AI at Scale

A handful of genuinely useful framings - token economics replacing PUE, power-and-cooling as one system, the silicon/concrete mismatch - but they are surrounded by significant filler, repetition, and high-level generalities. The episode rarely goes deep enough to give an operator an actionable or non-obvious idea they couldn't have reached themselves.

“PUE, that was the miles per gallon you got out of your data center. The tokenomics of what you're getting out of your infrastructure today is that new measurement”

“concrete is permanent. Silicon is not right. And that mismatch and that fundamental kind of alignment of the rapid innovation of silicon, it's really just changing the industry”

The token-economics-as-new-PUE analogy and the 'age of inference' framing are modestly fresh, but 'build the base, rent the spike' is a direct cloud-era recycling, and most observations about density increases and planning horizons circulate widely in data-center coverage. Nothing here is genuinely contrarian or first-principles.

“build the base, rent the spike”

“Not all tokens are equal in value and tokens have a longer and longer lifespan, if you will, just to touch at the highest level because of reasoning”

Chris Sharp as CTO of Digital Realty is a legitimate senior practitioner who has operated large-scale infrastructure at real volume; Steven Carlini carries the title 'Chief Advocate' at a vendor whose podcast this is, placing him closer to thought-leader/sales evangelist territory. The asymmetry and the branded context dilute overall caliber.

“Chris Sharp, the Chief Technology Officer at Digital Realty”

“Steven Carlini, Chief Advocate AI and Data center at Schneider Electric”

There are real numbers - 227 kW per rack, 8 - 10 kW historical baseline, doubling density per year, cooling at 40% of power draw - and named silicon generations (H100s, H200s, Blackwells, Vera Rubens). However, the striking 30x token-cost reduction claim is completely unsupported, and most evidence stays at illustrative rather than analytical depth.

“227 kilowatts per rack, when in the past we were talking about 10 to 15 kilowatts per rack”

“I have an actual private AI running an Hermes agent behind me on an Nvidia Spark, running a Kimmy K2 model. The ROI of the amount of tokens I can produce off of this is just game changing. It's like a 30x kind of reduction in rate”

The host asks broad, open-ended questions ('what has changed?', 'what hasn't changed?') and never challenges a single claim - including the extraordinary 30x cost reduction figure. The episode is a branded Schneider Electric production with a Schneider employee as co-guest, structurally precluding any meaningful pushback; it ends with the host reading back a bullet-point summary of what was just said.

“What would you say has been the major change when it comes to powering and cooling data centers in the last few years?”

“We should be checking in with you on this every 60 days. Um, it sounds. Because it changes so fast.”