Spotlight on an indulgent tech sector

We often hear of senior tech industry execs drawing an analogy between the tech sector and automotive manufacturing. Basically, if car development followed the same rules as the tech sector, by now cars would be capable of traveling at speeds close to the speed of light.

There is absolutely no demand for making a passenger car drive that fast. Car makers tend to take small, iterative steps, which is probably why they are struggling to introduce an entirely new drivetrain based on electric motors.

But what we tend to see in the tech sector is the cult that is Moore’s Law, an idea penned by Intel co-founder Gordon Moore, that the number of transistors on a chip would double every 18 months. The flip side of this rule is that, every 18 months, stuff gets cheaper. If the application needed a certain amount of computational power, measured by the number of transistors the central processor unit (CPU) uses, then, 18 months down the road, the application will still be able to run. And it will be able to run on the very same hardware, except this hardware is now much cheaper to buy new as it effectively uses an older generation of processor, compared to the latest technology, which has double the number of transistors.

That’s great for IT buyers but the tech sector is geared up to selling more and more new chips and applications that run best only when the latest hardware is used. So there is built-in obsolescence. It is why a perfectly good operating system like Windows 10 will reach end of life in October, when Microsoft releases its next Windows OS, and then the countdown clock for Windows 11 begins. Every commercial application and PC will need to be recertified; there is a boost in revenue across the industry and the tech sector can continue its growth strategy.

There will always be some who want to have the latest hardware and software. Unless there is a killer application out there somewhere, who really needs a PC with a neural processing unit (NPU)? Budgets are constrained and IT buyers are looking for their software to run more efficiently.

So it is interesting to hear what Nvidia co-founder, Jensen Huang is saying about AI reasoning models like DeepSeek, which are, indeed, able to run more efficiently and so require less expensive hardware. He claims that such AI models can consume 100 times more computational power. Perhaps that is what he’d like to see, as it means more sales of Nvidia’s expensive graphics processor for AI acceleration. But we believe this is not the approach the tech sector should be taking. We should, instead, be laser focused on greater efficiency and cost savings.