AI Infrastructure Podcasts

Start with firmware, then follow the bottleneck outward to cooling, networks, silicon and the power grid. That path runs through In The Arena by TechArena: Ken Sun of AMI on Firmware, Fragmentation & AI Infrastructure, then Phononic's Thermal Fabric: Precision Cooling for AI, then OCP's Zane Ball on the Open AI Era and Chip-to-Grid Vision. This is where the subject comes alive in audio: informed people worrying through how the machinery will work at production scale. The models get the headlines. The tension lives in everything required to keep them running.

Technology Now is easier on the ears without sanding away the engineering. The show asks plain questions, lets an explanation get technical, then translates it back into practical stakes. Quiet, more efficient, and liquid cooled: the future of data centres in the age of AI | Jason Zeiler takes the cooling route; Self-driving, intelligent, and built for AI: the future of networking in the 2020s | Rami Rahim turns to networks. Its two-part pairing, Inside the UK's fastest supercomputer: Isambard AI and Powering the UK's fastest supercomputer: Isambard AI, is a useful way into large-scale compute.

The two-host format of No Priors: Artificial Intelligence | Technology | Startups feels like a peer briefing between investors and engineers. Questions skip the elementary tour and head toward architecture or strategy. Re-engineering the Semiconductor Supply Chain with Intel CEO Lip-Bu Tan puts manufacturing capacity in view, while How Nuclear Will Unlock Energy Abundance with Valar Atomics Founder Isaiah Taylor follows the buildout to its energy requirements. Really Big Test-Time Compute in AI Changes Benchmarks, Safety and Research with OpenAI Research Scientist Noam Brown brings compute demand back to the behavior of the systems themselves.

For a quicker scan of possible choke points, 6Pages Market Shifts packs several signals into each edition. 3 Shifts Edition (Apr 10 2026): Anthropic's Mythos and Project Glasswing, Is advanced chip packaging the next AI bottleneck?, The business of weather modification puts packaging directly on the agenda. Other editions flag inference chips, fiber-optic interconnects and even Data centers in space alongside unrelated market shifts. It is brisk listening for people who want to know what might become scarce next.

Then there is the fight over where any of this gets built. Plain English with Derek Thompson makes a large question the spine of a measured, writerly interview, with technical language usually translated a beat later. Why Americans Hate Data Centers is the obvious entry here. It belongs beside the engineering conversations because local opposition is now part of the infrastructure problem, not an afterthought.

Decoder with Nilay Patel comes at the same pressure point with a drier, more skeptical tech-business voice. Its host likes to test a thesis, restate an answer sharply and see whether the guest will accept the framing. The left and right agree on one thing: no data centers places the political collision right in the title. Listeners who want to follow that collision beyond construction and capacity should continue across to AI in Society or AI Regulation, where the focus shifts from the buildout itself to its wider consequences and rules.

Once AI reaches an operating environment, the conversation changes. The New Stack Podcast sounds like engineers talking just outside a conference room: relaxed, acronym-heavy and skeptical of vendor-friendly claims. Why CPUs still matter in the age of AI agents pushes beyond accelerator fixation. Public cloud vs. on-prem: Summit on where each workload belongs is about placement, while Why your company should (try to) build its own AI SRE brings reliability work into the AI stack.

The InfoQ Podcast is denser. Long guest answers and clause-heavy questions suit listeners who already know the vocabulary and want edge cases rather than a primer. Cloud and DevOps InfoQ Trends Report 2026: AI, Resilience, Platforms, FinOps, and Sovereignty puts AI inside the existing concerns of platform teams. Spite-Driven Engineering: A New Blueprint for Cloud Security in the AI Native Era and How eBPF Empowers Developers to Observe Inside the Linux Kernel in a Safe and Unintrusive Way keep security and observability close to the metal.

A working developer may prefer The Stack Overflow Podcast, where short clarifying questions interrupt longer technical explanations and the host often paraphrases a claim before moving on. The show has Your MVP doesn't need a Kubernetes cluster for an antidote to premature complexity, then You need reliable AI context for your site reliability for the opposite problem: making AI dependable inside real operations. Even Your tokenmaxxing is not valuemaxxing sounds like the sort of reminder an infrastructure team should pin above its cost dashboard.

Making Data Simple approaches enterprise machinery through warmer, less polished conversations. It is interested in use, cost and implementation friction, with enough personal rapport to keep the discussion from turning into a conference panel. Making Developers and AI Agents more productive with an Enterprise MCP. An in depth review with Craig McLuckie, Founder and CEO of Stacklok. brings MCP into that setting. For a wider engineering detour, there is This Episode is Deep into the Tech! From Graph DB Fundamentals to Grid Security: Engineering the Next-Gen Infrastructure.