Formae's beta just stopped being a toy for AWS-only pilots. With beta support for GCP, Azure, Oracle Cloud (OCI) and OVHcloud, Platform Engineering Labs has flipped the switch: this is an attempt at a single open-source IaC surface for multi-cloud provisioning, not another cloud-specific wrapper.
That matters because the platform ecosystem has been fragmenting along two axes for years. On one side you have opinionated internal developer platforms (IDPs) that bake a narrow set of capabilities and run them well. On the other, you have extensible control-plane projects like Crossplane that let you model cloud primitives as CRDs. Formae is staking a middle ground — delivering opinionated APIs while expanding the underlying provider reach.
Two ecosystem moves make this moment interesting. First, Crossplane's growing adoption and recognition within the CNCF ecosystem turns what used to be "a nice extensibility pattern" into a de facto control-plane approach for Kubernetes-native infrastructure orchestration. Expect more vendor products and internal platforms to base provider models on Crossplane primitives or to offer Crossplane-compatible control planes. Second, Formae's multi-cloud beta signals that teams want a single IaC API that talks to multiple clouds without rewriting converters or managing a bespoke fleet of Terraform modules.
Here's the implication for platform teams: you're no longer choosing just an opinionated IDP versus a generic control plane. You're choosing how opinionated your abstractions must be to avoid constant provider divergence. Formae's promise is tempting — one set of APIs, one place to teach developers — but it also raises the control-plane complexity you have to own: provider feature parity, drift management, provider-specific lifecycle hooks, and testing matrices explode once you move beyond a single public cloud.
The other part of this week's signal is a shift in what platform engineering leaders are prioritizing. Recent research and commentary — including Google's platform engineering work tying IDP outcomes to Four Keys/DORA metrics — show platform maturity is now being measured by concrete developer velocity and quality signals, not just uptime. The list of emerging operational requirements reads like an AI-era platform checklist: non-human identity (token economies for agents and models), GPU provisioning and scheduling, agent-mediated operations, and real-time cost attribution tied to developer actions.
Call this: Formae's multi-cloud push is the right move for teams that want consistent developer APIs across clouds. But it's also the exact thing that will expose gaps in governance and identity. When you have agents or model-driven workflows that need non-human identities and ephemeral credentials (and you will), your control plane must support scoped tokens, short TTLs, and auditable credential issuance at scale. Most teams aren't set up for that yet.
Crossplane's growing adoption accelerates a different risk: CRD sprawl. Control planes that let you model every cloud primitive are powerful; without strong policies, composition patterns, and a curated provider catalog, you'll end up with ad-hoc CRDs, mismatched lifecycle semantics, and poor DX for developers. Opinionated platforms win at predictability. Extensible control planes win at coverage. Expect a wave of platform engineering work that tries to have both — and fails unless governance and developer experience are treated as first-class features.
If you're building the next-generation IDP, focus on three concrete pieces: short-lived non-human credential flows (agent-first architectures), clear GPU scheduling and quota semantics tied to cost attribution, and a curated set of provider compositions that hide cloud-specific failure modes. Formae and Crossplane give you tools; they don't give you the product thinking to make them safe at scale.
One small linkable note: we've covered Formae's beta multi-cloud support in depth here: [/article/formae-beta-multi-cloud-gcp-azure-oci-ovhcloud/]. For a spot-on read on measuring IDP impact with DORA/Four Keys, see Google's recent platform engineering research: [/article/google-cloud-research-measure-idp-four-keys-dora/].
Final take: the era of single-cloud, bespoke IaC is ending. Platforms that stitch a predictable, opinionated developer surface on top of a Crossplane-style control plane will dominate. But if your platform team treats multi-cloud support as a checkbox and not a governance, identity, and cost problem, you will create more operational work for your future self than you save for your developers.