Some of you have asked about my take on #DeepSeek’s recent announcement, and it’s sparked some interesting thoughts about where the AI industry might be heading…
If DeepSeek’s claims about #R1 training costs hold true – suggesting they achieved similar capabilities to GPT #o1 at apparently 1/10-1/20th of the cost – we might be approaching a fascinating inflection point in AI economics.
Let’s explore what this could mean:
- Owning the Experience Becomes Everything. What I have learned over the years is that technology, albeit fundamental, rarely creates sustainable advantage by itself. What matters is owning the user experience and the customer relationship. If AI models become commoditized, the real battleground likely shifts to who controls the ‘last mile’: the interface between AI and users, the workflows that deliver value, the experience that keeps users coming back. This is where defensible value can be created.
- Innovation Explodes at the Edge. Here’s the intriguing part: if AI models become commodities, what you can build on top of them remains nevertheless virtually limitless. And this is a good thing, as there would simply be too many possible applications, use cases, and specialized solutions for any single player – even the tech giants – to dominate. At Yahoo, we saw this pattern with advertising platforms: once the core technology commoditized, innovation exploded at the edges, creating entirely new categories of solutions.
- Value Creation Through Orchestration. In this commoditized AI world, competitive advantage might come from how effectively you orchestrate these capabilities. Think less about individual AI implementations and more about creating intelligent workflows that span entire value chains. At dunnhumby, we saw how connecting customer data across touchpoints created exponentially more value than point solutions. With commodity AI, this orchestration could happen through autonomous agents that navigate systems, understand context, and actively drive value – turning integration from a technical challenge into a strategic opportunity.
If this plays out, the commoditization of AI models wouldn’t just be an opportunity – it would be an invitation to reimagine every industry, every workflow, every customer interaction. Just as cloud computing democratized infrastructure and led to an explosion of innovation, cheaper AI could unlock possibilities we haven’t even imagined.
The game might be shifting from “who has the best model?” to “who can create the most value with these models?” – and that would be a game anyone could play.
What’s your take? How would your business strategy change if AI capability becomes truly democratized?