Conference Presentation, Fireside Chat
The Token-per-Watt Era: What's Next for AI Hyperscalers | Nebius x Accel | RAISE Summit 2026
- Global AI adoption is expected to continue growing rather than flattening, with significant future growth anticipated for Nebius.
- The company forecasts a $11 trillion global AI infrastructure build-out occurring between 2024 and 2029.
- GPU capacity remains sold out, with estimated hyperscaler demand comprising 50% to 70% or more of total market demand, though exact figures are unknown.
- Broad enterprise adoption is projected to lag behind digital natives, as real adoption for the majority of the enterprise market has not yet begun.
- Nebius plans to evolve its product stack from bare metal and managed infrastructure toward a layer featuring managed inference and agentic capabilities to enable customers to pay for "authentic outcomes" rather than compute or tokens.
- Efficiency gains are expected to drive increased demand by allowing companies to solve more complex tasks within existing budgets rather than reducing consumption.
- Frontier open-source models are predicted to unlock new use cases, shifting reliance from vanilla models toward trainable and tunable options that safely inject customer data.
- While lowering the barrier to build AI solutions is projected over the next two to three years, the current state of the industry is considered insufficient for mass developer production.
- Nebius intends to invest five years in building a platform to enable a broad population to develop production AI solutions through managed inference and agentic features.