How much did Stripe pay for OpenRouter?

Stripe has not published a price. Bloomberg reported on August 16, 2026 that the finalized agreement was worth more than $7 billion. The New York Times reported on August 19, 2026 that the figure is about $7.5 billion, split roughly $1.5 billion to OpenRouter's founders and $6 billion to investors. Axios reported more than $8 billion in cash and stock on August 17, 2026. Treat every number as reporting, not a disclosed term — the official Stripe newsroom announcement confirms the acquisition without stating consideration.

Will OpenRouter keep working after the Stripe acquisition?

Yes, as of August 24, 2026 nothing has changed operationally. The API, model catalog, credits and routing behaviour are all live and Stripe framed the deal around extending token routing and usage-based billing rather than winding the product down. The realistic risks are slower and structural: pricing model changes, deeper Stripe account coupling, and eventual consolidation into Stripe's billing stack. None of those have been announced.

Why would a payments company buy an AI model gateway?

Because token consumption is becoming a billing primitive. Stripe already sells usage-based billing, and an AI gateway sits exactly where tokens are metered — every request through OpenRouter is a priced, attributable unit of consumption across dozens of providers. Owning the routing layer gives Stripe the meter, not just the invoice, for AI spend, which is the same reason it built token-billing tooling before the acquisition.

Should I migrate off OpenRouter now?

No, but you should make migration cheap. The correct response to any acquisition of infrastructure you depend on is to verify your exit cost, not to pay it immediately. Keep at least one provider key that does not route through the gateway, confirm your code targets the OpenAI-compatible schema rather than gateway-specific fields, and log which model actually served each request so a future switch is a config change.

Quick Answer

What Is the Stripe OpenRouter Acquisition? (August 2026)

Published: August 24, 2026

The Short Answer

Stripe agreed to acquire OpenRouter, the AI model gateway and routing platform, in a deal announced publicly the week of August 17, 2026. Stripe’s own newsroom post confirms the agreement and frames it around helping businesses “optimize token routing and usage.” The price was not officially disclosed; reporting placed it between roughly $7 billion and $8 billion.

For anyone routing production LLM traffic, the practical answer as of August 24, 2026 is: nothing has broken, nothing has repriced, and the sensible move is to check your exit cost rather than to execute an exit.

	Detail
Acquirer	Stripe
Target	OpenRouter — multi-provider LLM gateway / model marketplace
Announced	Week of August 17, 2026 (Stripe newsroom)
Reported price	~$7.5B (NYT, Aug 19) · >$7B (Bloomberg, Aug 16) · >$8B cash + stock (Axios, Aug 17)
Officially disclosed price	None
Stated rationale	Token routing and usage optimization for businesses
Status of the product	Live and unchanged as of August 24, 2026

What OpenRouter Actually Is

OpenRouter is a single OpenAI-compatible API endpoint that brokers requests to hundreds of models across dozens of providers — Anthropic, OpenAI, Google, xAI, DeepSeek, Z.ai, Meta and a long tail of hosts for open-weight models.

Its value was never the models. It was three things:

One integration, many vendors. You write against one schema and change a model string to switch from Claude Opus 5 to Grok 4.6.
One bill. Credits at the gateway instead of a dozen provider accounts, each with its own invoice, tax treatment and minimums.
Price and availability arbitrage. Multiple hosts serve the same open-weight model at different prices and latencies; the gateway picks and fails over.

That third point is why the pricing surface matters so much. In August 2026 the spread between tiers is enormous — Claude Opus 5 at $5/$25 per million input/output tokens, GPT-5.6 Sol at $5/$30, Grok 4.6 at $2/$6, Gemini 3.7 Flash at an introductory $0.75/$3.75 through December 31, 2026, and GPT-5.6 Luna at $0.20/$1.20 after its 80% cut on July 30, 2026. A routing layer sitting on top of a 25x price spread is a genuinely valuable piece of infrastructure.

Why Stripe Bought It

The clean way to read this deal: Stripe did not buy an AI company. It bought a meter.

Stripe’s business is moving money and, increasingly, billing for consumption. Usage-based billing was already a Stripe product line, and it had shipped token-billing tooling before this acquisition. The problem with billing for AI usage from the outside is that you only ever see the invoice — you do not see the request.

An AI gateway sees the request. Every call through OpenRouter is a metered, attributable, priced event with a known model, a known provider, a token count and a cost. That is the raw material for:

Usage-based billing for companies reselling AI features to their own customers.
Cost attribution per customer, per feature, per agent run.
Agentic commerce, where software agents spend money and someone has to authorize, meter and settle it.

The last one is the strategic bet. If autonomous agents become meaningful buyers, the company that sits between the agent and the model — counting tokens and dollars — is holding a payments rail, not a developer tool.

What Changes for You (and What Doesn’t)

Unchanged as of August 24, 2026:

The API endpoint and OpenAI-compatible request schema
Model catalog and per-model pricing
Credits, keys and existing integrations
Routing and failover behaviour

Worth watching over the next two to four quarters:

Risk	What it would look like	How to hedge now
Pricing model change	Gateway margin restructured, or fees folded into Stripe billing	Track your effective per-token markup monthly
Account coupling	OpenRouter identity merged into Stripe accounts	Note which credentials are gateway-scoped vs Stripe-scoped
Product consolidation	Gateway repositioned as a Stripe billing feature	Keep one direct provider key, unused but working
Enterprise-first drift	Free/hobby tiers deprioritized	Know your fallback for low-volume workloads
Neutrality questions	A payments owner brokering model choice	Log the model that actually served each request

None of these are announced. All of them are the normal shape of post-acquisition drift, and all of them are cheap to insure against today.

The Three-Step Hedge

If you route production traffic through OpenRouter, spend an hour on this and then stop worrying:

Confirm you are schema-portable, not gateway-portable. If your code only uses standard OpenAI-compatible fields, you can point the base URL at LiteLLM, Cloudflare AI Gateway, Portkey or a provider directly. If you depend on gateway-specific routing fields, isolate them behind one function.
Keep one live direct key. One Anthropic or Google key with a $5 balance and a passing smoke test in CI is the entire disaster-recovery plan.
Log model provenance. Record which provider and model actually served each request. Without it, you cannot tell a routing change from a quality regression — which is exactly the ambiguity that makes gateway migrations painful.

The Bigger Signal

This deal is a data point about where the AI stack is consolidating. Model quality at the frontier is converging — Grok 4.6, GPT-5.6 Sol and Claude Opus 5 are close enough on general intelligence indices that price and reliability increasingly decide the choice. When the models converge, the layer that chooses between them gains power, and the layer that bills for them gains more.

A $7B+ price tag for a routing company, paid by a payments company, is the market saying that metering AI consumption is worth more than most people assumed.