consecutive quarters, thanks to high-performance AI driving product improvements.
SAP’s approach focuses on data connectivity, using AI to solve specific customer problems.
Big players like Google, Meta, and Microsoft spend $60-80B annually on data centers, assuming ever-growing hardware needs. But if training costs drop 95%, infrastructure spending will shift, impacting the entire chip industry.
3 Things CFOs Must Focus On
Understand AI’s real impact on business challenges. It’s not just about implementing AI models but about finding the right ones to increase productivity and drive growth.
Focus on reliable AI models that work with real data, ensuring no errors or "hallucinations" that could mislead decisions. Compliance and data integrity are key. Pick AI models that meet global standards and work with real-world data to ensure accuracy.
Strategically adopt AI, focusing on customer needs, ensuring data reliability, and staying ahead of the competition.
Berkeley Researchers Replicated DeepSeek's R1-Zero For Under $30
A Berkeley AI research team led by PhD candidate Jiayi Pan achieved what many thought impossible:
This shows that groundbreaking AI advancements don’t need huge budgets. The team proved that small models with 1.5B parameters can solve complex problems just like larger systems.
They used a countdown game to show how reinforcement learning can help models go from guessing to solving tough problems. The choice of algorithm wasn’t as important as expected, meaning smart thinking matters more than just size or tech.
This breakthrough means that AI systems can focus on specific tasks, like multiplication or verification, without needing tons of data.
For the price of a dinner, this research opens up AI to people all over the world.
What Everyone’s Missing
Cheaper AI can lead to bigger problems than you realize.
According to the Jevons Paradox, when technology becomes more efficient, people end up using more of it. This is exactly what’s happening with AI.
As DeepSeek’s AI becomes cheaper and more powerful, companies are running towards it. While this seems like a win, it brings a serious challenge:
Increased energy demand.
Look at coal. Once seen as an efficient energy source. It fueled industrial progress but caused massive environmental harm over time. AI’s efficiency can spark a similar crisis, an explosion of energy consumption and a global resource shortage.
You are already seeing AI models grow in size and complexity, wanting more powerful hardware and massive data centers. This trend will only accelerate, leading to even more strain on global energy resources.
The benefits of AI can’t be ignored.
But just like deepfakes or AI-generated misinformation, we’re blind to some of its unintended consequences.
To stay ahead, CFOs must consider these long-term risks in their AI strategies.
Modern CFOs don’t just lead financial strategy.
They also build sustainable AI practices.
The Bottom Line
DeepSeek didn't innovate.
They copied OpenAI's models and improved them.
Yes, AI is now cheaper, faster, and more powerful than ever.
But the real AHA moment will come if a Chinese AI company builds a model that can do something a U.S. AI company can’t.
That will be a bigger wake-up call than the moment we’re in today.
This is just the beginning.
And that's all for this week.
I’m Wouter Born. A CFOTech investor, advisor and entrepreneur.
Find me on LinkedIn
P.S. Read: OpenAI's CFO Sarah Friar is a standout in tech and finance
Whenever you're ready, there are 2 ways I can help you:
Promote your business to 7,000+ engaged finance leaders by sponsoring this newsletter. (Booking 3 months in advance.)
If you’re building an AI-powered CFOTech startup, I’d love to hear more and see if it’s a fit for our investment portfolio.