Nobody Knows Where Their AI Budget Is Going and Gartner Says It Will Hit 2.59 Trillion
Here’s a stat that should make every CFO lose sleep. Gartner is predicting that global AI spending will hit $2.59 trillion by 2028. Not billion. Trillion. With a T. And yet if you walk into most boardrooms and ask “where exactly is our AI budget going,” you’ll get a lot of blank stares and nervous throat-clearing.
That’s the paradox nobody wants to talk about. Companies are pouring money into AI faster than ever before, but most of them have zero visibility into what they’re actually getting for that spend. Token costs are spiraling. API calls add up in ways nobody predicted. And teams are running experiments that burn through budget before anyone thinks to ask whether the experiment is even working.
I’ve seen this firsthand. A mid-size SaaS company I consulted with last year had three different departments each running their own AI tools. Marketing was paying for one LLM provider. Product was using another. Customer support had quietly signed up for a third. Nobody was talking to each other. Nobody was tracking consolidated spend. Their total AI bill was north of $400,000 a year, and they couldn’t tell me which tool produced a single dollar of ROI.
The Token Tax Nobody Budgets For
Here’s what makes AI budgeting so tricky. Unlike traditional software where you pay a fixed monthly or annual license, AI costs are usage-based. Every prompt, every completion, every agent interaction burns tokens. And tokens cost money. Lots of it, when you multiply across thousands of users or millions of API calls.
The problem gets worse when you look at AI agent projects. Gartner also predicts that 40% of AI agent projects will be canceled by 2027. Think about what that means. Almost half the AI agent initiatives companies are funding right now won’t survive. That’s not a technology problem. That’s a spending discipline problem.
Most teams launch AI agents with big ambitions but no budget guardrails. They don’t set token limits. They don’t track cost-per-interaction. They don’t define what “success” looks like before they start burning through their allocation. And when the bill comes due, leadership panics and pulls the plug entirely instead of optimizing what’s actually working.
FinOps Is the Answer Nobody Wants to Hear
If you’re serious about getting your AI budget under control, you need to look at what the cloud computing world figured out years ago. The FinOps Foundation built an entire discipline around cloud cost management, and the exact same principles apply to AI spend.
FinOps isn’t glamorous. Nobody puts “implemented cloud cost tagging” on their resume. But it works. The core idea is simple. You need real-time visibility into where money is going. You need accountability at the team level. And you need automated alerts when spending crosses thresholds you’ve pre-defined.
Applied to AI, this means tagging every API call by project, team, and use case. It means setting hard budgets per department. It means reviewing token consumption weekly, not quarterly. And it means making people actually justify their AI spend instead of hand-waving about “innovation.”
Hot take? Most companies don’t have an AI problem. They have a transparency problem. They’re spending millions on something they can’t see, can’t measure, and can’t optimize. That’s not a technology strategy. That’s financial negligence.
What Good AI Cost Management Actually Looks Like
Let me paint a picture of what this looks like when it’s done right. First, you centralize all AI spending into a single dashboard. Every provider, every team, every project. If it’s not on the dashboard, it doesn’t exist. Period.
Second, you assign a cost center to every AI initiative. Marketing’s chatbot. Sales’ lead scoring model. Engineering’s code assistant. Each one gets its own budget line. Each one gets reviewed monthly. And each one has to prove its worth or get cut.
Third, you implement automated cost controls. Set maximum monthly spend per tool. Set alerts at 75% and 90% of budget. Use provider-level rate limiting where available. Don’t let a single rogue integration blow through your entire quarter’s allocation in a weekend.
Fourth, and this is the one most companies skip, you track the business outcome per dollar spent. Not just “we used 2 million tokens.” But “those tokens generated 500 qualified leads at a cost of $12 per lead.” Without that connection between AI spend and business results, you’re just burning cash and hoping something sticks.
The People Side of the AI Budget Problem
Here’s something nobody wants to admit. The AI budget problem is really a people problem. It’s about who has authority to spend, who’s accountable for results, and whether your workplace culture supports honest conversations about what’s working and what isn’t.
Too many companies treat AI spending like a lottery ticket. They throw money at it and hope for a big win. But responsible AI adoption means having the discipline to say “this isn’t working, let’s stop” just as much as it means saying “this is great, let’s scale it.”
And that brings up an uncomfortable question about how we value AI contributions. When an AI tool saves 200 hours of manual work per month, what’s that worth to your business? If you can’t answer that question, you can’t justify the spend. We wrote more about this exact tension in our piece on AI pay and what companies actually owe the people building these systems.
Five Practical Steps Starting This Week
Stop waiting for the perfect AI governance framework. Start with these five things right now.
Audit everything. Pull every AI-related invoice, subscription, and API bill from the last 90 days. You’ll probably be shocked at how much is scattered across credit cards and expense reports that nobody reviews.
Tag and categorize. Assign each expense to a team, project, and use case. Use a spreadsheet if you have to. The tool doesn’t matter. The visibility does.
Set budgets with teeth. Not soft guidelines that people ignore. Hard limits with automated enforcement. If a team needs more, they submit a business case with projected ROI.
Review monthly. Put AI cost reviews on the same cadence as financial reviews. Make it a standing agenda item. Make it someone’s job to care about this.
Measure outcomes. For every dollar spent on AI, track the business result it produced. Revenue influenced. Hours saved. Errors reduced. If you can’t measure it, question whether you should be spending on it.
The Bottom Line
The $2.59 trillion number from Gartner isn’t just a forecast. It’s a warning. Companies are about to spend more on AI than most nations spend on education. And a huge chunk of that spending is happening with no oversight, no measurement, and no accountability.
The companies that figure out AI cost management now will have a massive advantage. They’ll be able to scale what works, kill what doesn’t, and make smart bets with real data behind them. The companies that don’t? They’ll be the ones canceling 40% of their AI projects in two years wondering where it all went wrong.
Don’t be that company. Get your AI budget in order before it gets away from you. For more sharp takes on business strategy in the age of AI, check out The Business Series.
