Financing The Robots | Turning Compute Into Something Useful

Everybody NEEDS compute, but what are you willing to pay for it?

In Episode #530 of 'Meanderings', we unpacked what it might actually take to finance the robotic future, starting with the current AI boom, looming IPOs and whether we’re approaching a classic bubble moment. We explored the idea that even if some of the biggest AI companies stumble financially, the underlying value of compute, models and automation won’t just disappear — it may simply get cheaper and become even more useful. Along the way, we dug into inflation versus deflation, why so much capital is pouring into data centres and memory, and how falling compute costs could create the next wave of practical adoption rather than signal the end of the trend.

We also got into the more grounded question of where AI is already useful today. We compared model intelligence with the real unlock of agentic workflows and integrations, and discussed why the biggest near-term gains may come less from “smarter chatbots” and more from systems that can actually do things end-to-end. From there, we zoomed out to robotics, cybersecurity, privacy, physical-world constraints, and what kinds of jobs or human experiences may be hardest to replace. The big theme running through the conversation was this: digital AI may be close to a temporary overbuild, but that same excess could end up fuelling a much bigger shift once robots, real-world applications and human adoption catch up.

What are your thoughts on this, reach out and let us know!

Timeline:

(00:00:00) Intro

(00:00:30) Setting the scene for financing the robotic future

(00:03:20) Anthropic IPO talk and signs of an AI bubble peak

(00:06:01) Debt, inflation and whether AI could be deflationary

(00:10:45) Have large language models already peaked for daily use?

(00:17:00) Productivity gains, media workflows and why jobs remain

(00:24:38) Privacy concerns, open source tools and cybersecurity limits