As we move through the first few weeks of 2025, now is the time when we love to predict what’s coming for the new year and beyond.In 2025, it feels like we are entering a new era of innovationThe launch and rapid adoption of ChatGPT in late 2022 took many by surprise. Subsequently, the speedy evolution of generative AI to encompass the generation of images, video, music and more in less than a year also took many people by surprise. Already now we have moved beyond that onto agentic AI where AI can independently make decisions and take actions on our behalf.However, I’m not sure that anyone predicted that in 2025 it would be possible to buy a supercomputer, one that fits in the palm of your hand, for $3,000, and which is 1,000 times more powerful than a laptop costing about the same. When I think about this progress I personally am stunned, especially when I think I must barely be using a fraction of the power of my current i7 core laptop. This is a huge step change in technology, and it’s one that ignores Moore’s law, which says compute power doubles every 18-24 months.Nvidia’s DIGITS supercomputer makes AI compute power much more affordable for developers and corporations. Sure, most developers were already using AI in the cloud, and sure, cloud has historically been cheaper than buying your own computer and infrastructure. However, to me this is much like the shift from mainframes to PCs, but without the processing capacity compromise.What excites me is the potential for innovation to follow. Not only do we have the software to create AI solutions, but we can now easily access huge amounts of compute power, too. The author Malcom Gladwell attributes some of the success of Bill Gates to his access to a mainframe at a time when there weren’t many around and when they were expensive to run/use. In his book, Outliers, Gladwell argues that those that get their 10,000 hours practice of a skill have a much higher probability of success than those that do not. This applies to almost anything, from sports to technology or art.Once DIGITS units start shipping, I think it will be a race to get your 10,000 hours use in first. We will inevitably go through solutions that simply do things we have previously done just faster and better. We will also see the innovators reserving first shipments and laggards bewildered by what anyone actually does with all this compute power.But what excites me specifically is the potential to redefine businesses and create new business models. It feels like we are entering a new era of innovation as already we are seeing AI built into ordinary devices like washing machines, vacuums and watches.As I have written before about the “single brain”, the high-level model is very simple:Sense – detect events.
Decide – using appropriate data, decide how to react.
Respond – action accordingly.
This is a parallel to conventional computing: Input->Process->Output.This is not difficult to understand, and it’s
