Why ScaledNative
The hyperscalers modernize you onto their cloud and process your estate in theirs. ScaledNative recommends the target that fits your estate, runs the model where your code and data already live, and proves the reimagined behavior matches the legacy — with a human clearing the gate.
The honest comparison
AWS Transform, Google Cloud Dual Run, and Azure app-modernization are serious tools. These differences are structural — they come from how the business is built, not from a feature gap that closes next quarter.
Hyperscaler modernization
ScaledNative
Cloud target
AWS Transform modernizes to AWS. Google Cloud's reimagine + Dual Run target GCP; Azure app-modernization targets Azure. The destination is the vendor's cloud.
AWS Transform GA, May 2025 — AWS's modernization service produces AWS-targeted workloads.
We recommend and deploy to AWS, GCP, Azure, or on-premise — the target is scored for your estate and constraints, with a documented rationale. The cloud is chosen in your interest, not ours.
Where your code & data run
Source is sent to a foundation model in the vendor's cloud to be analyzed and reimagined. The estate leaves your perimeter to be processed.
A per-customer forked model runs inside your boundary. Estate source and data never leave your walls, and there is no external token meter on the model.
Proof the behavior didn't change
Equivalence is established by testing and review against the new system — the verdict, where one exists, lives in the vendor's pipeline.
A parallel-run oracle replays recorded fixtures through legacy and reimagined, diffs field-by-field, and emits a threshold-gated, Ed25519-signed equivalence report — in your boundary, fail-closed on restricted fixtures.
Who clears the cutover
Cutover is a project milestone managed by the integrator or the customer's team.
Cutover is a designed gate: a human reviews the signed equivalence verdict and clears it. The approval is part of the lineage, not a side conversation.
When value ships
Months of assessment precede the migration; meaningful value arrives at cutover.
A working workflow deploys while the rest of the estate migrates. Value is incremental — not a six-month report followed by a big-bang switch.
What you can hand an auditor
A migration report and the vendor's internal logs — trust the metadata.
Every change writes to an append-only, signed lineage spine: what changed, why, from which extracted rule, and who approved it — reconstructable and verifiable offline.
AWS Transform reached general availability in May 2025 and targets AWS. Reimagine and lineage are converging toward table stakes across the market — cloud-agnostic target selection, the in-boundary forked model, and a customer-owned signed equivalence verdict are the three things a hyperscaler can't match without working against its own cloud business.
Named, one at a time
Each is genuinely good at part of the job. The gap is the same one each time: the target is their cloud, the processing happens in their cloud, and the equivalence proof — if there is one — is theirs, not yours.
AWS Transform
Generally available since May 2025; modernizes to AWS.
A serious platform — but the destination is AWS, your source is processed in AWS's cloud, and the agnostic cost waterfall across clouds is not one a hyperscaler will ever show you.
Google Cloud (Gemini + Dual Run)
Reimagine + dual-run rigor, targeting Google Cloud.
Same parallel-run discipline and a Gemini-class reimagine — but your code and model never leave your walls with us, and you are not married to GCP at the end of it.
Azure app-modernization
Copilot-assisted app-modernization, targeting Azure.
We run the modernization where Copilot-class cloud tooling legally can't — air-gapped and inside a regulated boundary — and the target is still yours to choose.
Transpile-and-rehost tools
Convert COBOL as-is into a new language or runtime.
Moving COBOL to a new machine leaves you with the same procedural logic in a new costume. We extract the rule and rebuild around it — then prove the behavior matched before cutover.
What's live today
These capabilities are verified in source and runnable in the demo today. We sell the architecture honestly — what runs, runs; the rest is on the roadmap, not in this list.
/mainframe assessment on a real COBOL parser
Deterministic COBOL / copybook / JCL parse → LOC, dependency edges, dead-code candidates, and a per-asset disposition traced to a parsed fact.
The parallel-run equivalence gate
Replays fixtures through legacy vs. reimagined, diffs field-by-field with severity ranking, and emits a signed BehavioralEquivalenceReport — fail-closed on restricted-class fixtures.
In-boundary model + signed lineage
A per-customer forked model runs inside your perimeter, and every action lands on an append-only Ed25519 hash-chain you can verify offline.
Run where your estate lives
Single-tenant instances deploy to AWS, GCP, Azure, or on-premise — bring-your-own-cloud by default, with no run-commitment afterward.
Honest current state: in-boundary serving runs over a tunnel in the public demo as a stand-in for “inside your boundary”; production deploys in-cluster on the same code path. No customer fork is live in production yet, and connectors do not touch real customer systems in the demo.
Walk an estate from assessment to a signed, human-cleared cutover — on synthetic data, in the demo sandbox.