Here’s something that should concern every CFO: around 80-85% of enterprises miss their AI infrastructure forecasts by more than 25%, according to The State of AI Cost Governance. We’re not talking about rounding errors, we’re talking about budgets that are fundamentally broken before the fiscal year even starts.
AI infrastructure spending isn’t just another line item in your cloud budget. It’s becoming its own domain, with unique cost patterns that traditional IT planning completely misses. The difference between companies that get AI costs under control and those that don’t shows up in quarterly variance reports, margin erosion, and CFO confidence in future AI investments.
If you’re planning your 2026 AI strategy, or trying to explain why last quarter’s AI spend was 40% over forecast, these statistics will show you exactly where the industry stands and what you need to do differently.
Key Takeaways: The 5 Most Critical Statistics
- 80-85% of enterprises miss AI forecasts by 25%+ – Persistent underestimation and weak cost governance plague AI infrastructure planning (Mavvrik and BenchmarkIT)
- Companies plan to spend 1.7% of revenue on AI in 2026 – More than doubling from 0.8% in 2025, with 94% continuing to invest even without immediate returns (BCG)
- <1% report “significant ROI” – Less than 1% of executives report ROI of 20% or greater; 53% report only 1-5% ROI (Forbes Research)
- 30% of GenAI projects abandoned after POC – Expected abandonment by end of 2025, with 40%+ of agentic AI projects expected to be canceled by end of 2027 (Gartner)
- 60% see minimal or no value – Only 5% are AI leaders achieving significant returns; 35% scale with moderate success (BCG)
AI Investment Is Accelerating, But Returns Remain Elusive
The AI spending boom is real. But so is the struggle to show value.
85% of organizations increased AI investment in the prior 12 months
And it’s not slowing down. Some 91% plan to increase AI investment again in the next year, according to Deloitte’s AI ROI study released in October 2025.
This creates a paradox: investment is accelerating while CFOs struggle to measure returns. The pressure to “do AI” is overriding the discipline to “prove AI value.”
Organizations increased AI compute spending by 166% year-over-year in Q2 2025
Spending on AI compute and storage hardware reached approximately $82 billion for Q2 2025 alone, a staggering increase that shows no signs of slowing, according to IDC’s analysis.
One VC estimate from Menlo Ventures puts 2025 generative AI spend at $37 billion, up from $11.5 billion in 2024. Within that, roughly $1.5 billion is specifically labeled as “AI infrastructure” and $4.0 billion for model training infrastructure.
Companies plan to spend 1.7% of revenue on AI in 2026
More than doubling from 0.8% in 2025, this represents a massive commitment relative to overall revenue. According to BCG’s AI Radar Survey of 2,360 executives released in January 2026, every industry tracked plans to increase AI spending, led by technology companies at 2.1% of revenue and financial institutions at 2.0%.
94% of companies will continue investing even without immediate returns
Only 6% would pull back on AI investments if current initiatives don’t pay off in 2026. This unwavering commitment comes as CEOs take direct ownership of AI strategy, 72% now say they are the main decision-maker on AI in their organization, double last year’s proportion.
Half of CEOs believe their job stability depends on getting AI investments right. And they’re betting big on AI agents: 90% of CEOs believe AI agents will produce measurable returns in 2026, leading companies to commit more than 30% of their AI budgets to agentic AI.
Source: BCG – As AI Investments Surge, CEOs Take the Lead (Jan 15, 2026)
Analysts project $3-4 trillion of cumulative global AI infrastructure capex by the end of the decade
This massive investment is driven largely by hyperscalers building out data center capacity, according to Ropes Gray’s industry analysis.
One industry analysis from Environment + Energy Leader estimates about $6.7 trillion of data center capital investment required to keep pace with AI demand, with roughly $5.2 trillion tied to AI-intensive facilities versus $1.5 trillion for traditional IT workloads.
Hyperscalers are announcing individual multi-year data center programs in the $80-85 billion range for 2025 alone, with single vendors signaling several hundred billion of AI infrastructure capex over the next few years.
What this means for you: AI spending is growing faster than any technology investment in recent history. But unlike previous tech waves, the cost structures are more opaque and harder to predict. The shift from CIO-led to CEO-led AI initiatives (72% of CEOs now the main decision-maker) means AI spending decisions are happening at the highest levels—often without the financial controls that traditional IT investments require. If your organization is increasing AI investment without improving forecasting and cost management capabilities, you’re setting yourself up for budget disasters in 2026.
AI ROI: Closing the Gap Between Hype and Reality
Most organizations report “positive” ROI. But dig one level deeper and the picture gets concerning.
72-75% of organizations report positive ROI on generative AI initiatives
That sounds encouraging until you look at how they’re measuring “ROI.” Most frame it as productivity gains and cost avoidance, not margin expansion or revenue growth.
Sources: Wharton – 2025 AI Adoption Report (Oct 28, 2025); McKinsey – The State of AI: Global Survey 2025 (Nov 4, 2025)
Less than 1% of executives report “significant ROI” (≥20% profit or cost-savings uplift)
Only 3% report 10-20% ROI. The vast majority (53%) report only 1-5% ROI. This means most AI investments are delivering marginal returns at best.
Source: Forbes Research – 2025 AI Survey (Oct 9, 2025)
BCG identifies a widening AI ROI value gap
Only 5% of companies are “future-built” AI leaders. Another 35% are scaling with moderate success. But a full 60% see minimal or no material value from their AI investments.
The leaders achieve 5× the revenue increases of peers, see 3× the cost reductions, and can expect 2× revenue growth with ~40% greater cost reductions by 2028.
Source: BCG – Are You Generating Value from AI? The Widening Gap (Oct 15, 2025)
AI payback periods are longer than most tech investments
Only 6% of organizations see AI payback in under one year. Just 13% see payback within 12 months, even among top projects. Most expect 2-4 years to reach satisfactory ROI.
Source: Deloitte – AI ROI: The Paradox of Rising Investment and Elusive Returns (Oct 21, 2025)
What this means for you: If you’re measuring AI success purely by “productivity improvements” or “time saved,” you’re not measuring business value. The gap between AI leaders and laggards is growing. Leaders aren’t just spending more—they’re measuring differently, governing differently, and redesigning workflows to capture value. The 1-5% ROI that most companies report won’t justify continued investment once CFO patience runs out.
How Companies Are Actually Measuring AI ROI
The metrics companies use reveal a lot about why ROI remains unclear.
64% use operational efficiency as the primary ROI metric
Half use data quality improvements (50%), and 48% use employee productivity. Far fewer tie AI directly to P&L or margin impact.
Source: Forbes Research – 2025 AI Survey (Oct 9, 2025)
This is the measurement problem in a nutshell: operational efficiency is important, but it’s not a business outcome. You can be operationally efficient while losing market share or eroding margins.
Only 36% of CFOs feel assured they can achieve meaningful AI outcomes
Despite 39% of CFOs prioritizing “accelerating AI use in the finance function” as a top-5 action item for 2026, just 36% feel confident in their ability to deliver real enterprise impact from AI initiatives, according to Gartner’s Q4 2025 survey of more than 200 finance chiefs.
This confidence gap helps explain why many CFOs remain cautious about AI investments even as spending accelerates.
Source: Gartner – 2026 CFO Top Priorities (Dec 2025)
What this means for you: Metrics matter. If you’re measuring inputs (efficiency, productivity) instead of outputs (revenue, margin, customer value), you’ll never connect AI spend to business value. Finance and product teams need to collaborate on outcome-based metrics that tie AI investments to the P&L. Otherwise, you’re just tracking activity, not results.
The Hidden Cost of AI: The Rework Tax
AI tools are saving employees time—but organizations are failing to capture that value. A new productivity paradox is emerging that helps explain why AI ROI remains elusive.
85% of employees save 1-7 hours per week using AI
That sounds like a massive productivity gain. But here’s the problem: organizations aren’t converting that time into business value.
Source: Workday – Beyond Productivity: Measuring the Real Value of AI (Jan 14, 2026)
Nearly 40% of AI time savings are lost to rework
Employees spend those “saved” hours correcting errors, rewriting low-quality AI-generated content, and verifying outputs. Workday’s global survey of 3,200 employees found that for every 10 hours of efficiency gained through AI, nearly 4 hours are lost to fixing its output.
Only 14% of employees consistently achieve net-positive outcomes from AI use. The rest are caught in what Workday calls a “productivity paradox”—working faster but not actually achieving better results.
Source: Workday – Beyond Productivity: Measuring the Real Value of AI (Jan 14, 2026)
89% of organizations haven’t updated job roles for AI
Most organizations have introduced AI tools without redesigning roles, updating processes, or investing in workforce skills. Workday describes this as employees “using 2025 tools inside 2015 job structures.”
The burden isn’t evenly distributed. Employees aged 25-34 bear the heaviest load (46% of those dealing with the highest levels of AI correction), and the most frequent AI users spend the most time reviewing and correcting output—77% of daily AI users review AI-generated work just as carefully as they would their own.
Source: Workday – Beyond Productivity: Measuring the Real Value of AI (Jan 14, 2026)
Companies reinvest in technology (39%) over people (30%)
When organizations do capture AI time savings, they’re more likely to put those gains back into additional technology rather than employee development. And instead of using time saved to build skills, many simply increase workload (32%)—leaving employees to navigate AI on their own.
The disconnect between leadership priorities and employee experience is stark: 66% of leaders cite skills training as a top priority, but only 37% of employees experiencing the highest rework have access to that training.
Source: Workday – Beyond Productivity: Measuring the Real Value of AI (Jan 14, 2026)
What this means for you: The “rework tax” explains why your AI productivity metrics don’t translate to business outcomes. Employees might be working faster, but if 40% of that time goes to fixing AI mistakes, you haven’t actually gained productivity—you’ve just shifted where people spend their time. This is especially critical for Finance teams trying to measure AI ROI: speed ≠ value. The organizations seeing real returns don’t just deploy AI—they redesign roles, invest in skills, and modernize how work gets done so employees can actually use the time AI saves.
The CFO Dilemma: Cost Optimization vs. AI Investment
CFOs in 2026 face an impossible balancing act: they’re being asked to double AI spending while simultaneously making cost optimization their top priority.
56% of CFOs rank cost optimization as a top-5 priority for 2026
According to Gartner’s Q4 2025 survey of more than 200 finance chiefs, “achieving enterprise-wide cost optimization targets” is the #1 urgent action item over the next six months.
But here’s the tension: 47% also rank “allocating capital to new growth opportunities” in their top five priorities. CFOs are being pulled in opposite directions—cut costs and invest in growth simultaneously.
Source: Gartner – 2026 CFO Top Priorities (Dec 2025)
AI sits at the center of this tension
Some 39% of CFOs prioritize “accelerating AI use in the finance function” as a top-5 action item, while 33% prioritize “driving enterprise AI investment impact.” AI is clearly on the agenda—but it’s competing with cost optimization for resources and attention.
The problem? Only 36% of CFOs feel assured they can achieve meaningful AI outcomes. They’re being asked to invest billions while lacking confidence that those investments will deliver returns.
Source: Gartner – 2026 CFO Top Priorities (Dec 2025)
Dennis Gannon, VP analyst in Gartner’s Finance practice, describes this as a “very explicit tension around cost and growth goals” that “shows up every day” in conversations with finance chiefs.
What this means for you: This is the defining challenge for Finance leaders in 2026. You can’t optimize your way to growth, and you can’t invest your way to efficiency. The CFOs who succeed will be those who treat AI as a strategic capability that enables both—but only when it’s properly governed, measured, and tied to business outcomes. Cost optimization and AI investment aren’t competing priorities; they’re two sides of the same coin. AI should drive cost optimization through better forecasting, resource allocation, and waste elimination—not just add to the cost base.
AI Governance: Ownership Affects ROI Outcomes
Who manages AI matters as much as how you manage it.
Organizations with CIO/CTO + CFO jointly owning AI show higher realized ROI
Tech-only ownership correlates with weaker value capture. The best results come from cross-functional ownership that combines technical understanding with financial discipline and strategic alignment.
Source: Deloitte – C-Suite Tech Value and AI Governance Research (Dec 17, 2025)
Only 21% of organizations using gen AI have redesigned workflows
Those that do see up to 35% higher revenue growth and ~10% higher profit margins. This is the single strongest ROI driver according to McKinsey’s research.
Source: McKinsey – The State of AI: Global Survey 2025 (Nov 4, 2025)
Workflow redesign means fundamentally rethinking how work gets done—not just automating existing processes. Most organizations are bolting AI onto legacy workflows and wondering why they’re not seeing transformative results.
47-51% of organizations report at least one negative AI incident
This includes data privacy issues, security concerns, compliance problems, or accuracy issues. The rush to deploy AI without proper governance is creating real business risks.
Source: McKinsey – The State of AI: Global Survey 2025 (Nov 4, 2025)
What this means for you: Don’t let engineering run AI in isolation, and don’t let finance control it purely as a cost center. The best outcomes come from joint ownership that combines technical capability, financial discipline, and strategic alignment. And if you haven’t redesigned workflows to take advantage of AI capabilities, you’re leaving the biggest ROI opportunity on the table.
Project Success and Failure: The Pilot-to-Scale Problem
Getting AI projects into production remains a massive challenge.
30% of GenAI projects expected to be abandoned after proof of concept by end of 2025
Even more concerning: 40%+ of agentic AI projects are expected to be canceled by end of 2027. Primary causes include escalating costs, poor data quality, inadequate governance, and unclear business value.
Sources: Gartner – Predicts 2025: AI and Generative AI; Gartner – Top Trends in AI Governance
62% of organizations are experimenting with or deploying AI agents
But only 23% are actively scaling agents. Approximately 39% remain stuck in pilot phase, unable to move from experimentation to production deployment.
Source: RiskInfo.ai – AI Governance & Risk Update (Nov 17, 2025)
Among agentic AI leaders, 88% report ROI on at least one use case
More than 50% of these leaders report revenue gains of ~6-10% in growth-oriented use cases. Additionally, 52% of gen-AI users report deploying AI agents. This suggests that agentic AI can deliver strong returns—but only for organizations that figure out how to scale it.
Source: Google Cloud – The ROI of AI 2025 (Dec 21, 2025)
What this means for you: Pilot projects are not the problem—scaling is. The 30-40% abandonment rate reflects a fundamental inability to move from experimentation to production. Before you start another AI pilot, ask yourself: what’s different this time that will get this project into production? If you don’t have a clear answer around data quality, governance, and business value measurement, you’re just adding to the abandonment statistics.
AI Budget Allocation: Where the Money Is Going
AI is consuming a growing share of IT budgets, and that trend is accelerating.
Companies allocating ≥25% of IT budget to AI expected to rise from 27% → 52%
Companies allocating ≥50% of IT budget to AI expected to rise from 3% → 19%. This represents a massive shift in how enterprises are allocating technology resources.
Source: EY – AI-Driven Productivity and Investment Survey (Dec 9, 2025)
For context, this means that within a year or two, a majority of companies will be spending a quarter or more of their entire IT budget on AI. For large enterprises, that could represent hundreds of millions or even billions of dollars.
Investors expect ROI from embedded AI, not standalone tools
Durable enterprise AI ROI is concentrated in workflow-embedded and verticalized solutions. Standalone AI tools struggle to deliver sustained value according to venture capital analysis.
Source: Menlo Ventures – 2025: State of Generative AI in the Enterprise
This aligns with the workflow redesign finding: AI delivers value when it’s embedded into how work gets done, not when it sits as a separate tool that people have to remember to use.
What this means for you: If you’re spending 25-50% of your IT budget on AI, you need AI-specific financial controls and governance. You can’t manage that level of spend with the same processes you use for SaaS subscriptions or traditional cloud infrastructure. And if you’re building standalone AI tools rather than embedding AI into existing workflows, you’re likely investing in the wrong approach.
Practical Next Steps for 2026
Here’s how to translate these statistics into action for your organization:
For Finance Leaders (CFOs, Finance VPs)
1. Establish an AI cost taxonomy that separates training, inference, data platform, networking, and change-management spending. Make sure executive forecasts reference this taxonomy so everyone understands the full cost picture.
2. Shift to scenario-based forecasting. Stop using single-point estimates. Build explicit expected, committed, and stress scenarios with at least 10-15% buffer for unexpected spikes.
3. Demand outcome-based metrics. Don’t accept “productivity improvements” as ROI. Insist on metrics that tie to revenue, margin, or customer value. Work with product and engineering to define these metrics upfront.
4. Joint ownership with technology leaders. The data shows that tech-only or finance-only ownership produces worse outcomes. Establish joint accountability for AI investments between finance and technology leadership.
5. Address the rework tax. Don’t just measure time saved—measure actual business outcomes. If 40% of AI time savings are lost to rework, your productivity metrics are lying to you.
For Product and Engineering Leaders
1. Redesign workflows, don’t just automate existing ones. The 35% revenue growth and 10% margin improvement only come when you fundamentally rethink how work gets done.
2. Embed cost metrics into the ML lifecycle. Make cost a first-class metric alongside accuracy, latency, and throughput. Engineers should see cost impact in their development workflows, not just in monthly finance reports.
3. Set explicit waste budgets. Make the cost of experimentation intentional. Allocate specific budget for failed experiments, idle capacity, and learning. This removes the stigma from failure while keeping waste visible and controlled.
4. Focus on scaling, not pilots. Before starting a new AI project, define the path to production. What data quality, governance, and business value metrics need to be met before this moves from pilot to scaled deployment?
5. Update job structures for AI. 89% of organizations are asking employees to use 2025 tools in 2015 job structures. Redesign roles to reflect how AI changes work, not just how fast it gets done.
For FinOps Teams
1. Stand up AI-specific FinOps capabilities. AI workloads need different practices than traditional cloud workloads. Own GPU utilization targets, commitment strategy, and scenario forecasts specifically for AI.
2. Implement quarterly growth checkpoints. Don’t wait for monthly cloud bills to detect problems. Set up weekly or daily anomaly detection for AI workloads and quarterly reviews of growth against forecasts.
3. Build commitment strategies that preserve flexibility. Use the 60-80% baseline coverage model, commit to your predictable workload while preserving headroom for spikes and experimentation.
4. Create AI cost dashboards for engineering teams. Don’t just track costs—make them visible and actionable for the teams that control them. Show cost per model, cost per training run, cost per customer inference request.
For All Teams
Stop treating AI like just another cloud workload. The statistics are clear: traditional IT planning, traditional cloud FinOps, and traditional ROI metrics all fail when applied to AI. Organizations that recognize AI as its own domain, with unique cost patterns, unique governance needs, and unique success metrics, are the ones seeing actual returns on their investments.
The gap between AI leaders and laggards is widening. The leaders aren’t just spending more money, they’re forecasting differently, measuring differently, governing differently, and redesigning work differently. If you want to be in the 5% achieving significant ROI rather than the 60% seeing minimal value, you need to approach AI costs as a strategic discipline, not a tactical problem.
What’s Next?
These statistics paint a clear picture: AI investment is accelerating, but most organizations are struggling to show meaningful returns. The 80-85% who miss their forecasts, the 53% reporting only 1-5% ROI, the 30% abandoning projects after proof of concept—these aren’t isolated failures. They’re symptoms of a systemic problem.
The problem isn’t AI. It’s how we’re managing AI costs, measuring AI value, and governing AI investments.
The organizations that figure this out in 2026 will build sustainable competitive advantages. They’ll be able to continue investing in AI with confidence while their competitors face budget cuts and CFO skepticism. They’ll move from pilots to production while others add to the 30-40% abandonment rate. They’ll capture the 5× revenue increases and 3× cost reductions that leaders are achieving.
But it requires treating AI as its own FinOps domain—with specialized forecasting, dedicated cost management capabilities, cross-functional governance, and outcome-based success metrics.
The statistics show where the industry is today. The question is: where will your organization be at the end of 2026?
About Mavvrik
Mavvrik is the financial control platform for the AI era, helping organizations restore financial control by serving as the control layer between AI spend sources and business outcomes. We help finance, product, and FinOps teams understand, forecast, and optimize AI infrastructure costs.
Learn more about Mavvrik | Book a demo
Sources & References
This article synthesizes data from the following authoritative sources:
FAQ Section
These are optimized for Google AI Overviews, Perplexity, and ChatGPT citation pickup.
FAQ: AI Cost Statistics 2026
Q. Why do most companies miss AI cost forecasts?
A. Most enterprises miss AI cost forecasts because AI spending spans GPUs, cloud services, model inference, agents, and shared infrastructure. Traditional IT forecasting models fail to capture this complexity, leading 80–85% of organizations to miss forecasts by more than 25%.
Q. How much are companies spending on AI in 2026?
A. Companies plan to spend an average of 1.7% of revenue on AI in 2026, more than doubling 2025 levels. In many enterprises, AI is expected to consume 25–50% of total IT budgets within the next two years.
Q. What is the average ROI on AI investments?
A. While over 70% of organizations report “positive” AI ROI, fewer than 1% report significant ROI of 20% or more. Most companies see only 1–5% returns, often measured as productivity gains rather than financial impact.
Q. Why are so many AI projects abandoned after pilots?
A. Approximately 30% of generative AI projects are abandoned after proof of concept due to escalating costs, unclear business value, poor data quality, and lack of financial governance. Agentic AI projects face even higher failure rates.
Q. How should companies manage AI costs differently?
A. AI costs require dedicated financial governance, including AI-specific cost taxonomies, scenario-based forecasting, outcome-based ROI metrics, and joint ownership between finance and technology leaders.