Love to see any evidence that spacehopper and the crazy gang are fixing or saving anything. For sure the main benefactors if any will be themselves. It is a bit like setting fire to your house thinking the insurance money will make you rich.

Mea culpa, teach me to proof read before pressing post, though they will be giving it to each other so perhaps a bit correct given they will be receiving the benefits.

AI is both capital and energy intensive though. Expenditure on computer hardware and related infrastructure is growing ever larger. Running into hundreds of billions of $'s in 2025. Massaging EPS by extending the economic lifespan can only defer so much of the cost into the future. Shines the spotlight firmly onto the revenues that are generated.

I do worry that we're basing our investment in this space on today's power and computation needs. I'm not sure many traders really grasp what DeepSeek achieved, with exponentially less compute infrastructure.

AI's ability to optimise itself might catch people by surprise too. After all, the human brain has 100 trillion connections, whereas DeepSeek you're talking 1.5-70 billion. Yet the human brain only needs 20 watts. Computer AI will close in on that level of efficiency, and probably surpass it. Maybe very quickly.

Energy use will eventually come down, but AI systems will always need storage and memory. Companies like Micron, with a P/E ratio of 12, are more certain winners than those with much higher ratios who are likely to be disrupted or commoditised by software companies. I suspect HG Capital and other SaaS companies will become winners in the foreseeable future, as humans (and paying them) will still be needed alongside leveraging AI.

Timescale is important. For now, NVIDIA's results this week will be crucial for them. I suspect it will perform well for a while and remains a top 2 or 3 holding in my SIPP before disruption, but companies within HG Capital are still in the top 10.

Currently, software I use for my job, found in companies like HVPE and Atlassian, use AI and is useful. However, actually using AI to analyse qualitative data is a game-changer, as in the past, only quantitative data was queryable. If the qualitative data is structured properly, its power is remarkable.

To explain as an example on a basic level, what schools are doing is using AI to generate individual exam papers for students based on the examination board syllabus (and other trainings), then uploading the weaknesses of each student. This is a simplified but powerful example of AI in use.

Once we see use cases like this, it becomes easier to find new applications and uses by connecting the dots. For a creative like me, AI has been snowballing... and it seems that AI will eventually have to be commoditsed (cheap as edible chips), resulting in the current crop of high P/E companies in my portfolio (via ATT) being in a bubble. I am sure they know it but have little choice but to spend.

Edit: Apologies as this is going somewhat well off topic. Perhaps the Chinese know all this and events like DeepSeek was only a shock to the west playing catch up!