The GenAI Revolution And Product Shortages Send IT Budgets Skyward
November 16, 2026 Timothy Prickett Morgan
It sure is a pity that your pay and your company’s annual IT budget increases – indeed my pay, while we are at it – has not been indexed to the IT spending forecasts put out by the likes of market researchers Gartner and IDC. If that had happened, you would without a doubt be able to better afford the higher prices for just about everything in your life – and me, too!
The latest forecast from Gartner came out a few weeks ago, before Huangmageddon happened and therefore the long-term prospects for increases in IT spending, driven by massive investments in GenAI infrastructure by the hyperscalers, cloud builders, neoclouds, sovereign clouds, and AI model builders, may be impacted in the longer term by what the new execs at Nvidia do and do not do. But in the short term, the order books for Nvidia GPUs are filled for 2027 and are filling up for 2028 and the near-term forecast will probably not change all that much. Much of the growth is from increases in the costs of GPUs (and their HBM memory), CPUs, DRAM, flash, and networking components, but some of that growth – I reckon about half of it – is from increasing shipments of GenAI systems as well as more traditional back-end and front-end infrastructure systems that are also seeing their own upgrade cycles.

For the first time, Gartner is breaking out sales of cloud compute, storage, and networking – what is commonly called Infrastructure as a Service, or IaaS – free of other kinds of services so you can see how infrastructure sales and infrastructure rentals compare and can be added together to get a better sense of what overall spending on datacenter systems – that’s servers, operating systems, storage, and switching – stacks up against spending on enterprise software and other kinds of IT services.
In 2026, Gartner is projecting that datacenter systems revenues will rise by an incredible 73.8 percent to $879 billion, and says further that spending on datacenter systems will rise by 55.5 percent in 2027 to reach $1.37 trillion. This is absolutely astounding. If you add IaaS rentals to datacenter systems acquisitions, then 2026 is forecast to have $1.17 trillion in spending, up 60.8 percent, and is expected to have another 48 percent growth to $1.73 trillion. This is the first time that hardware spending in the aggregate will be larger than spending on enterprise software and IT services (minus the IaaS, of course).

This is, to use a technical term in the IT market, freaking crazy. The last time hardware spending was larger than enterprise software spending or IT services spending, many of us were in diapers in the early years of the advance of mainframes in corporate computing. And that is when software and services were included, gratis, because companies spent a relative fortune on hardware.
Given that the size of the IT market has doubled thanks to the AI boom, you might be wondering what your company’s spending should be as you experiment with AI or roll it into production either embedded in or wrapped around back office (transactional) and front end (analytical and predictive) applications. Or, in many cases, as the basis of brand new types of applications. It is reasonable to assume this. The pattern of the hyperscalers and cloud builders offers some guidance.
In the early days of the GenAI boom, the hyperscalers and cloud builders focused their capital spending on GenAI systems for training and then for inference and they stretched out the lifetimes of their other systems underpinning their businesses. Gradually, they realized that they needed to upgrade their existing business systems so they could have enough capacity to do their jobs better and also to feed into AI systems, and an upgrade cycle for traditional gear got underway at the same time budgets were created to support GenAI workloads in their own right. Having built the AI training systems, now the focus is on getting AI inference systems to scale up in terms of capacity and to scale down in terms of the cost per token processed or output.
Small and medium businesses as well as most large enterprises do not need to train their own AI models – they can adopt closed source models from the likes of OpenAI and Anthropic or open weight models from Nvidia – including the zillions of very specific models on Hugging Face as well as its own Nemotron models. So you at least get to skip the truly expensive – and hardware intensive – part of GenAI. That said, picking an AI inference platform is not trivial. While IBM’s Spyre accelerators might be fine for certain kinds of workloads, particularly if they are paired with the matrix math units in Power10 and Power11 processors, the most advanced, low latency, and highest performance inference is being done on disaggregated inference platforms that run the prefill stage of inference – where the context for a query is processed – on GPUs and the decode phase – where a response to a query is generated with token streams – is run on specialized parallel processors from Cerebras Systems, SambaNova Systems, or Grok. Be mindful of this before you commit to a long-term inference hardware platform, and press IBM to explain how Spyre will – or will not – drive down latency and costs for inference as you need to scale it up.
It is always good for IBM i shops to go with the most integrated approach to any IT problem. But not to the detriment of the future company IT budget. By all means, experiment now. But choose that long-term inference platform very carefully.
RELATED STORIES
Have You Tried To Buy A Server Lately?
IT Spending Forecast Keeps Going Up And Up, But It Won’t Go Away
Gartner Raises 2025 IT Spending Forecast, Puts Out 2026 Prediction
The GenAI Boom Is Only Slightly Louder Than The Dot Com Boom
Picking Apart An Ebullient GenAI Spending Forecast
Gartner Raises IT Spending Forecast For 2024 Again
You Guessed It: AI Will Drive Cloud Adoption
The Numbers For Global IT Spending Are Up And To The Right
Gartner: IT Spending To Grow Faster Than Expected In 2024
Server And Storage Spending To Recover In The Years Ahead
Many Different Kinds Of Cloud, Very Big Piles Of Money
IDC Boosts IT Spending Forecasts For 2023 And Beyond
Software Price Inflation Helps Boost IT Spending In 2023
Spending On Legacy Systems Stalls In Q1, 2023 Forecast Looks Weak
How Does Your Infrastructure Spending Stack Up To The World?
IBM i Has a Future ‘If Kept Up To Date,’ IDC Says
IT Spending Growth Will Not Be As Robust In 2023
The Numbers For Global IT Spending Are Up And To The Right
Inflation Pumps Up Global IT Spending, Supply Chain Deflates It
The Downshifting Of IT Spending Growth Continues Apace
IT Spending Growth To Slow In The Coming Years
Gartner Revises IT Spending Forecasts Upward For 2021 And 2022
Forecast: Systems Spending Steady, Up For Services And Software

