Same Math, New Meter: Why AI Tokens Are Cloud Bills 2.0
Yep, AI smells a lot like the cloud.
I’ve been in technology since 1982 — Fortran programmer, military communications specialist, Cisco engineer, InfoSec product specialist, and for the last 28 years, the guy standing in front of a whiteboard telling executives what a new platform will actually cost them. So when I hear a vendor promise that AI will optimize spend, cut complexity, and shrink headcount, I get déjà vu. I heard the same pitch about the cloud fifteen years ago. And I watched it, and it didn’t entirely deliver.
That’s not cynicism. It’s pattern recognition. And the pattern is worth naming before your organization signs another “unlimited” AI seat license.
We’ve Heard This Promise Before
Cloud’s original pitch was simple: stop buying servers, stop hiring people to babysit them, and watch your IT line item shrink. For many workloads, in the early years, that was true. Then the scale happened.
Andreessen Horowitz called this out directly in their now-famous cloud economics analysis, describing how cloud clearly pays off early in a company’s life but starts eating margin as usage scales — a shift so gradual that few finance teams see it coming. Their prescribed fix, once the bill gets bad enough, is often full repatriation off the cloud entirely — the exact opposite of the original promise.
I wrote about this same dynamic from the security side back in 2022, watching clients deploy “the latest, most significant, automated” cloud security stack with a managed-services bow on top, only to see SecOps costs blow past the predetermined budget number every single time. The tools didn’t fail because they were bad tools. They failed because “optimize and reduce” was never actually engineered into the deployment — it was a slide in the sales deck.
The Snake Oil Playbook, Rebottled for AI
Now swap “cloud” for “AI” and watch the same slide reappear. Do more with less. Automate the headcount. Optimize the spend.
Fortune’s reporting on the current AI budget crunch is blunt about where that pitch has landed inside real companies: internal leaderboards with names like “Claudeonomics” tracking who burns the most AI tokens, and employees at Amazon openly encouraged to “toxenmaxx” — use as many tokens as possible, cost be damned. Goldman Sachs is now forecasting a 24-fold increase in enterprise token consumption by 2030. That is not an optimization curve. That is the multi-cloud sprawl story with a chatbot interface.
Arvind Jain, CEO of Glean, told CNBC something I don’t think enterprise boards have fully absorbed yet: this is the first time technology has cost roughly what people cost, which means the “AI vs. headcount” conversation isn’t hypothetical anymore — it’s a live budget line CFOs are staring at this quarter.
The Complexity Tax Cloud Never Paid Off
Here’s the part I lived through personally, and the part vendors never lead with: cloud didn’t just fail to shrink complexity — it multiplied it. Zero trust running inside SASE, eleven MSSPs covering twenty cloud apps, CASB stacked on top of DLP stacked on top of MFA. I described this exact spiral in Cybersecurity is a Successfully Failure — a system with a hundred adaptive controls that still gets breached, because more layers mean more attack surface, not less.
AI is already adding its own version of that layer, and healthcare is where it shows up first. Coeus has flagged “Shadow AI” — unauthorized use of AI tools by staff — as a leading cause of data exposure and HIPAA compliance failures in community health organizations right now. Nobody budgeted for that complexity either. It just showed up, the same way the eleventh MSSP always does.
AI Cost Management Is Repeating Cloud’s Biggest Mistake
The mechanics are almost identical to cloud sprawl, just faster. The FinOps Foundation’s 2026 State of FinOps report found that 73% of enterprises say their AI costs have blown past original projections — even as the per-token price has actually been falling. Total spend is price times volume, and volume is the variable nobody modeled.
OpenAI’s own head of enterprise, Alexander Embiricos, told TechCrunch the customer conversation has completely changed in six months — it’s no longer about capability, it’s about visibility, auditability, and token controls. That quote could have been lifted from a 2016 FinOps conference about cloud sprawl. Which is exactly why the Linux Foundation just launched something called the Tokenomics Foundation — a standards body built to bring the same cost discipline to AI tokens that FinOps eventually brought to cloud, years after the bills got out of control.
What Coeus Is Doing Differently
This is why Coeus Consulting isn’t chasing an “AI-first” headline. We’re treating AI adoption the way we wish more organizations had treated cloud migration a decade ago: as an infrastructure, security, and governance decision first, and a tool decision second.
That’s the thinking behind our new alliance with Hummingbird Advisory Partners to guide Phoenix-area healthcare organizations through AI adoption — deliberately starting in one of the most regulated, highest-consequence verticals we serve, as the first phase of a much deeper research effort into responsible AI governance across every industry we support. Hummingbird’s Managing Principal, Curt Schatz, put it exactly right: AI adoption in healthcare gets sold as a software decision, but it’s really an operating, security, and infrastructure decision happening at the same time. That’s the governance conversation the token bill eventually forces anyway — we’d rather have it before the invoice arrives, not after.
The Bottom Line — Optimize Before You Adopt
Cloud wasn’t a bad idea. It was a good idea sold without the complexity, cost, and governance conversation attached — and enterprises paid for that omission in SecOps overtime and surprise invoices for a decade. AI is following the identical script, just compressed into months instead of years.
The organizations that get ahead of it will be the ones that build cost visibility, model routing, and security governance into the plan before the rollout — not after the first token bill lands on the CFO’s desk. If you want that conversation before your Phase 2 renewal, your next audit, or your next AI budget review, talk to Coeus.
Frequently Asked Questions
Is AI really more expensive than cloud computing was?
It’s following a similar trajectory. Per-unit prices for both cloud compute and AI tokens have fallen over time, but total spend keeps rising because usage volume grows faster than prices drop — the FinOps Foundation found 73% of enterprises exceeded their 2026 AI budget projections despite falling per-token costs.
Why are AI token costs rising if the price per token is falling?
Total spend is price multiplied by volume. Agentic AI, background monitoring agents, and “always-on” workflows consume tokens continuously and multiply geometrically, so volume is outpacing the price drops — the same dynamic that made cloud bills balloon even as per-instance pricing fell.
What is “shadow AI” and why does it matter for compliance?
Shadow AI is the unauthorized use of AI tools by staff outside of IT-approved, governed systems. In regulated industries like healthcare, it’s become a leading cause of data exposure and compliance failures because sensitive data can end up inside ungoverned AI tools without anyone realizing it.
How is Coeus Consulting approaching AI adoption differently?
Coeus treats AI adoption as a security and governance decision first, not just a software purchase. Its alliance with Hummingbird Advisory Partners is the first phase of a broader initiative to build responsible AI governance frameworks, starting in healthcare — one of the highest-stakes, most regulated environments Coeus serves.
What can businesses do now to avoid an AI cost blowout?
Build cost visibility and model-routing governance before scaling AI usage — not after the bill arrives. That means tracking token consumption by team and use case, routing simple tasks to cheaper models, and treating AI security and compliance as part of the initial rollout plan rather than an afterthought.
About Coeus Consulting
Coeus Consulting is a Phoenix-based managed IT, cybersecurity, cloud, and compliance provider serving healthcare, aerospace & defense, construction, automotive, and legal organizations across Arizona, California, Nevada, and the broader Southwest. Coeus is BBB A+ rated, holds a 4.9★ Google rating, and was a 2025 Southwest MSP Titans of the Industry finalist. Through its alliance with Hummingbird Advisory Partners, Coeus helps regulated organizations adopt AI without trading security and compliance for speed. Learn more at coe.us.
About the Author
John Gormally, MBA, is the Digital Marketing Coordinator for Coeus Consulting and a 28-year cybersecurity professional. A U.S. Marine Corps veteran, John holds an MBA and co-authored the graduate cybersecurity curriculum at Cal State San Marcos. He writes and speaks frequently on cybersecurity, cloud, and AI governance trends, and has presented at ISSA, ISACA, and FBI InfraGard events nationwide. Connect with him on LinkedIn or read more of his writing on Medium.