As of October 2026, AWS Data Exports gives you two serious options for cost data: Cost and Usage Report 2.0, and FOCUS 1.2 with AWS columns. The FOCUS specification has moved on to version 1.4, but AWS still exports 1.2 at most. The answer to CUR 2.0 vs FOCUS depends on who reads the data. CUR 2.0 is the detailed, AWS-native dataset; FOCUS 1.2 gives you standard cost and usage semantics you can use across providers.

Start from the decisions the data has to support and the fields those decisions need. If finance wants one model across providers while engineering digs into AWS-specific discounts, resources or Bedrock callers, two well-controlled exports are often cleaner than one export stretched to do both jobs. If Google Cloud is one of those providers, the equivalent groundwork there is to set up the Google Cloud billing export for cost reporting.

Where things stand in October 2026

Data Exports offers standard exports of five tables: CUR 2.0, cost optimisation recommendations, FOCUS 1.2 with AWS columns, FOCUS 1.0 with AWS columns, and carbon emissions. On top of that sit the QuickSight dashboard export and legacy CUR, which uses its own API.

Three facts shape the choice:

  • A version gap. The FOCUS Steering Committee ratified FOCUS 1.4 on 4 June 2026, adding datasets for invoice detail and billing periods. On 10 October 2026 the AWS documentation lists no table for FOCUS 1.3 or 1.4. If you need 1.4 fields, you derive them yourself or wait.
  • Quotas. AWS allows five CUR 2.0 exports but only two per FOCUS table, so there is little room for experiments on the FOCUS side.
  • Cross-account delivery. Since April 2026, standard exports can be delivered straight into an S3 bucket in another authorised AWS account, such as a central analytics account fed from the management account. The copy job between accounts goes away.

The short decision

NeedPreferred starting pointWhy
Migrate an existing CUR pipelineCUR 2.0AWS documents a backwards-compatible path using an SQL query in the old schema
Investigate AWS-specific costCUR 2.0Nested product, discount, tag and cost-category data
Resource-level or split cost allocation dataCUR 2.0Only CUR 2.0 has table configurations for them
Bedrock cost by calling IAM identityCUR 2.0The line_item_iam_principal column exists only there
One model across providersFOCUS 1.2Standard terms and columns across provider boundaries
Reconcile invoices in a FOCUS pipelineFOCUS 1.2Includes InvoiceId and InvoiceIssuerId
Find unused capacity reservationsFOCUS 1.2Dedicated identifier and status fields, separate used and unused rows
Deep AWS analysis and multi-provider reportingBothSeparate stable consumers justify separate reconciled data products

For a new FOCUS export, we would not start on 1.0 any more. AWS explicitly documents breaking changes between 1.0 and 1.2.

What CUR 2.0 is good at

Unlike legacy CUR, CUR 2.0 has a fixed schema. The formerly dynamic column groups resource_tags, cost_category, product and discount become nested key-value fields. Four columns exist only in CUR 2.0: bill_payer_account_name, line_item_usage_account_name, line_item_iam_principal and line_item_user_identifier. An additional tags column combines resource tags and cost categories and, unlike them, is not normalised; AWS recommends selecting it instead of the two separate columns.

Table configurations set hourly, daily or monthly granularity, resource IDs, and split cost allocation data. The SQL selection supports SELECT, WHERE, AS and LIMIT, but no joins or aggregation.

New for AI spend: with the option "Include caller identity (IAM principal) allocation data", CUR 2.0 records the IAM identity behind each Bedrock call, and tags on those identities appear with the prefix iamPrincipal/. AWS warns that row count grows by a factor of the number of calling identities per model.

CUR 2.0 fits when existing AWS queries need migrating, analysis depends on AWS product or discount detail, columns and rows should be cut down in the export definition, and AWS engineering is the main consumer.

What FOCUS 1.2 with AWS columns is good at

FOCUS defines shared concepts such as Billed Cost, Effective Cost, Service, SKU, Charge Category and Commitment Discount. The AWS table FOCUS_1_2_AWS carries 57 FOCUS columns plus three AWS columns: x_Discounts, x_Operation and x_ServiceCode. Its only table configuration is granularity.

If you are moving from 1.0, plan for these breaks:

  • x_UsageType is gone and becomes SkuMeter; x_CostCategories moves into the Tags column with the prefix aws:tags:CostCategory/.
  • On-Demand Capacity Reservations and EC2 Capacity Blocks for ML are split into used and unused rows, so row counts rise.
  • Reserved Instance and Savings Plan upfront fees now show "Standard" in PricingCategory instead of "Committed".
  • ConsumedQuantity is null for unused commitments instead of 0.
  • New columns include InvoiceIssuerId, ServiceSubcategory, SkuPriceDetails and four pricing-currency columns.

FOCUS fits when one set of downstream logic has to read AWS, Azure or other sources. It cuts translation work, but providers still differ: null values, conformance gaps and the AWS columns all need handling on purpose. The glossary entry on FOCUS covers the term itself.

Do not turn a shared column name directly into policy

Two providers using the same column name does not mean your company uses it the same way. Settle these rules before anyone writes a pipeline:

Business termRule to define
Monthly costBilled, amortised or effective cost?
Consumption ownerAccount, tag, cost category, IAM identity or an internal rule?
Commitment purchaseIn which report and period does it appear?
Unused commitmentWhich owner receives the residual?
Credit and refundOriginal period or posting period?
InvoiceWhich identifiers and totals are the reconciliation basis?
CurrencyBilling, pricing or reporting currency?

Put these rules into mappings and tests, not just a dashboard. Currency matters more in the EU and UK than the US documentation suggests: AWS lists its prices in US dollars, while a European company may be invoiced in euros and report in euros or pounds. The guide to AWS and Azure invoices in euros shows which amounts the invoice actually carries.

When running both is justified

A second dataset earns its place only when it serves a different consumer with stable needs. A pattern that works:

  • FOCUS 1.2 feeds normalised cross-provider reporting.
  • CUR 2.0 feeds AWS-specific investigations, such as resources or Bedrock callers.
  • Both reconcile to the same AWS invoice for each billing period.
  • No measure sums both datasets.

If FOCUS already covers every decision you make today, an extra CUR 2.0 feed is something to run and monitor with nothing to show for it.

FOCUS 1.2 with AWS columnsCross-provider reportingCUR 2.0AWS investigations: resources, discounts, Bedrock callersThe same AWS invoice per billing period
Two exports, two consumers, one invoice: each dataset reconciles to the same billing period, and no measure sums both

Process delivery and corrections safely

AWS delivers data files plus a Manifest.json, and the manifest appears only once every file is written. Exports refresh whenever the source data changes, at least daily for CUR 2.0. After a billing period ends, AWS may redeliver it within the first two weeks. In overwrite mode, surplus chunks are overwritten with empty data when an export shrinks. Table configurations and the chosen table cannot be changed after creation; switching means a new export.

A pipeline that survives this:

  1. Reads the delivery inventory from the manifest, never from a file glob.
  2. Processes execution and billing period idempotently.
  3. Accepts corrections to the previous month until mid-month as a new version.
  4. Never reads empty chunks as a zero-cost period.
  5. Reconciles source total, loaded rows and transformed total.
  6. Blocks the management view when schema or nullability checks fail.
Yes, until mid-monthNoYesNoManifest.json names the delivered filesLoad per execution and billing period, idempotentlyCorrection of a closed period?New version of that periodReconcile source total, loaded rows and transformed totalSchema and nullability checks pass?Management viewView blocked, failure reported
A delivery pipeline that survives refreshes: the manifest names the files, each execution loads idempotently, a late correction becomes a new version, and the management view opens only when the checks pass

What the export costs

The AWS pages checked state no price for the export itself. Cost appears where the data is stored and read: S3 storage, especially in "create new" mode, which keeps every refresh, plus Athena, Redshift or your own warehouse. Resource IDs, hourly granularity and IAM principal data all make files bigger. Measure one month's volume before switching those options on by default.

A short selection test

Take one closed month and put the same questions to each candidate:

  • Does the defined total reconcile to the invoice?
  • Are account, service, SKU, Region and resource available at sufficient detail?
  • Can commitments, credits, refunds, tax and support be classified correctly?
  • Does the existing ownership and allocation policy work?
  • Can the consumer read the data without provider-specific exceptions?
  • Do uncovered cases remain visible?

Choose the export that answers your questions with the least extra logic. The choice shows up first in the monthly cloud cost review, which needs one declared cost basis.

What Costfluent covers

The AWS connector reads a FOCUS 1.2 export from an S3 bucket in your account through a read-only IAM role. Until the first delivery arrives it reads Cost Explorer, then replaces those figures with the export rows. A separate CUR 2.0 export for deeper AWS analysis is unaffected. The AWS integration page has the details.

Sources and method

Checked on 2026-10-10. AWS What is Data Exports, migrating to CUR 2.0, Data Exports table dictionary, FOCUS 1.2 with AWS columns, migrating from FOCUS 1.0 to 1.2, data query, export delivery, quotas and restrictions, IAM principal cost allocation and the cross-account delivery announcement; FinOps Foundation™, Introducing FOCUS 1.4 and the FOCUS specification. The decision table, rule table and checks are Costfluent recommendations; recheck the table dictionary and conformance notes before fixing a production schema.