Product Update — May 2026: Everywhere Your Spend Lives

This release extends a single promise: every dollar spent on cloud, AI, and SaaS belongs in one governed view, with one path from seeing a cost to acting on it. Three new sources of spend went GA, AI cost governance gained 31 new savings scenarios, and two areas FinOps teams told us still felt opaque — commitments and Kubernetes — got materially deeper.

Three new connectors went GA: Anthropic, Databricks, and Alibaba Cloud

Usage and cost data from Anthropic, Databricks, and Alibaba Cloud now flow in alongside the rest of the estate. Anthropic spend — every Claude call across every workspace — comes in via a single admin API key. Databricks connects through an OAuth service principal and a SQL warehouse, landing AI platform costs in the same chargeback model as the cloud infrastructure underneath them. Alibaba Cloud customers can bring billing exports in through OSS with a few RAM credentials, normalised into FOCUS alongside AWS, Azure, GCP, OCI, and Google Workspace.

AI cost governance gets serious

Azure OpenAI and Google Vertex AI are now first-class citizens in the savings stream. New Azure OpenAI and Cognitive Services scenarios catch the things that quietly bleed AI budgets — under-utilised provisioned throughput units, poor cache hit rates, idle fine-tuned models, deprecated model versions still attached to billing, and deployments with zero token usage. New GCP Vertex AI scenarios cover the same blind spots for Vector Search indexes, index endpoints, metadata stores, and Tensorboards. Each carries a concrete dollar figure, an owner, and a path to remediation, following the same workflow a team already uses for cloud and SaaS savings.

A predefined AI Dashboard built around unit economics

The new AI Dashboard gives FinOps practitioners and AI platform leads a purpose-built starting view — AI/ML totals, trends, top services, top regions, top resources, and unit economics widgets for the metrics a business actually cares about: cost per token, cost per request, cost per GPU-hour. Sub-cent rates now render at the right precision rather than collapsing to “$0.00” — a small fix, but the difference between a useful chart and a useless one.

Commitments get their own perspectives and dashboard widget

Saved filters — perspectives — now apply to the Commitments page itself, so “all production Savings Plans” or “all expiring Reserved Instances in EMEA” is one click away every visit. A new custom-dashboard commitments widget puts that same slice into any executive view.

Kubernetes gets GPU and storage cost lenses

Kubernetes spend is no longer just CPU and memory. A resource lens switcher inside the Kubernetes view exposes GPU and Storage alongside the existing CPU & Memory view, with per-workload, per-namespace, and per-cluster figures. Persistent Volume costs now land as real catalogue-priced FOCUS rows — pro-rated hourly and tied to a stable identifier so a volume’s cost trail survives bind, release, and re-bind cycles. New structural filters let platform engineers slice the dashboard by cluster and namespace without scrolling through hundreds of pods, and a long-standing irritation — total spend changing depending on which grouping you picked — is gone, with totals now group-invariant and infrastructure idle bucketed cleanly.

A clearer name for the most-used workflow

What was previously called “Recommendations” is now Savings. Every page title, breadcrumb, tooltip, and dialog reflects what the workflow actually does — surface money that can be saved, and let a team act on it.

A more consistent filter experience, and quieter under the hood

Cost Explorer, Chargeback, Radar (anomalies), and the Kubernetes dashboard now share a unified filter pattern — edits buffer locally, commit on Apply, reset cleanly with Clear — removing the accidental refetches and confusion of the previous behaviour. And some of the biggest wins this release are invisible: Kubernetes data ingestion runs 40% faster on the metrics path and 80% faster on storage, AWS account enrichment for customers with thousands of S3 buckets dropped from roughly twenty-one minutes to under two, database failovers no longer cascade into visible errors, and Cost Explorer queries hitting million-row cardinality now degrade gracefully with a clear “results truncated” message rather than timing out.

What this all adds up to

There are more places to govern spend — Anthropic, Databricks, Alibaba — and more savings to find inside what was already governed, particularly across AI inference. And the workflows that mattered most — commitments, Kubernetes, savings, dashboards — got materially deeper without getting harder.

See what this looks like against your own multi-cloud and AI estate. Book a 30-minute session and we’ll map it to what you’re actually running. Request a demo →

Frequently asked questions

What’s new about DigiUsher’s Anthropic integration, and why does it matter? It ingests Anthropic API usage and cost data via a single admin API key and normalises it into the same FOCUS schema used by every other data source. AI spend at most enterprises is fragmented across multiple providers, each exposing cost data in its own format — without a unified layer, cross-provider model selection and CFO-level AI cost reporting become effectively impossible. Bringing Anthropic into the same FOCUS-formatted storage means the same allocation rules and unit-economics formulas apply uniformly across every model and provider.

How does DigiUsher govern Azure OpenAI and Google Vertex AI cost natively? Through 15 native Azure OpenAI and Cognitive Services savings scenarios (covering PTU underutilisation, poor cache hit rates, idle fine-tuned deployments, deprecated models, zero-token deployments) and 16 native Vertex AI scenarios covering the Vector Search surface (indexes, index endpoints, metadata stores, Tensorboards). Both sets surface through the same Savings stream a FinOps analyst already uses, with the same severity, status, owner, and remediation workflow as cloud infrastructure recommendations.

What is a unit-economics dashboard for AI, and what does the new AI Dashboard actually show? It’s a view expressing total AI inference cost in the business metrics finance and product leaders actually use — cost per token, cost per request, cost per GPU-hour — rather than raw provider billing line items, since a large AI bill is only meaningful once divided by the interactions or outcomes it supported. The predefined AI Dashboard combines standard cost overview widgets (totals, trends, top services) with three dedicated unit-economics widgets, rendering sub-cent rates at the precision needed to actually compare models meaningfully.

How does the new Kubernetes cost lens handle GPU and storage spend specifically? GPU and Storage now sit as dedicated resource lenses alongside CPU & Memory. The GPU lens surfaces per-workload, per-namespace, and per-cluster GPU-hour consumption and utilisation efficiency — the most common waste pattern in production ML workloads, and the hardest to catch from cloud billing alone since GPU spend normally lands as instance-hour billing on the node rather than attributed to the actual workload. The Storage lens treats Persistent Volume cost as a first-class line item, pro-rated hourly with a stable identifier so the cost trail survives bind-release-rebind cycles that normally fragment attribution.

Why did DigiUsher rename “Recommendations” to “Savings”? Customer interviews converged on one finding: new users couldn’t infer from the word “Recommendations” what the workflow actually did. “Recommendations” is technically accurate but semantically vague; “Savings” names the outcome directly — a list of dollar amounts the organisation can recover, with an action attached to each one.

Related reading

• Product Update — May 2026: From Insight to Action — May 4, 2026
• Product Updates - DigiUsher Ships Its First AI Cost Attribution Feature — July 13, 2026
• Product Updates - Kubernetes GPU Governance and Resource Intelligence — June 29, 2026
• AI Unit Economics: Why Your AI Invoice Can No Longer Answer the Board’s Question — July 16, 2026

© 2026 DigiUsher. All rights reserved. Privacy Policy · Terms of Service · Trust Center