The Mavka team is excited to release Data Center Technology: AI Inference Infrastructure Deep Dive, analyzing how the center of gravity in AI has shifted from model training to inference as the primary economic engine of the value chain. The report examines how energy, heat, latency, and capital allocation now define access to intelligence, margins, and control over AI capacity.
The also report highlights three structural shifts reshaping infrastructure strategy: physics as the new boundary for progress, capital allocators emerging as “compute landlords,” and the economics of milliseconds as the organizing principle for user experience and pricing.
Rack power densities have risen from 40 kW to as high as 250 kW per rack, while global colocation pricing has climbed to over 200 dollars per kW per month amid record-low vacancy and heavy pre-leasing of new capacity.
For additional insights or to discuss specific opportunities, please contact:
Misha Edel
Partner
Deep Tech & Infrastructure
Investment Banking
or
Vitaly Golomb
Managing Partner
