For more than three years, the central question around artificial intelligence was whether inference demand would materialize; the evidence is now here. Inference workloads today account for roughly 40%+ of installed AI compute and are expected to account for 75%+ of go-forward demand, while the Big Five US hyperscalers have collectively guided to US$750+  billion of CapEx for 2026, with all five pointing to supply as the binding deployment constraint, not demand.

Our latest report frames "Neoscalers"  as a new challenger to hyperscaler dominance across the AI stack, dividing these into “Model Neoscalers” (frontier AI model engines mainly leasing infrastructure) and “Infra Neoscalers” (GPU cloud platforms and AI-grade data centre landlords).

The report explores the variables that will determine who wins the upcoming infrastructure cycle, including:
 

• How the AI stack dynamic is evolving from the app to the infrastructure layers, including where hyperscalers still dominate, and Neoscalers challenge.
   
• How to assess neoscaler risk through three dimensions: use-case risk, business model risk, and access to capital risk.
   
• How to evaluate global market archetypes for AI infrastructure investment, mapped across major economies including the US, Australia, and the broader APAC region.
   
• The four distinct business models emerging among Infra Neoscalers, and how capital markets are responding to them.
   
• How GPU obsolescence is not a tail risk, but a base-case assumption that should be priced into every capital structure from day one.
   
• Why contract quality,is now the single most deterministic driver of project economics.
  
  Download the full report to understand the risk framework, financing evolution, and what it means for investors, operators, and lenders navigating this rapidly evolving sector.
  
  Download