With DRAM and HBM memory makers running at full capacity and unable to boost capacity much, the one thing that will start slowing down is the constant upward revisions in AI spending and therefore global IT spending. There will be price increases that get passed through to customers, which will inch up the forecasts a little, but because HBM capacity is the gating factor in the AI accelerator market, there will not be leaps and bounds in revenue forecasts like we have seen every couple of months for the past several years.

This is certainly borne out by the latest forecast put together by the market researchers at Gartner. The company’s latest forecast for AI spending shows some modest uptick in spending for AI datacenter infrastructure, software, and services in both 2026 and 2027, with the biggest change coming for the servers and networks that underpin AI systems.

For instance, the forecast for sales of AI infrastructure was upped by 4.8 points for 2026 and by 8.1 points for 2027, and I think a lot of this is expected opportunistic pricing more than capacity expansion. AI software forecasts, in contrast, we raised only by a few tenths of a point since the last forecast back in January of this year. The forecast for AI services revenues this year and next were actually tweaked downwards by a few tenths of a point in the most recent forecast.

That said, we are still talking about absolutely enormous amounts of incremental money that has come into the IT sector, which is ironic given how much people used to complain about IT spending growth being between 5 percent to 8 percent in the years before the GenAI boom started in late 2022.

Here is the complete and updated dataset from Gartner for its AI spending forecasts, which we have compiled and updated:

Back in 2024, Gartner characterized the market a little differently, but we have added the appropriate AI segments as we could to give you a four year horizon on AI budgets around the world as expressed in US dollars.

As you can see, the growth from 2024 to 2025 was huge across all AI spending fronts, and by our math using Gartner’s overall IT spending and these AI spending forecasts, AI went from accounting for 13.7 percent of total IT spending in 2024 to 31.7 percent of the IT pie in 2025. Part of that growth in the AI slice is because the rest of the IT spending pie shrunk by 12.6 percent. So anyone who tells you AI is not eating into other parts of the IT budget, they are wrong. And that trend, as best as I can figure, is going to continue indefinitely – even as there is a spending boom to modernize and consolidate traditional servers to make room and budget available for AI projects.

This is a big hurrah, to be sure, which we have seen in the recent financial results from Dell and Hewlett Packard Enterprise. But it won’t last forever. At some point, companies will go back to incremental server installations for traditional servers and things will return to their lowest energy state, and spending across traditional back office and analytics systems will continue their slide.

As for 2027, Gartner has not put out a spending forecast for next year, so as you can see in the bold red italics, I took a stab at predicting what it might be. If my guess turns out to be right, for the first time in history, aggregate AI spending for IT hardware, software, and services will exceed spending on non-AI stuff by a smidgen for the first time in history. And I strongly suspect that the gap will widen for the foreseeable future.

Of course, that assumes that AI proponents will be able to prove that all of that investment is paying off in increased efficiencies and productivity as well as reduced costs and increased revenues due to AI projects. This is not a foregone conclusion, and the winds of change could blow down the straw, stick, or brick house that is AI. (Depending on your own assessment of the state of the world economy and AI.) One good recession could either stop the AI spending spree, or maintain it, or accelerate it at the expense of legacy IT systems.

The only way to predict what will happen for sure is to live through it. So, hang on to your hard hats, everyone.

In-depth coverage of high end computing