Subject: Disruptive innovation

Good morning!

Disruption often feels like it comes out of nowhere—until you step back and realize it was entirely predictable. This week, we explore a fascinating case of disruptive innovation: DeepSeek, a Chinese AI startup that just shook up the industry in a big way. If you thought OpenAI, Meta, and Google were untouchable in the AI race, this development might make you rethink everything.

DeepSeek’s success is a classic case of Clayton Christensen’s Disruptive Innovation Theory in action: start at the low end, provide a “good enough” alternative, and methodically work up the value chain until the incumbents are forced to react.

Here’s what this means for businesses, investors, and the future of AI.

Why DeepSeek Matters

DeepSeek’s latest model, DeepSeek-R1, has gained traction by offering a cheaper, more efficient, and surprisingly competitive alternative to the dominant Western AI models. While OpenAI and Google have invested tens (or hundreds) of millions into training their models on advanced NVIDIA GPUs, DeepSeek has taken a different approach:

Cost Efficiency: DeepSeek trained its model for just $5.6M—compared to OpenAI’s and Google’s $40M–$200M per model.
Hardware Strategy: Due to export restrictions on advanced chips, Chinese firms have mastered distributed training across less powerful GPUs, achieving impressive results without top-tier hardware.
Open-Source Advantage: DeepSeek has released its models under the MIT open-source license, allowing widespread adoption.
Targeted Applications: While Western AI models focus on general-purpose tasks, many Chinese AI models are narrower but optimized for specific industries, making them incredibly efficient.

This development challenges the traditional assumption that the best AI must be the most expensive AI. Instead, it follows a well-documented path of technological disruption—one that echoes past shifts in industries like steel production, where mini-mills eventually overtook traditional steel plants.

The Business Implications: What Happens Next?

If we apply the principles of disruptive innovation, we can anticipate some key shifts:

1. The Price of AI Will Drop—Dramatically

DeepSeek’s emergence signals a deflationary trend in AI costs. Just like smartphones democratized computing power, AI is about to become more accessible and cheaper. This is great for businesses that want to integrate AI without massive investment.

➡ Takeaway: If you’re in business, this means more affordable AI tools for automating tasks, improving decision-making, and driving efficiency.

2. Small Language Models (SLMs) Will Rise

We’re seeing the rise of small language models (SLMs)—AI systems that use fewer resources and produce targeted, domain-specific results. DeepSeek proves that a leaner model can still deliver competitive performance, and this will pressure Western AI companies to rethink their expensive, high-end approach.

➡ Takeaway: Expect companies to adopt custom AI models tailored to their unique needs, rather than relying solely on massive, general-purpose models.

3. The AI Market Will Fragment

Businesses will face a key decision: should they stick with Western AI models (OpenAI, Google, Meta), adopt Chinese AI alternatives, or hedge their bets by using both?

Companies that rely on AI will need to consider:

Data privacy risks (especially if using Chinese AI models)
Regulatory concerns in different countries
The advantage of using multiple AI models (reducing reliance on a single provider)

➡ Takeaway: Firms will likely diversify their AI strategies to balance cost, security, and innovation.

4. American AI Companies Might Struggle to Disrupt Themselves

History shows that large incumbents often fail to adapt to disruption. Even when they see it coming, they resist change due to:

Sunk cost fallacy: They’ve spent too much on high-end infrastructure.
Profit margins: Premium AI models make more money per user.
Internal incentives: Employees are rewarded for maintaining the status quo, not for disrupting it.

Unless OpenAI and Google actively disrupt themselves, they might lose market share not just to China, but to new Western startups building low-cost alternatives.

➡ Takeaway: Expect AI startups to emerge in the U.S. and Europe, using lower-cost architectures similar to DeepSeek’s.

What This Means for You

So, what’s the actionable insight here? Whether you’re a CEO, entrepreneur, investor, or just an AI enthusiast, here’s how you can stay ahead of the curve:

Adopt AI early and experiment—if AI costs are coming down, now’s the time to test how it can transform your business.
Follow the disruptors—DeepSeek is just the beginning. Watch for AI startups that challenge the big players.
Don’t overpay for AI—The idea that only the most expensive AI is useful is outdated. Look at leaner alternatives that do the job efficiently.
Diversify your AI strategy—If your business relies on AI, consider using multiple providers to mitigate risks and costs.

The bottom line? AI is becoming a commodity. The businesses that embrace this shift early will have a major edge.

For the full details on DeepSeek’s disruption and what it means for AI strategy, read the full article here: HBR - WhyDeepSeek Shouldn’t Have Been a Surprise.

Until next week,
Werner Mouton, CGMA

Monday Insights #5: The Disruption You Didn’t See Coming