Calculating the GPUaaS opportunity

Cost reduction and operational efficiency remain the primary objectives for most AI deployments worldwide.  However, new AI deployments in the telecoms sector show a growing emphasis on external revenue generation, particularly in Asia Pacific, MENA and North America. In the past six months, 35% of new AI deployments have a revenue objective, according to the GSMA Intelligence Telco AI Tracker. 

One revenue model that is gaining particular attention is GPUaaS, which involves operators purchasing GPUs and then renting compute to businesses and public sector organisations on either a short-term (e.g. hourly) or long-term basis. More than 30 operators worldwide have launched GPUaaS solutions, and this number likely to grow in the coming years as new data centre facilities come online and compute shortages begin to ease.  
GSMA Intelligence estimates that telco GPUaaS revenues will reach almost $9 billion by 2030, up from less than $1 billion today. Revenues will be concentrated in markets with strong sovereign AI demand, stemming from national AI strategies, data-residency requirements and sovereign procurement mandates. Indonesia provides an early example of the revenue opportunity for telcos.

Indosat Ooredoo Hutchison ‘s neocloud business reported revenues of $33 million in the first half of 2026. This figure is expected to grow sharply with the business securing an annual recurring revenue pipeline of more than $300 million, which will begin contributing from the start of 2027. In total, the operator has locked in contracts worth well in excess of $1 billion over the next few years.   GPUaaS sales are a key part of this, alongside revenues derived from selling access to its own LLM, applications and other capabilities.  
Such early proof points highlight the attractiveness of the GPUaaS opportunity, but operators must carefully weigh up market entry. The main risk is spending vast sums on GPUs that could become quickly obsolete as newer versions are introduced. This means operators should take a phased approach to launching GPUaaS offerings, targeting specific verticals or regions within a country. They should also seek partners to help spread the financial burden and reduce risk exposure.

To help operators and their partners better understand the GPUaaS opportunity, GSMA Intelligence recently launched its GPUaaS calculator. The calculator allows users to assess the business case for GPUaaS based on their own assumptions, over a five-year horizon. Built on GSMA Intelligence data and modelling, it assesses the infrastructure needed to build a GPUaaS offering and the customer demand required to fill it. Users can model both training and inference workloads, helping them to see how different usage patterns affect return on investment.  

The GPUaaS calculator forms part of GSMA Intelligence’s growing suite of tools to help operators better understand the economics of new market opportunities. So far, beyond GPUaaS, we have introduced calculators covering AI inference use cases at the edge and will be extending this to sovereign AI. We will soon be introducing 5G monetisation options covering speed-based tariffs and differential connectivity (e.g. for paid priority lanes), as well as calculators on satellite D2D and telco energy optimisation. 
All of these are done with one thing in mind: to help our customers understand the business case before deployment. In a fast changing world of tech and AI, geopolitical dynamics having unprecedented influence, and with growth under pressure, intelligence to support decision making has never been more important.