Share this:
The State of Tokenomics 2026 report found that three in four enterprises cannot confidently prove AI business outcomes to the CFO. The biggest reason is not price. It is visibility. For IT Directors, the lesson is simple: AI spend runs on cloud services, rented GPUs, and hardware in your own data center, so you cannot explain it until on-prem, cloud, and AI costs sit in one view.
The report, released September 23, 2026 by the Tokenomics Foundation, surveyed 472 organizations across 11 industries representing $4.6 trillion in combined revenue. Its closing line sums up where most IT teams are right now: “Enterprises are not asking providers to charge less. They are asking for a bill they can explain to a CFO.”
Key takeaways
- Three in four enterprises cannot confidently prove AI outcomes to the CFO, and 43% name proving ROI as their single largest AI cost challenge.
- Enterprises want transparency, not discounts. 23% asked providers for more granular data. Only 4% asked for cheaper prices.
- AI spend is not just a token invoice. 31% of organizations run AI on rented hardware and 29% on private hardware.
- Ownership matters. Organizations with defined tokenomics ownership are 3.7x more likely to show value to the CFO.
- A complete FinOps practice covers on-prem, cloud, and AI. Leave one out and total IT spend stays unexplained.
What is the State of Tokenomics report?
The State of Tokenomics is an annual survey of how enterprises buy, measure, and govern AI spend. It is published by the Tokenomics Foundation, a Linux Foundation project announced at FinOps X 2026 to build open standards for measuring the cost, value, and return of AI.
Tokenomics is the practice of managing AI cost and value at the unit level: the input tokens a model consumes (prompts, retrieved context, system instructions) and the output tokens it produces (responses, tool calls, reasoning). Think of it as FinOps for AI. The 2026 edition drew responses from companies with a median revenue of $1.8 billion, so it reflects large enterprise reality, not startup experimentation.
What did the State of Tokenomics 2026 find?
The short version: AI adoption is broad, but accountability is not. Here are the findings that matter most for infrastructure and operations leaders.
| Finding | The number | What it means for IT Directors |
|---|---|---|
| Proving AI value | 3 in 4 cannot confidently prove AI outcomes to the CFO | AI budgets will face harder questions in 2027 planning |
| Top challenge | 43% say proving ROI; 27% say visibility and attribution of spend | You cannot prove a return on a cost you cannot fully see |
| What buyers want | 23% want transparency and granular data; 4% want lower prices | Clean, attributable cost data is now the priority |
| Ownership | Defined owners are 3.7x more likely to show value | Someone has to own AI cost, not just AI projects |
| Model routing | 86% are evaluating or using a model router; users are 4x more likely to show value | Routing decisions need cost data behind them |
| Where AI runs | 87% use cloud token providers; 31% rented hardware; 29% private hardware | AI cost lands on cloud bills and in your data center |
Why can’t most enterprises prove AI ROI?
Because the cost side of the equation is incomplete. Proving ROI (43%) and visibility and attribution of spend (27%) were the top two challenges in the survey, and they are really the same problem. If you cannot see what an AI workload costs end to end, you cannot tie it to a business outcome.
Governance is also early. The report notes that most organizations still rely on budget caps and token limits as their primary controls, and focus on monitoring rather than measuring return. Caps keep spend from running away. They do not tell the CFO whether the spend was worth it.
This is the same gap many IT teams already face with infrastructure. We covered the value side in AI Value Mapping: Moving From Adoption to Accountability. The State of Tokenomics data shows the cost side needs the same discipline.
Where does enterprise AI spend actually live?
Everywhere. The report shows AI spend spread across at least six procurement channels, and several of them never appear on an AI provider invoice.
- 96% use frontier model providers directly
- 87% buy tokens through cloud platforms such as AWS Bedrock, Google Vertex, or Azure Foundry
- 64% pay for AI embedded in other tools
- 32% run AI at the edge or on local devices
- 31% use rented hardware
- 29% run AI on private hardware
That last group is where infrastructure teams come in. When AI runs on private hardware, its cost shows up as GPU servers, storage for training data and retrieval, network capacity, power, and cooling. 38% of respondents are already considering energy consumption in their tokenomics calculations. None of that is on a token bill.
The trend points toward more of it, not less. 51% of respondents describe themselves as heavy frontier model users today, but only 24% expect to be there next year. As organizations move toward open-weight models, more of them will run those models on rented or owned infrastructure, which moves more AI cost into the data center and cloud infrastructure lines IT Directors already manage.
Why does Hybrid FinOps™ need on-prem, cloud, and AI?
Because total IT spend is the sum of all three, and each one hides costs the others need. A complete Hybrid FinOps™ practice tracks:
- On-prem infrastructure: storage, compute, VMware, backup, and the hardware AI workloads increasingly run on.
- Cloud: IaaS, PaaS, and the cloud platforms where most enterprises buy AI tokens.
- AI: tokens, model subscriptions, embedded AI tools, and the infrastructure allocated to AI workloads.
Here is how the gap shows up in practice. Say your team buys a GPU cluster to run an open-weight model in the data center. If AI cost tracking only counts tokens, that project looks nearly free. If on-prem tracking only counts hardware, nobody connects the cluster to the AI outcome it supports. Either way, the CFO gets half a story.
Cloud-only FinOps has the same blind spot. We made that case in On-Prem FinOps: Why It Matters for Your Organization, and AI makes it more urgent. Without all three, IT Directors do not have a clear picture of their overall IT spend.
What should IT Directors do with these findings?
Start treating AI as a cost category you manage alongside infrastructure, not a separate budget someone else explains. Five steps you can take before 2027 budgets lock:
- Name an owner. 12% of respondents have no defined tokenomics owner and 26% share it across functions. Defined owners are 3.7x more likely to show value.
- Map every place AI runs. Include cloud token platforms, embedded tools, rented GPUs, and private hardware.
- Allocate shared infrastructure. Assign the storage, compute, and power AI workloads consume, the same way you would with showback or chargeback.
- Standardize the data. 7% of respondents named FOCUS, the open billing data standard, without being asked. Consistent cost data makes on-prem, cloud, and AI comparable.
- Report cost per outcome. Tie spend to a business unit, application, or result so the CFO sees return, not just a total.
Where Visual One Intelligence® fits
Visual One Intelligence® was recognized by the FinOps Foundation as a FinOps Certified Platform for hybrid infrastructure. Its Hybrid FinOps™ platform brings on-prem storage, compute, VMware, and cloud into one cost and capacity model, including the infrastructure your AI workloads run on.
That gives IT Directors a piece of the picture the State of Tokenomics says most enterprises are missing: attributable cost data for the infrastructure AI runs on, in the same view as the rest of the estate. Pair it with your token and subscription spend, and you can show Finance the full cost of AI and defend the number. Price every asset, surface every risk, defend every number.
Frequently asked questions
What is tokenomics in AI?
Tokenomics is the practice of managing the cost and value of AI at the unit level. It measures the input tokens a model consumes and the output tokens it produces, then connects that spend to business outcomes. It applies FinOps principles such as visibility, allocation, and optimization to AI workloads.
Who publishes the State of Tokenomics report?
The Tokenomics Foundation, a Linux Foundation project announced at FinOps X 2026. The 2026 report, released September 23, 2026, surveyed 472 organizations across 11 industries representing $4.6 trillion in combined revenue.
What is the biggest AI cost challenge for enterprises in 2026?
Proving value. 43% of State of Tokenomics respondents named proving ROI as their single largest challenge, followed by visibility and attribution of spend at 27%. Three in four enterprises cannot confidently prove AI business outcomes to the CFO.
Does AI spend include on-prem infrastructure?
Often, yes. The State of Tokenomics found that 29% of organizations run AI on private hardware and 31% on rented hardware. Those workloads consume servers, GPUs, storage, network, power, and cooling, none of which appear on a token invoice. Accurate AI cost tracking has to include that infrastructure.
What is Hybrid FinOps?
Hybrid FinOps is a financial operations practice that tracks on-prem infrastructure, cloud, and AI spend together. Without all three, IT leaders cannot see total IT spend or explain it to Finance. Visual One Intelligence® provides a FinOps Certified Platform built for hybrid infrastructure.
See your whole IT bill in one view
AI is now part of your infrastructure budget, whether it runs in the cloud or in your data center. Schedule a demo and see on-prem, cloud, and the infrastructure behind your AI workloads in one place.
