FinOps for AI: what FOCUS 1.4 gives you now, and what 1.5 adds in December

Abstract cover: streams of amber and blue particles flowing into translucent ledger rows on a deep navy background

Almost every FinOps team now manages AI spend, but most still cannot answer three basic questions about it: which model, which kind of tokens, and who called it. The FinOps Foundation’s FOCUS specification is about to make those answers part of the standard billing record. Here is what FOCUS 1.4 gives you today, what 1.5 is set to add in December, and what to do before it lands.

Why this matters now

According to the FinOps Foundation’s State of FinOps 2026, 98 percent of respondents now manage AI spend, up from 63 percent in 2025 and 31 percent in 2024. FinOps for AI is the top forward-looking priority for the next twelve months, even though workload optimization and waste reduction remain the top current priority.

The problem is the data. AI charges arrive from cloud providers, direct model API vendors and AI features inside SaaS products, each in its own format and units. FOCUS, the FinOps Open Cost and Usage Specification, exists to put all of that into one schema with one set of column definitions, so a cost report does not need a different pipeline for every provider.

Comparison of FOCUS 1.4, ratified June 4, 2026, with invoice and commitment features, and FOCUS 1.5, targeted for December 3, 2026, with AI model identity, token labels and requester attribution
FOCUS 1.4 is available now; the 1.5 scope is as listed by the FOCUS working group in October 2026.

What changed: FOCUS 1.4 is available today

The FOCUS Steering Committee ratified version 1.4 on June 4, 2026. It adds 2 datasets, 47 columns, 6 attributes, 17 glossary entries and 2 supported features, with no incompatible changes. Its theme is finance, not AI: for the first time, FinOps and accounts payable can work from the same billing facts.

What 1.4 addsWhat it lets you do
Invoice Detail datasetSee charges as they appear on the issued invoice, including payment currency, terms, due date and PO number
Billing Period datasetCompare periods correctly when a provider’s billing period is not a calendar month
Invoice Reconciliation (supported feature)Join invoice lines to cost and usage rows on Invoice ID, so usage and invoices tie out
Contract Commitment expansion (13 to 30 columns)Compare reservations, savings plans and other commitments across providers
Commitment Program Eligibility DetailsSee which commitment programs a charge qualified for, even where none was applied
Integrity attributesKnow how a provider handles corrections, delivery, completeness and configuration of its dataset

The FinOps Foundation’s 1.4 announcement does not describe AI-specific changes; those are scoped to the next release. It also said the FOCUS Validator supports 1.3 conformance testing, with 1.4 support planned for later in Q3 2026.

What changed: what FOCUS 1.5 is set to add

The FOCUS working group’s 1.5 release scope page sets ratification for December 3, 2026. Member approval closed on October 8, and work not approved by then moves to the next release. The working draft includes:

  • AI model identity. ModelDeveloper, ModelFamily, ModelId and ModelVersion, as recommended FOCUS-defined properties in SKU Price Details. Recommended, not required.
  • Token labels. TokenDirection (Input or Output) and TokenCacheAction (Uncached, Read, Write or Other), the first SKU Price Details properties with a fixed set of allowed values.
  • Caller attribution. Three new Cost and Usage columns: PrincipalId for the identity granted access, CredentialId for the credential presented, such as an API key or session, and RequesterDetails for further requester attributes.
  • Allocation. An AllocatedServiceName column naming the service a split charge is allocated to, while the original service name stays on the row.
  • Units and currency. Character, Image, Job, Message, Page and Session join Token and Request as recommended unit names, and provider-issued units such as platform credits can be recorded as a consumption currency.

Still in final review at the time of writing: a SKU Price dataset that publishes prices separately from consumption, six worked examples for mapping AI billing, and tighter definitions of list and contracted pricing. Not in 1.5: a Recommendation dataset, and requests for reasoning-token billing, input modality and batch inference tiers. The working group states that nothing on the page is a commitment until ratification.

Diagram: a FOCUS 1.5 AI charge row draws model identity and token labels from SKU Price Details properties, and caller identity from new PrincipalId, CredentialId, RequesterDetails and AllocatedServiceName columns, producing cost per team, per model and per token type
How one AI charge is expected to read in FOCUS 1.5, based on the October 2026 working draft. The FOCUS 1.5 deep dive explains each key.

Where it is heading (analysis)

Token economics become comparable. Input and output tokens, and cached versus uncached reads, are often priced differently. Once those labels sit in a standard place, cost per request and cost per outcome can be compared across vendors without hand-built mapping tables.

Identity becomes the allocation key for AI. Tags work well for infrastructure, but model API calls are made by applications and agents holding a credential. Standard PrincipalId and CredentialId columns mean the API key or service identity you issue becomes the most reliable way to charge AI spend back to a team. That turns credential hygiene into a FinOps concern, and ties FinOps for AI directly to agent governance.

Coverage will be uneven at first. Model identity properties are recommended rather than required. Expect a period in 2027 where some providers populate these fields fully, some partially and some not at all.

What to do this quarter

  • Map every source of AI spend. List the AI services on each cloud bill, every direct model API invoice, and AI features bundled into SaaS licenses. Note which sources already offer a FOCUS export and which version it conforms to.
  • Turn on FOCUS exports where they exist. Build reports on the FOCUS columns rather than provider-native ones, so a 1.5 upgrade is a schema change, not a rewrite.
  • Issue AI credentials per application or team, now. Shared API keys will make CredentialId useless for chargeback. Separate keys or service identities today are what make attribution work when 1.5 data arrives.
  • Split input, output and cached tokens in your unit economics. Where a provider export already separates them, use that now; where it does not, ask. Cost per thousand requests is misleading if one application is mostly long outputs and another mostly cached reads.
  • Pilot invoice reconciliation. If a provider supports 1.4, try joining Invoice Detail to cost and usage for one month and show the result to accounts payable.
  • Ask your vendors. Put FOCUS exports, and the version they conform to, in your next renewal conversation.
  • Reserve a migration window in Q1 2027. Plan time to adopt 1.5 columns once your providers ship them, and to retire any interim token-mapping tables.

Discuss

How are you allocating AI and token spend today: showback, chargeback, a shared pool, or not yet? And are your AI credentials already issued per team? Share your approach in the comments below.

Sources

Last verified: November 5, 2026. This article will be updated when FOCUS 1.5 is ratified.

Leave a Reply