Hyperscaler Capacity & Supply Forecasting

Overview

Capacity planning failures are rarely modelling failures — they are supply surprises. A region goes allocated, a lead time stretches from weeks to quarters, a committed block fails to materialise when it is needed. This product builds a forward view of accelerator availability across regions and providers, combining observable signals with infrastructure lead-time modelling so that commitments are made against realistic supply.

What is covered

• Regional availability tracking per accelerator generation and instance shape
• Lead time observation and trend, including allocation-request behaviour
• Infrastructure pipeline analysis: announced capacity, permitting and construction timelines, power constraints
• Power and cooling constraint assessment as the binding limit on new capacity
• Supply chain signals: component lead times, packaging and memory availability
• Provider commitment behaviour: how reliably announced capacity converts to allocatable capacity
• Scenario modelling for demand shocks and supply disruption

Deliverables

• Forward availability view by region, accelerator and horizon
• Lead time index with observed versus announced comparison
• Constraint analysis identifying the binding limit per region
• Sourcing strategy recommendation: where to commit, where to stay flexible, where to dual-source
• Risk register for the capacity plan with modelled impact

Who this is for

Organisations making multi-quarter accelerator commitments who need to know whether the supply behind those commitments is real.

Specification

Horizon

Rolling 4 to 12 quarters

Granularity

Per region, per accelerator generation, per instance family

Constraint modelling

Power, cooling, construction, component supply, allocation policy

Signal sources

Public announcements, observed availability, lead-time telemetry

Output

Forecast, lead-time index, sourcing strategy, risk register

Related services