The economics of compute are undergoing a seismic shift. With NVIDIA's Blackwell architecture reducing total cost of ownership by over 85%, and newer models like Rubin poised to scale GPU density even further, compute capacity is becoming dramatically cheaper to deploy. Meanwhile, the rise of advanced reasoning AI agents and compact SLMs (small language models) is fueling an explosion in inferencing workloads. These agents require 100x to 1,000x more compute than traditional GenAI models.
As demand soars, spending is diversifying. While the Big Four (Amazon, Microsoft, Alphabet, Meta) still lead compute capex, emerging hyperscalers like Tesla, CoreWeave, and OpenAI are quickly scaling their infrastructure. National governments are also entering the race, launching Sovereign AI projects to secure local capacity.
With compute usage projected to double by 2028 and valuations across semiconductors and hyperscalers still compelling, this report explores who stands to gain most from the insatiable demand ahead
Inside this report:
- Why compute demand is accelerating beyond GenAI workloads
- How NVIDIA's Blackwell & Rubin architectures are reshaping TCO
- Which emerging hyperscalers are building their own data centers
- Why DeepSeek and SLMs are catalyzing a surge in inferencing
- Who benefits most as data center spend crosses the $1 trillion mark
Download the full report to understand why computing is becoming the new oil of the AI era, how to position your portfolio as the next compute supercycle takes shape, and what risks (like oversupply or tariffs) investors should monitor.