The fundamental argument about why DeepSeek is good for the ecosystem

There's been a lot of handwringing (and, perhaps, blind panic in some quarters) over the past few days as a result of the general release and availability of DeepSeek.

I wasn't surprised.

I wasn't shocked.

That's not because I'm some brilliant genius. It's simply because I have been expecting this. This is the deal. This is how it works. We're all playing with exponential technology here and the biggest challenge is to try and avoid applying linear thinking because that's the base-level habit.

It's always tempting to try and think you've got a handle on what's going on, especially when it comes to AI.

When I speak to clients in a commercial advisory capacity, I always make a point of being clear that I don't know what's going to happen. I can speculate. I can assume and predict. But I don't know. No one does.

"Ahh, but Sam Altman does," some have commented to me.

Yes. Yes, indeed. He, they – the key players in the US ecosystem - have got an idea on what they and others are doing.

But there is still – as DeepSeek's announcements evidences – a significant opportunity for novel approaches, innovations, ideas and implementations.

If I reflect over the past weeks and months, I could feel a sense of normality and stability begin to influence and govern the AI world. Davos was brilliant, wasn't it, listening to all these perspectives and points of view about AI, from commentators and the people doing the innovation. Only for us to all be surprised just a few days later – and for the material impacts on share price to be noted.

It was getting all too familiar wasn't it?

Billions-upon-billions being spent.

Huge CAPEX moats rising around Silicon Valley.

Defensive positions being built and exploited.

It's great to have a shakeout.

It's really good to have some different thinking in the ecosystem.

I've been reading widely since the DeepSeek emergence and I was particularly excited to read this post by Julie Bort over at TechCrunch featuring an interview with former Intel CEO Pat Gelsinger discussing DeepSeek.

I liked this quote in particular:

DeepSeek proves that AI can be moved forward “by engineering creativity, not throwing more hardware power and compute resources at the problem. So that’s thrilling,” he said.

Thrilling is a great word, Pat. Julie continues:

Gelsinger wrote that DeepSeek should remind the tech industry of its three most important lessons: lower costs mean wider-spread adoption; ingenuity flourishes under constraints; and “Open Wins. DeepSeek will help reset the increasingly closed world of foundational AI model work,” he wrote. OpenAI and Anthropic are both closed source.

Gelsinger and his team have put their money where their mouth is and started using DeepSeek R1 – and they claim to be in the process of rebuilding their systems with their own (DeepSeek) foundational model, "that's all open source".

The speed at which things are being adopted is absolutely fascinating – but also requires a good bit of thought (and, I'd suggest, rigorous testing) before heading into production.

Elsewhere, I also liked this DeepSeek analysis by Reuven Cohen on LinkedIn summarised with these two specific quotes:

Even more impressive, their expert system intelligently activates specialized modules only when necessary, ensuring that out of 671 billion total parameters, only 37 billion are active at any given time.

and..

This ingenious optimization slashes training costs from $100 million to around $5 million and reduces hardware needs from 100,000 GPUs to a mere 2,000, making AI development accessible on standard gaming GPUs. Think training your model using PlayStation 5 clusters.

So, more, please.

More, more, more.

Keep the innovation coming!