GPuuaS Calculator
The business case for GPU-as-a-Service modelled on your own assumptions, over a five-year horizon.
Built on GSMA Intelligence data, modelling and industry expertise. Model both sides of the GPUaaS equation: the infrastructure you need to build and the customer demand required to fill it.
As AI compute demand accelerates, operators and infrastructure providers face a critical question: is there a viable business case for building GPUaaS capacity and how do you size it correctly?
The GPUaaS Calculator lets you model both sides of that equation using your own assumptions. Set your infrastructure supply (GPU capacity, power draw, hardware costs and growth plans), or set your demand side (customers, pricing, usage patterns and attachment rates). Then see how the two align over five years, including where you risk overcapacity and where demand outstrips supply.
The tool is designed for operators evaluating whether to enter the GPU-as-a-Service market, infrastructure providers planning capacity expansion, and strategy teams building the business case for AI compute investment.
It is free to use and requires no login or configuration. It is built on the same modelling and analytical framework that underpins GSMA Intelligence research for operators and policymakers worldwide.
How the calculator works
The calculator has two input panels and two output views. Set your supply and demand assumptions, then see how infrastructure capacity matches up against projected workloads over five years.
Demand inputs
Set the demand side: number of training and inference customers in year one and their growth rates, average GPUs per customer, and attachment rates for managed services and connectivity. Then set pricing — on-demand GPU hour rates, storage, data transfer, managed service fees, setup fees, electricity and usage hours per day.
Capacity and workloads
See how your GPU estate and available GPU hours grow over five years, alongside projected customers, sold GPU hours and utilisation rates. Utilisation is the key signal it tells you whether your supply matches demand, and where the crossover happens.
Supplier inputs
Set GPU capacity — how many GPUs you plan to install in year one for training and inference, and your annual growth rate over years one to five. Then set power draw per GPU and server, and the full cost stack: hardware, servers, CPU, RAM, SSD, networking, setup, support staff, colocation and connectivity.
What the outputs show
Two output views — one for each side of the equation. Together they answer the question: do you have the right amount of infrastructure to meet demand, and does the business work?
Infrastructure capacity
GPUs installed and servers installed over years one to five, split between training and inference workloads. Available GPU hours per year shows the total compute capacity you are building toward — and how that capacity is allocated. This view answers: what am I building, how fast is it growing, and what can it serve?
Workloads
Projected customers, sold GPU hours and utilisation rates over five years, split between training and inference. Utilisation above 100% signals unmet demand. Well below 100% signals overcapacity and stranded investment. The five-year view shows where the crossover happens so you can plan your build-out accordingly.
About GSMA Intelligence
GSMA Intelligence is the definitive source of global mobile operator data, analysis and forecasts. Its datasets cover every operator group, network and MVNO worldwide, delivering one of the industry's most comprehensive and trusted data foundations.
The AI Inference Calculator reflects GSMA Intelligence's independent analytical expertise in supporting the telecoms industry's understanding of emerging technology deployment at scale.