Practical thinking on AI infrastructure
Field notes from designing, deploying and optimising AI environments across 20 sectors. No hype. No vendor cheerleading.
The real cost of cloud GPU at scale
When sustained workloads cross the break-even line, and what to do about it.
Designing for the EU AI Act, not just compliant to it
How to bake governance into the AI infrastructure layer rather than bolting it on later.
Why AI traffic breaks traditional network design
Spine-leaf, lossless Ethernet and 800GbE, the new shape of enterprise networks.
Per-model cost allocation done right
Allocating AI spend by model, team and tenant, without slowing teams down.
Why MSPs need an AI infrastructure capability now
Customer demand has overtaken provider readiness. Here's how to close the gap.
UK sovereign AI: what it means in practice
Beyond the slogan, what data sovereignty actually requires from your AI stack.
Why GPU infrastructure needs its own security layer
AI factories are now a primary attack surface. Here's where the perimeter actually sits.
The new bottleneck isn't chips, it's power
Why grid capacity, not GPU supply, now sets the pace of AI infrastructure builds.
Why a single-vendor GPU strategy is now a risk, not a simplification
How the shift toward heterogeneous and custom silicon changes infrastructure planning.