This new foundation is a vendor‑neutral standards body charged with answering a deceptively simple question that has become a board‑level obsession: What does AI actually cost, and is it worth it?
Many AI companies throw money at AI like there’s no tomorrow. Businesses that use AI, however, have grown far more cautious. The days of tokenmaxing are history. Instead, they’re looking for ways to get a grip on AI token pricing. That’s where the Linux Foundation’s newest organization, the Tokenomics Foundation, comes in.
Formally established on Aug. 4 with 30 founding members, the Tokenomics Foundation aims to create open frameworks, specifications, and best practices for measuring the cost, value, and return on AI spend across clouds, models, and vendors.
Hosted by the Linux Foundation but governed independently, the group brings together large enterprise buyers such as JPMorgan Chase, BNY, GoDaddy, Hitachi, and Lenovo alongside infrastructure vendors and cost‑optimization players including Oracle, SAP, ServiceNow, Broadcom, Accenture, IBM, Cast AI, and Flexera.
In a statement, Tokenomics Foundation executive director J.R. Storment explained the new foundation’s mission: “Businesses are reinventing how they deliver value with AI faster than they can measure it. Every model release changes the math on cost, consumption, and ROI. Tokens are only the visible tip. The real total cost of AI spans compute, storage, data, and the people who build with it. Every CEO is being asked to show returns on all of that without a shared way to count it. That is why this Foundation exists and is working together on pre-competitive frameworks, benchmarks, and specifications, built in the open, so the entire global economy can accelerate value from AI rather than just account for the spend.”
Under the banner of “tokenomics,” the foundation is taking direct aim at the opaque billing models and fragmented dashboards that currently govern AI usage. In an interview with Fortune, Storment said, “Tokenomics Foundation arrives at a defining moment for the global technology economy when every company in the world is struggling to quantify the value of AI. While per-token costs fell heavily during 2023-2025, they have leveled off—and new model token prices are rising—making AI costs the largest and fastest-growing line item on enterprise technology budgets…This has made tokenomics a CEO-level concern, and organizations are looking for alignment on industry best practices and standards for AI ROI.”
So what are tokens anyway? At FinOps X 2026, Storment calls them “the atomic unit of AI.” In his keynote, Storment said that “tokens serve more roles in the modern economy than almost any other commodity has in modern history, maybe, maybe oil in the 20th century.” Tokens, he told the audience, are simultaneously “the unit of output from all of the hardware and compute and data centers,” “how the labs price their outputs and inputs,” and “the value unit that enterprises are looking to monetize.”
In its first draft output, “Big‑T Notation,” the Tokenomics Foundation introduces a structured way to classify tokens and AI workload complexity ahead of routing traffic to different models. That’s only the start.
Tokenomics’s plans are ambitious. They include forming shared definitions and vocabulary. This will provide a formal definition of tokenomics in the AI context (distinct from its crypto usage), plus terms such as token value, token density, and the distinctions between input, output, reasoning, and cached token types.
The Foundation is also working on a full cost‑of‑AI reference model. This will include a nine‑layer cost stack that places token charges in context with compute, storage, data, networking, SaaS embedding, engineering labor, training costs, “shadow AI,” and tokens themselves.
They’re also considering a shift from “cost per token” to “cost per API call.” This will tie economics to a unit of work rather than raw compute consumption. In addition, they’re working on methodologies for relating AI spend to outcomes, starting with the share of work completed without human involvement, benchmarked against what the same process costs today.
Speaking of ambition, the foundation expects “nearly monthly” releases of frameworks and metric definitions through the end of 2026, backed by a governing board that convened July 30 and a technical steering committee now being formed. Its first in‑person gathering will run as “Tokenomicon + FinOps X Amsterdam” in September, with a flagship annual conference slated for June 2027 in San Diego, following in the footsteps of the FinOps conference.
It’s a good thing then that the Tokenomics Foundation is not starting from a blank slate. It is designed to sit alongside the FinOps Foundation. This is another Linux Foundation project that has become the de facto home for cloud financial management. These new open AI standards are also intended to plug into existing open specifications such as the FinOps Open Cost and Usage Specification (FOCUS).
FinOps has already extended its framework to cover AI and GPU spending, publishing “FinOps for AI” guidance that borrows from cloud disciplines: tracking AI costs, tagging workloads, setting budgets, assigning ownership, and optimizing GPU allocation. FOCUS, meanwhile, normalizes cloud cost and usage data across providers and is now being extended to account for AI spending and token‑based consumption.
Where FinOps focuses on cloud resources and financial operations in general, Tokenomics narrows in, for now, on tokens as the atomic unit of AI value.
The difference is that while FinOps for AI frameworks addresses a broad landscape of GPU, storage, and model costs, Tokenomics centers on token production, distribution, and consumption, treating tokens as the basic unit that connects energy and capital to business outcomes. Moreover, FinOps produces lifecycle guidance, practices, and training; Tokenomics is chartered to ship formal taxonomies, metrics (such as cost‑to‑serve), and AI Value Frameworks that can be implemented by vendors and enterprises alike.
When all is said and done, FOCUS and Tokenomics are expected to be tightly coupled. FOCUS will define how cost and usage data is represented, while Tokenomics defines what needs to be measured and compared for AI ROI.
These groups aren’t the only ones dealing with these issues. Many vendor frameworks for “AI cost management” and “AI cost governance” are trying to solve similar problems at a tooling level. Playbooks from Flexera, CloudZero, and others detail how to inventory AI tools, attribute costs to teams or products, and enforce budgets and guardrails for token consumption.
These efforts, however, are inherently product‑centric. The Tokenomics Foundation’s pitch is that enterprises need a neutral, open standard that can be implemented across competing platforms. This is a role closer to accounting management than to a single vendor’s dashboard.
Nishant Gupta, Salesforce’s Chief Availability Officer, framed the challenge starkly in a statement, “Token economics is fundamentally more abstract and more opaque than anything we’ve managed at this scale before. Input versus output tokens, cached versus non-cached, pricing structures that don’t behave like compute or storage. Yet at the same time, executives are being asked to make multi‑year commitments to AI platforms without a reliable way to compare self‑hosted, cloud‑hosted, and SaaS LLM options on an apples‑to‑apples basis.”
This is a mess in the making. The Tokenomics Foundation thinks the industry needs something more precise. That is a Generally Accepted Accounting Principles (GAAP) layer for AI that says what counts, how to count it, and how to compare value across an increasingly fragmented model landscape.
Good luck, folks. Whether enterprises adopt its metrics as the canonical way to talk about AI ROI may determine if “tokenomics” becomes a core competency of finance and product teams or remains another vendor buzzword in an already noisy AI cost‑management market.
Usually, I’d bet on the Linux Foundation and open standards to win the day. However, I can’t help but notice those who aren’t members: Anthropic, OpenAI, Meta, Google, and Microsoft, the AI vendor giants. We’ll see how far this initiative can get without their support.
