CUDO Compute

Why is gas suddenly everywhere in AI infrastructure? It comes down to speed-to-token due to constraints on the electrical grids in regions where gas popularity has grown. It's just the fastest way to power the GPUs generating the tokens customers actually want, and right now, speed wins. But sustainability isn't going anywhere. Every operator I talk to is signed up to ESG goals, SDGs, Scope 3 reporting especially in Europe. That stuff is shaping how people are thinking long-term, even if it's not driving today's decisions. Which is why fuel cells like Bloom Energy are starting to get really interesting. Same gas, but you're running it through an electrochemical process instead of burning it, so you strip out the sulphites and nitrogen oxides. You're still emitting CO2 unless you go hydrogen, but it's a real step forward. There's a quiet hand-off happening. Speed-to-token is calling the shots today, but the cleaner pathways are catching up faster than most people realise. Really enjoyed being part of the conversation at SES Engineering Services roundtable. Check out a clip from the day below.

96% of organisations have adjusted their AI infrastructure location strategy in response to geopolitical and grid conditions. Don't build your infrastructure strategy on outdated assumptions. Our latest research tracks the massive shift in AI deployment economics across the US, UK, and Europe. See the exact data on construction bottlenecks, power consumption, and shifting hardware timelines.

Site selection is no longer about proximity to users. It is about proximity to power. Established hubs are saturated. Where are you building? Infrastructure breaks when land, power, and permitting fail to line up. Based on insights from 701 AI leaders, our new report details how organisations are bypassing grid congestion and relocating their clusters. Download the free 2026 PDF report to see the data driving this shift.

37% of AI leaders cite GPU investment outpacing infrastructure as a key market concern. You have the compute, but do you have the power? Traditional construction timelines can't match AI deployment speed. Read the full 2026 Land, Power & Compute Report online to see exactly how 701 organisations are adjusting their facility locations and adapting to grid delays.

Making AI cheaper to run doesn't solve the power crunch. It increases demand. The hard l limits of the grid are here. Many organisations remain highly confident in their ability to scale, but that confidence is out of step with conditions on the ground. Read the 2026 Land Power. Compute. Report to benchmark your strategy against the realities of the market.

Infrastructure, not silicon, is now driving the cost of AI deployment. Electricity now dictates the real economics of compute scale. GPU pricing matters, but grid delays and energy availability matter more. Don't build your infrastructure strategy on outdated assumptions. Read the 2026 Land, Power & Compute Report online to see how 701 leaders are adapting to the physical limits of the grid.