The State of Enterprise AI
What peer data reveals about spending and adoption
- Vendor spending, performance and switching intent
- Vendor excitement and legacy displacement
- Technology adoption and vendors in use or in plan
- Special focus on AI procurement
Enterprise AI became a material line of IT spend in 2026, and buying and cost governance practices have not fully caught up. This joint study by NPI and Arcana Research shows what large enterprises spent on AI, where that spending was heading, which vendors are gaining or exposed, and how AI is being bought. It is written for 2027 planning, whether your AI budget is still open or already set.
AI is now a material IT category, and buying practices have not caught up.
At the end of 2025, two-thirds of the companies in this study spent $25M to $75M on AI, and 91% expected to spend more in 2026, most by 20% to 40%, which proved correct. Since then, AI spend and use has reached a new trajectory. In the $2B to $10B segment of enterprises, AI already accounts for as much as 14% to 20% of external IT spend.
Adoption is broad. Data platforms, foundation models, AI compute and coding tools are each in use at more than half of enterprises. No vendor leads any category by a wide margin, and all 23 vendors rated in the study score higher on execution than on value for money. Sourcing typically joins after the technical shortlist, and the fastest-growing spend is the hardest to see: 66% lack visibility into model and API consumption.
AI is a budget category now.
The scale of enterprise AI investment going into 2026, and how large a share of the IT budget it represents.
Why we asked
Every 2027 plan, renewal and budget review starts from a base. We asked what enterprises actually spent on AI in 2025, how much they planned to grow it in 2026, and how that compares with total IT spend, so the plan starts from evidence rather than from a vendor’s forecast.
What the data tells us
65% of enterprises spent $25M to $75M on AI in 2025, and 91% expected that to rise in 2026, half of them by 20% to 40%. Only 2% expected a cut. Nearly one in five spent more than $75M, and the tail below $20M was thin at 16%.
Share of IT spend tells a more useful story. In the $2B to $10B segment, AI is up to 14% to 20% of external IT spend, large enough to shape the whole budget. In the largest enterprises it is 1% to 2.5%, small enough to sit unnoticed inside cloud and SaaS lines, which is where the visibility gaps in Chapter 5 begin.
What this means for 2027. The base is eight figures. Check 2027 plans against this base and the growth rate, then ask which contracts carry the spend and when they renew. Most will renew on consumption volumes that haven’t been benchmarked.
For Sourcing
- Manage AI as a category, with one owner, a preferred-vendor list and unit-price benchmarks.
- Map the 2027 renewal calendar; consumption contracts will renew on much larger volumes.
- Consolidate where the same model is bought through several contracts.
For IT
- Share a 2027 demand forecast with sourcing and finance.
- Flag which 2026 pilots are heading to production.
- Separate AI infrastructure from application spend.
For Finance
- Give AI its own budget line, with entitlements next to actuals.
- Use the share-of-IT ranges as a sanity check on your own number.
- Plan for 20% to 40% growth and ask which controls come with it.
Six takeaways, written for 2027 planning.
Even if your 2027 budget is set, you still have a lot of say in how far it goes. From the terms you negotiate at renewal to when sourcing gets involved, to how you track usage and how you structure new AI contracts. Each chapter lays out practical steps sourcing, IT and finance can take right now.
- 01AI spending
The 2025 spending baseline and AI’s share of IT - 02Six core technologies
Adoption, ratings and renewal intent across models, data platforms, coding tools, compute, consulting and governance - 03Vendor profiles
The value gap, market signals and seven vendor relationships - 04Emerging technologies
What is being piloted and planned versus what is deployed - 05Special focus: AI procurement
Entry point, visibility, controls and levers - 06Five moves for 2027
NPI’s playbook, study method and about the authors
Download the full report.
51 pages of peer data from NPI and Arcana Research on what enterprises spent on AI, which vendors are gaining or exposed, and how AI is being bought, with a five-move playbook for 2027. Fill in the form and we’ll send you the PDF.
Tuesday, Nov 17, 1pm ET. Jon Winsett of NPI and Arcana Research walk through the findings and the five moves for 2027.
Save your seatTwo teams, one view of enterprise AI.
NPI is the premier provider of data, services and tools to help large enterprises identify and eliminate overspending on IT purchases. NPI delivers transaction-level price benchmark analysis, license and service optimization analysis, and vendor-specific negotiation intel that enables IT buying teams to drive material savings and measurable ROI.
NPI analyzes billions of dollars in spend each year for clients spanning all industries that invest heavily in technology.
npifinancial.com
Arcana Research, the de-facto enterprise AI peer network for benchmarks, fact-based data on spend, adoption and vendor performance, all delivered with no spin or bias.
This report may be shared in its entirety with attribution. Excerpts, charts and data points may be quoted with attribution to “NPI and Arcana Research, The State of Enterprise AI, 2027 Planning Edition.”