San Diego, Day 1
FinOps X opened on June 8 in San Diego with 2,500 practitioners in the room and a keynote that did something unusual for a cloud conference: it named a discipline that didn't have a name before.
Token Economics. Tokenomics. The idea that AI tokens are the new atomic unit of technology spend, and that understanding how energy and capital convert into tokens, and how efficiently those tokens generate business value, is a distinct professional discipline that FinOps practitioners need to own.
I wasn't in the room. I've been following the keynote closely from the published recap and the session recordings. What follows is my read on what Day 1 actually means for practitioners — the announcements that matter, the ones that sound important but aren't, and the uncomfortable question that nobody at the conference seemed to ask directly.
The announcement that will matter in five years: the Tokenomics Foundation
J.R. Storment opened by framing the challenge simply. Tokens are how AI consumes resources. Token costs are variable, hard to forecast, and growing faster than any other category of technology spend. The organisations that figure out how to govern token consumption efficiently will have a structural advantage over those that don't. And right now nobody has a standard way to even measure it.
The FinOps Foundation and Linux Foundation announced the intent to form the Tokenomics Foundation. Early supporting organisations include Oracle, Google, Microsoft, Accenture, Booking.com, Flexera, IBM, JPMorganChase, KPMG, Salesforce, SAP, and ServiceNow.
The ambition is clear: open standards for AI billing, the same way FOCUS created a common billing schema across cloud providers.
The honest read on this: it will matter a great deal, but not immediately. Foundation announcements take time. FOCUS 1.0 was years in development before it was practically useful to practitioners. The Tokenomics Foundation is the right call at the right moment, but most practitioners won't feel its impact for two to three years at minimum.
What it signals right now is more important than what it delivers. The industry is acknowledging that AI billing is broken in the same way cloud billing was broken in 2012. Costs are invisible, attribution is inconsistent across providers, and there are no shared metrics for measuring whether token consumption is efficient. The Foundation is the first formal attempt to fix that at an industry level rather than a vendor level.
If you care about the long arc of this discipline, the Tokenomics Foundation is the most significant announcement from Day 1. It just won't change your Monday morning much.
The announcement that will change your Monday morning: the AWS FinOps Agent
Bradford Lyman, Director of Product Management at AWS, came to the keynote with a list of new features. Some of them were incremental — more idle resource recommendations, credit transparency improvements, extended Bedrock attribution (which I covered in a separate post earlier this month). All useful. None of them are the story.
The story is the AWS FinOps Agent.
Natural language queries to surface savings opportunities. Automatic distribution of those recommendations to the right teams. Autonomous execution of approved optimisation actions at scale.
Let me say that more plainly. AWS has built a product that can identify savings in your environment, route the recommendations to the right people, and in certain conditions execute the changes without a human manually doing it.
This is not Amazon Q with a FinOps filter. This is a different category of product. The distinction between "AI that surfaces information" and "AI that takes action" is the most significant line in cloud cost management right now, and AWS just announced they've crossed it.
The details of what actions it can and cannot take autonomously are still being fleshed out from the session recording. Savings Plans purchasing decisions appear to be in the "surface and recommend" category rather than fully autonomous at this stage. Rightsizing recommendations for development environments appear to have an autonomous execution mode with guardrails. I'll cover this more thoroughly as more session detail becomes available.
What matters for now is the direction. AWS is building toward a world where routine cost optimisation actions — the kind that currently require a FinOps practitioner to review a report, identify the recommendation, notify an account owner, and track the outcome — are handled autonomously. That is not a future state speculation. It is a product announcement at the world's largest FinOps conference.
The other AWS features worth noting
Beyond the Agent, Bradford's announcement list included several features that deserve a closer look.
Target Coverage for Savings Plans. AWS will now let you set a target coverage percentage and track toward it directly in Cost Explorer rather than manually calculating your current coverage and comparing it to a spreadsheet. This sounds minor. For teams that do quarterly Savings Plans reviews, it compresses what is currently a multi-step analysis into a configured target with visual tracking. Worth setting up.
Automatic Cost and Forecast Explanations. Cost Explorer will now automatically generate written explanations for significant month-over-month cost changes, similar to what Amazon Q does on request but without requiring the user to ask. If your Bedrock spend spikes, you get an explanation without going to type a query. This is the Cost Comparison feature that launched last year, extended with AI-generated narrative on top.
Credit Level Sharing and Improved Credit Transparency. Credits applied by AWS — refunds, promotional credits, enterprise discount programme credits — have historically been difficult to trace at the team or account level. The new credit features improve attribution and sharing controls. For organisations with complex credit positions, this reduces a significant manual reconciliation burden.
What the practitioners said
Three enterprise practitioner stories came through the keynote and each landed a line worth keeping.
Frederik Pohl from SAP, running FinOps for AI at global scale: the metrics and governance structures from traditional cloud FinOps do not directly translate to AI workloads. You cannot apply EC2 rightsizing thinking to token consumption. The inputs, the variability, and the attribution challenges are different in kind, not just in scale.
Pooja Kumar from Prudential: "Traditional FinOps is now table stakes." She framed the shift as needing to "shift wild" rather than just "shift left" — the discipline needs to move beyond its current operating model into something more strategically integrated with how organisations decide to invest in AI. Her framing was direct in a way that most keynote presentations aren't. If your FinOps practice is still primarily focused on tagging governance and Savings Plans optimisation, you're doing necessary work. You're not doing sufficient work.
Courtney Totten from Shutterstock: applied FinOps to AI spend, shared the actual decisions and trade-offs, and explicitly said the journey involved a lot of things that didn't work before finding the ones that did. Honest practitioner storytelling is rare at conferences. When a CIO/CISO gets on stage and talks about what didn't work, that's worth more than a polished success story.
The question nobody asked directly
Here it is.
If the AWS FinOps Agent can autonomously surface savings, route recommendations, and in some cases execute changes — what does that mean for the FinOps practitioner's role?
The conference theme was "AI Value: The Era of FinOps for AI, Token Economics, and Agentic FinOps." The industry is deliberately moving toward agentic FinOps as the next model. AI systems managing cloud costs, not just surfacing them.
Pooja Kumar's "traditional FinOps is table stakes" line pointed directly at this without naming it. If routine cost identification and optimisation becomes automated, the practitioner's value has to come from somewhere else. Strategic alignment. Business value measurement. Governance design. The human judgment that an autonomous agent cannot replicate.
I don't think the practitioner role disappears. I think it changes faster than most practitioners are preparing for. The FinOps practitioners who will be most valuable in two to three years are the ones who can operate at the level of AI governance and business value attribution — the Tokenomics layer — not the ones who are fastest at reading a Cost Explorer dashboard.
The Tokenomics Foundation and the AWS FinOps Agent are two announcements from the same day that together describe the same future. The discipline is being redefined from two directions simultaneously. Upward into strategy and AI value measurement. Downward into autonomous execution.
Practitioners who understand what's happening have time to position accordingly. That's the real takeaway from Day 1.
What to watch on Day 2 and beyond
The Day 2 keynote livestream starts at 8:20am PT on June 10. Based on what Day 1 established, the sessions worth tracking:
Any session that goes deeper on the AWS FinOps Agent — specifically what it can and cannot execute autonomously, what the guardrail model looks like, and how practitioners retain oversight of autonomous actions.
FOCUS standard updates. The FinOps Foundation has maintainers at the conference for direct Q&A. Any changes to the FOCUS schema that touch AI cost attribution or token-level billing are worth following closely.
The new FinOps certification path. FinOps Certified: Technology Value launched at Day 1, and the path to FinOps Professional was restructured. The certification changes tell you something about where the Foundation thinks the required skillset is heading.
I'll publish a follow-up once Day 2 wraps.
Following FinOps X and working through what the shift to AI cost governance means for your practice? Get in touch — the Bedrock attribution and token economics questions are the ones most teams are trying to get ahead of right now.