Global AI spending is heading for $2.7 trillion
Gartner raised its 2026 forecast again in September. Most of that money is going into infrastructure, which tells a smaller business where not to compete.
On 16 September 2026 Gartner raised its forecast for worldwide AI spending to $2.7 trillion for the year, a rise of 49.5 per cent on 2025. It is worth pausing on the shape of that number rather than the size of it.
Where the money is actually going
Just over $1.48 trillion of the total is AI infrastructure: servers, data centres and the power to run them. That is 55 per cent of all AI spending in the world going into the plumbing. Services account for roughly $577bn and software for $462bn. The parts most people think of as artificial intelligence are far smaller: AI agents and assistants at around $29bn, generative AI models at around $28bn.
Gartner's own summary is blunt about the driver. Capacity growth from hyperscalers and service providers buying AI optimised servers remains the single largest area of spend.
Why that is good news for a smaller employer
An organisation of thirty or three hundred people is never going to compete on infrastructure, and it does not need to. The infrastructure is being built by other people at enormous cost and rented back by the hour. What is left to compete on is judgement: knowing which of your own processes are worth changing, and being able to tell a working product from a convincing demonstration.
That is a skills question, and it is a much cheaper one to answer.
The practical read
- Do not benchmark yourself against the infrastructure spending in the headlines
- Assume the tools will keep getting cheaper and better without your help
- Put the effort into knowing your own processes well enough to pick the right one
- Buy capability, not capacity
Source: Gartner press release, worldwide AI spending forecast, 16 September 2026.