Mapping the public market side of the AI factory buildout
Been spending time lately mapping the public market side of the AI factory buildout; who's actually winning, and where the capital is flowing at the infrastructure layer.
CIO Thoughts:
Been spending time lately mapping the public market side of the AI factory buildout; who's actually winning, and where the capital is flowing at the infrastructure layer. Typical another day is looking at some heatmaps!
It is not a rocket science actually, it is a simple question:
which public companies have delivered the highest YTD returns in 2026 among names tied to AI compute, training, and inference infrastructure?
The usual suspects showed up (a bit of yawning here): Nvidia, Broadcom, TSMC. Solid performers, all up.
But two names stopped me (my head tilted a bit when I saw the returns)
CoreWeave. Up 60%+ YTD. IPO'd at $40 less than a year ago. Now sitting on a $66B contracted backlog, a $21B deal with Meta, and guiding $12–13B in revenue for 2026... simply a 140% growth!
And Nebius Group, this Nebius is up 93% YTD. $46B in signed GPU capacity contracts. A $2B strategic investment from Nvidia itself. Q4 revenue growth of 547% year-over-year.
Both are what the market now calls "neoclouds", fancy name I know but I stopped asking about terms sometime ago, and what struck me isn't just the returns. It's what they signal.
Cloud computing, as we've understood it for 15 years, was built around a simple model: virtualize CPU compute, make it fractional, rent it by the hour. AWS, Azure, GCP perfected that.
What CoreWeave and Nebius represent is something structurally different. Infrastructure designed from the ground up where the GPU is the first-class citizen, the MVP in a basketball match. "Non-blocking InfiniBand fabrics" as they call it, I need to do some research here. Gang scheduling for 10,000-GPU training runs. NVMe storage optimized for checkpoint throughput. ML libraries baked in, not installed later.
The hyperscalers built roads. These companies are building formula one tracks, and the hyperscalers themselves are now their biggest customers.
When Microsoft and Meta are redirecting tens of billions in compute spend to specialized GPU cloud operators rather than building it all themselves, that's not a niche trend. That's a structural shift in how AI infrastructure gets delivered.
The returns in the stock price are a symptom. The real story is that "cloud" is being redefined under our feet, and the public markets are pricing it in faster than most people realize.
Still digging. But this one is worth watching closely.
Published in
Eighteen years in investment management, M&A and financial technology. A CFA Charterholder, he leads investment analysis, valuation and fund structuring across Significa's advisory and investment practices, and teaches valuation and fund structures in the CVC Builder curriculum. Concurrently Chief M&A and Growth Officer at AlternativeSoft in London; co-founder of MarkeScope, which developed the first stock-screening solution to incorporate risk measures. Treasurer of CFA Society Egypt; BSc Electrical, Electronics and Communications Engineering, Cairo University.