TL;DR: The venture-backed usage-billing specialists were bought during 2026 — Stripe took Metronome, Adyen took Orb for $335 million (closed 1 July), Salesforce took m3ter, Kong took OpenMeter. The category did not disappear with them. What remains independent is real and worth naming precisely: Lago (AGPLv3), Flexprice (AGPL-3.0, credits-native, publicly priced), OpenMeter (Apache 2.0 — acquired, but permissively licensed, which is a different kind of safe), Amberflo (commercial, independent), the subscription-first suites, and owning the meter yourself. This is the roster and how to choose between them. For what happened and why, see the market page.
"There are no independent usage-billing platforms left" is a claim you will see repeated, usually in a paragraph that then names Lago. It is not true, and the imprecision matters, because "independent" is doing at least three different jobs in that sentence.
Three kinds of independence, and they are not interchangeable
- Corporate independence. Nobody owns the company. Fragile by definition — it is exactly what Metronome, Orb, m3ter and OpenMeter each had until they did not.
- Processor neutrality. The product does not assume one payment rail. This is what most buyers actually mean, and it survives an acquisition only as long as the acquirer wants it to. Adyen's phrasing that multi-PSP support "continues initially" is the precise sound of this being undecided.
- Licence independence. The code is yours under terms nobody can revoke. This is the only one an acquisition cannot touch, and it is why OpenMeter being bought by Kong is a much smaller event than Orb being bought by Adyen.
Rank those against your actual risk before shortlisting, because they point at different vendors.
What is actually left
| Platform | Status | Licence | Scope | The honest caveat |
|---|---|---|---|---|
| Lago | Independent | AGPLv3 | Full billing — subscription, usage, prepaid, hybrid | The free stack (Docker, Postgres, Redis) is not the ClickHouse-and-Kafka configuration behind its throughput numbers |
| Flexprice | Independent | AGPL-3.0 | Billing with credits as the primary object | Younger; fewer public references at enterprise scale |
| OpenMeter | Acquired by Kong | Apache 2.0 | Metering and entitlements — stops before the invoice | Direction now points at Kong Konnect; the licence means that cannot reach the code you have |
| Amberflo | Independent | Commercial | Metering, with a cost-governance slant | Corporately independent today, with none of the licence protection |
| Chargebee, Zuora, Recurly | Independent | Commercial | Subscription-first, usage as an add-on | Not metering-native; fine when usage is the minority of the bill |
| Your own meter | Yours | — | Whatever you build | You now own dedupe, late events, retention and a restore drill |
Note the row that breaks the pattern. OpenMeter was acquired and is still the safest of these on the axis most people say they care about, because Apache 2.0 has no take-back. Meanwhile Amberflo is corporately independent with no licence protection at all — which is the position every one of the acquired vendors was in the day before their announcement.
The decision matrix
| Your situation | Reasonable default | Why |
|---|---|---|
| Processor neutrality is the requirement, and you want invoices too | Lago | The most complete open-source biller, genuinely processor-agnostic across Stripe, Adyen, GoCardless and others |
| Your pricing is credits and prepaid balances | Flexprice | Grants, top-ups, per-grant expiry and rollover caps are the data model rather than an approximation. Publicly priced at $500/$1,000 per month |
| You need metering only, and the licence must be permissive | OpenMeter | Apache 2.0, no network clause. Especially strong if you already run Kong |
| You want managed metering and can accept a commercial vendor | Amberflo | Strong metering; decide whether you are buying monetisation or cost governance |
| Usage is a line item on a subscription business | Chargebee / Recurly / Zuora | A metering pure-play is overkill, and their invoicing and tax coverage is better |
| AI economics — per-model, per-customer, per-task attribution | Own the meter, rent the biller | Write-time attribution is a product requirement that no billing export can reconstruct afterwards |
Why the last row keeps winning the argument
The acquisitions are themselves the evidence. Stripe did not pay a reported billion dollars for invoice templates — invoicing is well-understood, commoditised work. It paid for the real-time metering layer. Three sophisticated engineering organisations independently concluded that the meter was worth buying rather than building.
That should update your own build-versus-buy estimate in both directions. Metering properly is harder than it looks — the sum is trivial, and the correctness guarantees around it are not: deduplicating retried events, deciding what happens to usage that arrives after a period closed, keeping a balance fresh enough to enforce a spend cap, retaining raw events so a line can be proven months later. But it also means the meter is the layer worth keeping under your own control, because it is the layer everyone else wants to own.
The practical form of that is modest: whatever vendor you pick, land a copy of every meterable event, with its dimensions and its idempotency key, in a store you control. Then a vendor being acquired, or doubling its price, is a projection change and some glue work rather than a re-instrumentation of your product.
Four questions to ask any vendor now
- What is my exit path, concretely? Can I export raw usage events — not invoices — in a documented format, continuously and at termination? If leaving requires re-instrumenting my product, I am not a customer.
- Is your pricing coupled to payment volume? Percent-of-revenue billing fees plus processing fees compound quietly. Ask for the all-in take rate across billing, metering and processing as one number.
- What survives a change of ownership? No longer a rude question; it is the base case. Price locks at renewal, data-export guarantees and notice periods are worth more than roadmap promises.
- Where do my usage events live, and who can see them? Consumption data is effectively your price book. If the party holding it also sits across the table in a processing negotiation, know that going in.
The honest take
Nobody should panic-migrate. Acquisition-day migrations turn someone else's strategic decision into your own outage, and these products will likely improve under well-funded owners before anything about them gets captive.
What ended in 2026 is the assumption that a metering vendor is a neutral utility. The durable posture is unglamorous and cheap: own your usage data at write time, keep your instrumentation vendor-neutral, and get your export guarantee in the contract rather than in a blog post. Billing vendors turned out to be temporary. The meter is the part you keep.