Tech vendors are ploughing roughly $1 trillion of their own money into AI infrastructure this year, and Gartner's AI servers data study confirms the bill is already being passed to customers through pricier licences, chips and cloud capacity. For UK IT buyers building 2026-27 refresh budgets, the pass-through is no longer theoretical.
View the data behind this chart
| Global tech spend ($ trillion) | February | April | July |
|---|---|---|---|
| Gartner forecast | 6.15 | 6.31 | 6.37 |
A trillion-dollar spend that lands on your invoice
Gartner's John-David Lovelock told The Register that tech companies' own technology spending has reached around $1 trillion and is set to grow 34.7 percent in 2026, a splurge he described as "the largest infrastructure project humanity has ever undertaken." That corporate spending spree is dragging global tech expenditure to $6.37 trillion this year, a forecast Gartner has revised upward three times: from $6.15 trillion in February, to $6.31 trillion in April, to today's figure — a rise of 14.2 percent year-on-year.
For UK finance and IT leaders, the direction of travel matters more than any single number. Every upward revision has coincided with vendors quietly repricing contracts, and there's no indication that trend is levelling off.
Where the hardware squeeze hits UK budgets first
Gartner's device category — laptops, workstations and related kit — is projected to grow 9.8 percent in 2026, and Lovelock was explicit that a significant share of that growth is simply higher prices for memory and chips rather than higher volumes. That tracks with what UK buyers are already seeing on server and workstation refresh quotes, a dynamic explored in depth in our piece on why AI servers cost more.
This is compounded by the sheer scale of physical build-out competing for the same components. Nvidia has confirmed UK deployments including 10,000 Blackwell GPUs with Nscale and 4,000 with Nebius by the end of 2026, part of a wider plan to scale up to 120,000 Blackwell GPUs across UK AI factories backed by roughly £11 billion of infrastructure rollout. Microsoft has separately committed £2.5 billion to UK AI datacentre capacity, including 20,000 advanced GPUs landing this year. Every one of those GPUs is drawing on the same constrained memory and silicon supply chains pushing up device prices generally.
Software licensing inflation and the AI tax
As vendors embed AI features and foundation-model partnerships from providers such as OpenAI and Anthropic into their core products, CIOs are reporting price increases across almost every vendor relationship. Lovelock's assessment to The Register was blunt: buyers are pushing back hard everywhere, but succeeding in only one place — IT services, where providers adding AI to their offering are, counter-intuitively, rewarded with lower price points from customers rather than higher ones.
That's a useful data point for renewal negotiations. If your managed services or IT services provider is bundling AI capability, that's the lever most likely to yield a genuine discount rather than a rate increase. Elsewhere — core software licences, SaaS platforms, model-access billing — expect the increases to stick, and build that into IT procurement services planning rather than assuming negotiation alone will hold the line.
Cloud, IaaS growth and the UK's parallel compute build-out
Infrastructure as a service is on pace to grow 29.3 percent in 2026 to reach $287 billion, up from 25.3 percent growth in 2025, according to Gartner — driven largely by tech companies equipping datacentres for anticipated AI demand. That commercial build-out is running alongside a UK public sector one: government plans include over £2 billion in public support for AI infrastructure, £1 billion to expand the AI Research Resource compute capacity twentyfold, up to £750 million for a new national supercomputer in Edinburgh, and a target of at least 6GW of AI-capable datacentre capacity by 2030 through AI Growth Zones of at least 500MW each.
A separate DSIT procurement is seeking up to £250 million of cloud capacity over four years, running from May 2026 to March 2030 — a signal of how long-dated these compute commitments now are. UK buyers weighing cloud versus owned infrastructure should model that ongoing capacity crunch explicitly using a cloud vs. on-premise TCO comparison, rather than assuming today's cloud pricing holds through a multi-year contract term.
Practical TCO planning for the 2026-27 refresh cycle
Buyers should assume list prices on AI-adjacent software and hardware will keep climbing through the refresh window and plan accordingly.
Sensible steps for the current budget cycle include building an explicit inflation buffer into hardware refresh business cases via an IT hardware refresh planner, sizing GPU requirements precisely with an AI GPU calculator rather than over-provisioning against inflated forecasts, and exploring server financing options to smooth capital outlay against volatile hardware pricing. Where workloads don't need bleeding-edge silicon, refurbished servers and careful server configuration remain a legitimate hedge against the device-price inflation Gartner is now confirming.
View the data behind this chart
| Devices | Services | Telecom | IaaS | |
|---|---|---|---|---|
| 2026 growth (%) | %9.8 | %5.3 | %4.4 | %29.3 |
The open question buyers should keep asking vendors
Lovelock told The Register there's an unresolved debate about whether these price increases are commercially sustainable, or a defensive move by vendors protecting market share as AI reshapes their products — his example being search engines adding generative AI to fend off disruption rather than to generate new revenue. He also flagged that users are already managing costs by switching from capped subscriptions to usage-based billing, turning to lower-cost Chinese models, or adopting open-source alternatives where they'll do the job.
That gives UK buyers real negotiating leverage this cycle: ask vendors directly whether a price rise reflects genuine cost pass-through or defensive repricing, and use the emergence of cheaper open-source and Chinese model alternatives as a credible fallback in commercial discussions, even where you don't intend to deploy them.
- 01The Register — Tech sector pours $1T into AI and sends customers the bill · 27 July 2026
- 02The Register — Suits won't quit AI spending even if they can't prove ROI · 10 April 2026
- 03DataCenterDynamics — UK pledges £1 billion AI compute investment, Nscale to deploy 10,000 Nvidia GPUs · 13 January 2025
- 04DataCenterDynamics — New UK compute roadmap says country needs 6GW of AI-capable data centre capacity by 2030 · 10 June 2025
- 05TechRadar Pro — Microsoft is investing billions to bring AI to the UK · 15 September 2025
- 06DataCenterDynamics — UK govt turns to cloud providers for £250m AI compute capacity procurement · 4 February 2026
- 07Nvidia News — Nvidia and United Kingdom build nation's AI infrastructure and ecosystem · 10 June 2025
- 08ComputerWeekly — Starmer announces sovereign compute strategy amid £11bn chip investment · 11 June 2025