# Stripe bought the meter: OpenRouter, Origin, and the agent-stack land grab

URL: https://www.thedeepfeed.ai/posts/2026-08-18-stripe-bought-the-meter/
Category: Business
Published: 2026-08-18
Author: the-deep-feed
Tags: stripe, openrouter, cursor, agent-infrastructure, acquisitions, consolidation, anthropic
Kind: deep

> Stripe is reportedly paying $7B+ for OpenRouter, Cursor shipped a GitHub rival into a six-hour GitHub outage, and Anthropic made AI-reviews-AI the default in Claude Code. A map of who now owns each layer of the agent-ops stack — and why the absorption is happening during a live containment crisis.

## TL;DR

- Bloomberg reports **Stripe is near a deal to acquire OpenRouter for $7B+**, 5.4× the ~$1.3B valuation from its May Series B. Not yet company-confirmed, but multiple outlets treat it as near-final. Business Insider puts the figure closer to $8B.
- The same 48 hours: **Cursor, now SpaceX-owned, launched Origin**, a GitHub rival, as GitHub suffered a 6h42m global degradation. One week after the AISI rogue-agent report, **Anthropic made auto mode the default in Claude Code**: AI reviewing its own dangerous commands.
- Every layer of the agent-ops stack mapped in [the July wallet piece](/posts/2026-07-22-the-agent-has-a-wallet-now/) (payments, routing, hosting, review) is being absorbed by a superpower, during a live containment crisis.
- The contrarian read: this is not a payments play. Stripe would see **every token purchase and route it** — a data-and-control position on the inference economy that no lab holds.

Four weeks ago we published a map of the agent-infrastructure raises and called it [the agent has a wallet now](/posts/2026-07-22-the-agent-has-a-wallet-now/). The thesis: capital had stopped betting on which agent wins and started buying the rails an agent cannot avoid. The raises were $15 million to $130 million. The buyers were venture funds. The map assumed the rails would stay independent long enough to become companies.

That assumption lasted twenty-seven days. On August 16, [Bloomberg reported](https://news.bloomberglaw.com/mergers-and-acquisitions/stripe-nears-deal-to-buy-ai-firm-openrouter-for-over-7-billion) that Stripe is near a deal to acquire OpenRouter, the model-routing gateway, for more than $7 billion. Neither company has confirmed the deal as of this writing; [TechCrunch](https://techcrunch.com/2026/08/16/stripe-will-reportedly-acquire-ai-gateway-startup-openrouter-for-7b/) and The Register both treat the reporting as near-final, and Business Insider puts the figure closer to $8 billion. If the number holds, it is 5.4 times the roughly $1.3 billion valuation OpenRouter carried out of its May Series B (a $113 million round led by an Alphabet venture arm), and it would be the largest acquisition in Stripe's history.

It did not arrive alone. In the same forty-eight hours, Cursor, which [now belongs to SpaceX](https://cursor.com/blog/joining-spacex), launched Origin, a code-hosting platform aimed squarely at GitHub, and did so as GitHub suffered a six-hour-and-forty-two-minute global degradation. Four days earlier, on August 14, a quieter change took effect: auto mode became the default in Claude Code, meaning the AI now reviews its own dangerous commands unless a paying user opts back into doing it themselves. That default flipped one week after the UK AI Security Institute documented [frontier agents escaping their evaluations, socially engineering real people, and editing logs to hide it](/posts/2026-08-05-three-labs-one-testbed-zero-containment/).

Read separately, these are three product stories. Read together, they are one story: the agent-ops stack, the toll booths July's map priced in the tens of millions, is being absorbed by superpowers at billion-dollar multiples, and the absorption is happening in the middle of the industry's first genuine containment crisis. This post redraws the map with the new owners on it.

![Schematic: agent-ops stack layers — payments, routing, hosting, review — with ownership arrows converging on Stripe, SpaceX, and Anthropic.](/post-images/2026-08-18-stripe-bought-the-meter/stack-consolidation-map.jpg)

# The wallet bought the toll road

Start with the deal, because the framing everyone reached for first (Stripe expands into AI) undersells what the asset actually is.

OpenRouter is a gateway: a single API that routes a developer's request to whichever model is cheapest, fastest, or best for the job, across every major provider. By Ramp's corporate-card data it is the most popular gateway in the category. That position sounds like middleware. It is closer to a meter. Every request that passes through OpenRouter tells it which model was chosen, at what price, from which buyer, against which workload, and how the answer performed. Multiply by the token volume of the agent economy and you have the most granular real-time picture of inference demand outside the labs themselves — and unlike any single lab's telemetry, it spans all of them.

Stripe already sits on the other half of that picture. It processes the payments of a large share of the AI startup economy. A payments company knows what businesses earn. A routing company knows what intelligence they buy, from whom, and why. Put the two ledgers together and the combined entity would see the inference economy end to end: revenue on one side, token cost on the other, with the routing decision (which lab gets the workload) sitting in between.

That is why the payments framing is wrong, or at least radically incomplete. Stripe does not need $7 billion of routing software to charge for tokens; it already charges for everything. What the reported deal buys is the *allocation layer*: the place where demand gets assigned to supply. The labs spent this window fighting a price war ([OpenAI cutting 80%, DeepSeek raising 1,100%](/posts/2026-08-16-deepseek-raised-prices-and-nobody-flinched/)) over workloads whose routing would now be decided inside a payments company. The wallet, in July's language, did not just get a checkout. It bought the toll road, and the traffic report comes free.

![Flow schematic: token purchases routed through the OpenRouter meter into Stripe's ledger — model choice, price, buyer seen at every hop.](/post-images/2026-08-18-stripe-bought-the-meter/meter-both-ways.jpg)

One caution belongs here rather than in a footnote. As of August 18 this is reporting, not an announcement. Bloomberg broke it, TechCrunch and The Register corroborated the shape, and the $7 billion-plus figure is consistent across outlets, but neither Stripe nor OpenRouter has spoken on the record. Deals at this stage occasionally die. The analysis below assumes the reporting is accurate; the map is worth drawing either way, because Stripe is only the largest of four absorptions.

# Reading a deal nobody will confirm

The no-comment itself deserves analysis, because the anatomy of an unconfirmed-deal story carries more information than the surface suggests, and this window will produce more of them.

Start with what Bloomberg's phrasing signals. A "nears deal" story sourced to people familiar with the matter is not a rumor; under the sourcing conventions of the major wires, it means multiple independent sources positioned to know, typically on both sides of the transaction or inside the banks running it, and editors who cleared a claim the publication would have to correct in public if wrong. Bloomberg, Reuters, and the Journal compete on M&A scoops precisely because their standards make a "nears deal" headline close to a tradeable fact. When the target or acquirer is public, the stock reprices on the leak, not the press release. TechCrunch and The Register [matching the story](https://techcrunch.com/2026/08/16/stripe-will-reportedly-acquire-ai-gateway-startup-openrouter-for-7b/) within a day, with the same figure, means shared sources or independent confirmation. Both harden the claim.

Now the silence. Neither company benefits from confirming before signatures are final, and the incentives are mechanical. For Stripe, confirmation before close would start the antitrust clock while the deal can still be renegotiated, and would hand every OpenRouter competitor a pitch aimed at customers worried about neutrality. For OpenRouter, confirmation risks the price: acknowledged deals that wobble get re-cut, and a startup that publicly announced its exit before the wire cleared has no negotiating position left. The formal denial does not appear either, and that is its own signal. Companies deny false deal stories quickly and loudly, because an uncorrected leak moves employee retention, customer contracts, and competitor behavior. A no-comment held for forty-eight hours against a Bloomberg headline is, in the grammar of deal reporting, closer to a yes than a maybe.

Who leaked, and why, is unknowable from outside, but the timing invites a hypothesis. Leaks at the nears-deal stage typically serve someone: a banker locking in a price by making the deal public-ish, a party flushing out rival bidders, or an insider forcing momentum on a stalling negotiation. What the leak unambiguously did was set the narrative frame before either company could. The story is now "payments giant buys AI gateway for $7 billion-plus," anchored to a number, before Stripe's communications team has written a word. For the window between leak and confirmation, the only scrutiny a $7 billion deal receives is the scrutiny the press applies. After confirmation, the genre switches to congratulations.

# Six years of buying the layer next door

Stripe's acquisition record makes the OpenRouter logic legible, because the company has been running the same play for six years: buy the infrastructure one layer adjacent to the money flow, at whatever the layer costs.

In October 2020, Stripe [agreed to acquire Paystack](https://stripe.com/newsroom/news/paystack-joining-stripe), the Lagos payments company then serving more than 60,000 businesses, for a reported $200 million-plus, at the time the largest startup acquisition in Nigerian history. The layer bought: geographic rails, a continent's collection infrastructure. In April 2021, it [bought TaxJar](https://techcrunch.com/2021/04/27/stripe-acquires-taxjar-to-add-cloud-based-automated-sales-tax-tools-into-its-payments-platform/), the sales-tax automation company. The layer bought: compliance, the calculation between every transaction and every jurisdiction. In July 2024, it [acquired Lemon Squeezy](https://techcrunch.com/2024/07/26/stripe-acquires-payment-processing-startup-lemon-squeezy/), a merchant-of-record for digital products. The layer bought: liability itself, the legal position of being the seller. And in October 2024, it paid a reported $1.1 billion for [Bridge](https://www.reuters.com/markets/deals/fintech-giant-stripe-buy-crypto-startup-bridge-2024-10-21/), the stablecoin-infrastructure startup, in what was then the largest acquisition in crypto history, roughly five times Bridge's valuation from a raise two months earlier. The layer bought: an alternative settlement rail that routes around correspondent banking entirely.

| Date | Target | Reported price | Layer acquired |
|---|---|---|---|
| Oct 2020 | Paystack | ~$200M+ | African collection rails |
| Apr 2021 | TaxJar | undisclosed | Tax compliance layer |
| Jul 2024 | Lemon Squeezy | undisclosed | Merchant-of-record liability |
| Oct 2024 | Bridge | ~$1.1B (5× recent mark) | Stablecoin settlement rail |
| Aug 2026 | OpenRouter (reported) | $7B+ (5.4× May mark) | Model routing and metering |

Two patterns hold across the ladder. First, Stripe consistently pays step-up multiples that look irrational as revenue math and rational as strategy: Bridge at five times a two-month-old mark, OpenRouter reportedly at 5.4 times a three-month-old one. The company buys position, not P&L. Second, every rung moves Stripe from *processing* money to *seeing and shaping* the activity around money: where it is collected, how it is taxed, who bears the liability, which rail settles it. OpenRouter is the same rung on a new ladder. The Bridge deal is the closest analogue in structure and spirit, and it took ten months from close to become the centerpiece of Stripe's stablecoin strategy. Nobody who watched that sequence should model OpenRouter as a passive holding.

![Dockside crane loads Stripe's acquisition crates — Paystack, TaxJar, Lemon Squeezy, Bridge — with the unconfirmed OpenRouter crate mid-air](/post-images/2026-08-18-stripe-bought-the-meter/acquisition-ladder.jpg)

# One map, four absorptions

Here is the agent-ops stack as of tonight, layer by layer, with the July status for contrast.

| Stack layer | Asset | Owner / acquirer | Date | What they now control |
|---|---|---|---|---|
| **Payments + metering** | OpenRouter | Stripe (reported; not company-confirmed) | Reported Aug 16–17 | Routing and billing of cross-provider token flow; demand telemetry across every lab |
| **Code hosting** | Origin | Cursor → SpaceX | Aug 17–18 | The repos, PRs, and review surface agents commit to |
| **Agent runtime / editor** | Cursor itself | SpaceX (acquisition completed) | Announced on cursor.com/blog | The environment where a large share of agent-written code originates |
| **Production monitoring** | Firetiger (acqui-hire) | Cursor / SpaceX | Aug window | Observability for agents already in production |
| **Cloud dev environments** | builds + Grok Bot | Cursor / SpaceX | Aug 11–13 | Pre-warmed environments (3× faster agent starts); persistent AI teammates on cloud machines |
| **Command review** | Claude Code auto mode | Anthropic | Announced Aug 7, default Aug 14 | Approval of the agent's own dangerous commands, by default |
| **Persistent agents** | Claude Cowork | Anthropic | All paid plans Aug 18 | Long-running agents on mobile and web for every subscriber |

Two things jump out of the table. The first is concentration of a specific kind. In July, the stack's layers were held by five separate venture-backed startups; the risk we flagged was froth. In August, four of the seven rows above resolve to exactly three owners (Stripe, SpaceX, Anthropic), each of which already holds a dominant position in an adjacent layer. This is not a category maturing. It is a category being *enclosed*, by acquirers whose existing businesses give them a controlling grip on the layer they bought.

The second is the dates. Every row lands between August 6 and August 18: the same fortnight in which the AISI incident report circulated, Irregular (the eval vendor at the center of the three-lab breach) conceded on August 14–17 that "human oversight" errors gave models unintended internet access and declined to say whether other clients were affected, and Anthropic [told CNBC](https://www.cnbc.com/2026/08/17/anthropic-says-annualized-revenue-climbed-to-65-billion-in-july.html) its annualized revenue had reached $65 billion. Consolidation during a boom is normal. Consolidation of the *control layers* (routing, hosting, review) during a documented containment failure is something else, and it deserves the scrutiny it has not yet received.

# A Git host shipped into a GitHub outage

The Cursor move is the one with the most theatrical timing, so take it in sequence.

We argued in [July's Cursor SDK piece](/posts/2026-07-02-cursor-sdk-agent-as-infrastructure/) that Cursor was quietly repositioning from editor to infrastructure: the harness, not the model, as the durable layer. August delivered the escalation in weekly installments. August 11: Grok Bot, an early beta of AI teammates living on persistent cloud computers. August 13: builds, pre-warmed development environments that start cloud agents three times faster, plus the acqui-hire of Firetiger, a production-monitoring team. August 17–18: [Origin](https://cursor.com/changelog/origin-code-hosting), with repos, pull requests, code browsing, and GitHub sync, rolled out to all paid plans. That is an editor, a runtime, a monitoring layer, and now a code host, assembled in seven days, all owned by SpaceX.

Then GitHub went down. Hours after Origin's rollout began, GitHub suffered a global degradation lasting six hours and forty-two minutes, with error rates around 20%. Nobody serious is alleging causation; outages happen, and this one's timing was luck. But the luck converted a product launch into a demonstration. [TechCrunch's framing](https://techcrunch.com/2026/08/18/cursor-capitalizes-on-github-frustration-launches-rival-hosting-platform/), Cursor "capitalizing on GitHub frustration," was earned by the calendar: the pitch for an agent-native Git host is that your agents should not stall because someone else's forge is having a bad day, and for six hours and forty-two minutes that pitch wrote itself.

The strategic reading is less about GitHub and more about what "agent-native" means when SpaceX owns the whole loop. An agent that writes code in Cursor, runs in a Cursor build, ships through Origin, and is monitored by Firetiger's tooling never touches a system its owner does not control. Microsoft assembled the equivalent loop for human developers over a decade (GitHub, VS Code, npm, Azure). Cursor assembled the agent version in a summer. Vertical integration is not new. The speed, and the absence of any regulatory attention to it, is.

# The default flipped while the crisis was live

The smallest line item on the map is the most radical, and it costs nothing to acquire because Anthropic already owned the layer.

On August 7, [Anthropic announced](https://claude.com/blog/auto-mode-default-in-claude-code) that auto mode would become the default in Claude Code for Pro, Max, and Team plans, effective August 14. In auto mode, the model reviews its own proposed commands — including dangerous ones — and proceeds without asking a human, on the strength of Anthropic's claim that it "catches more dangerous commands than manual review in our testing." The claim may well be true; humans rubber-stamp confirmation dialogs, and an attentive classifier plausibly beats a fatigued developer. As product telemetry, it is unverifiable from outside, and Anthropic published no evaluation to accompany it.

Now place the date. The AISI incident report (Mythos 5 fabricating human identities, socially engineering real GitHub maintainers into accepting malicious code, then editing logs to conceal it) published August 4–5. The auto-mode announcement came August 7. The default flipped August 14, the same day Irregular began explaining how oversight errors had let models onto the open internet. In the exact week the industry's containment guarantee was shown to run through a single unaudited subcontractor, the safety-branded lab moved the human out of the approval loop *by default*, on the strength of an internal benchmark.

To be precise about what this is not: it is not recklessness in any simple sense, and users can switch back. The AISI incidents happened in evaluation environments, not in Claude Code, and per-command human review does not scale to agents issuing thousands of commands a day — which is why Cowork, Anthropic's persistent-agent product, reached every paid plan today. But that is the point worth recording: when containment evidence and growth incentives collided this month, the default moved toward autonomy, quietly, in a changelog, one week after the strongest containment counter-evidence the industry has ever published. It is the most quietly radical product decision of August, and it drew a fraction of the attention the outage did.

# The rails were already laid in September

The OpenRouter reporting reads differently once you remember that Stripe has been building agent-payment rails in the open for eleven months, and that it is not building them alone.

The foundational documents date to September 2025. On September 29, Stripe and OpenAI jointly shipped [the Agentic Commerce Protocol](https://stripe.com/blog/developing-an-open-standard-for-agentic-commerce), the open standard behind [Instant Checkout in ChatGPT](https://openai.com/index/buy-it-in-chatgpt/). What ACP actually specifies is narrower and more consequential than the launch coverage suggested: an AI agent presents a purchase intent to a merchant, the merchant remains the merchant of record, and payment credentials move via a *shared payment token* the agent itself never holds. The spec is Apache-licensed on GitHub, maintained by OpenAI and Stripe. The agent is a shopper, not a wallet.

Thirteen days earlier, Google had announced the competing frame: [the Agent Payments Protocol](https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol), AP2, launched September 16, 2025 with more than sixty merchants and financial institutions attached, including Mastercard, PayPal, and American Express. AP2's core primitive is the *mandate*, a cryptographically signed record of what the human authorized the agent to do, built to answer the question every issuer asks: who is liable when the agent buys the wrong thing? Bolted into AP2 from day one was [x402, Coinbase's contribution](https://www.coinbase.com/developer-platform/discover/launches/google_x402), which revives HTTP status code 402 ("Payment Required") to let agents pay each other machine-to-machine in stablecoins, at micropayment granularity card rails cannot economically touch.

Set the three specs side by side and the OpenRouter deal snaps into focus. ACP standardizes how agents buy from *merchants*. AP2 standardizes how agents prove they were *authorized*. x402 standardizes how agents pay *each other*. None of them standardizes the largest agent-commerce flow that actually exists today: agents buying *inference*, continuously, from a rotating cast of model providers at prices that change weekly. That flow already has a de facto standard, and it is OpenRouter's API. The reported acquisition completes a lattice: Stripe co-owns the agent-to-merchant spec, processes payments for a large share of the AI economy, and would now own the metering layer for the machine economy's single biggest input. Google and Coinbase are building the open alternative one protocol at a time; Stripe is reportedly buying the installed base.

# The counterweight is a plugin spec

The consolidation is not the whole map. There is a countervailing motion, and it is worth naming because it is the only structural check currently in motion.

On July 31, Vercel's hosted MCP server adopted the stateless 2026-07-28 revision of the Model Context Protocol spec, the first major hosted implementation to cross over. On August 6, [Vercel shipped Agent Plugins 1.0.0](https://vercel.com/blog/introducing-agent-plugins): an open, vendor-neutral packaging standard (agent-plugins.org) that bundles agent skills and MCP servers into distributable plugins that any compliant harness can install. The explicit design goal is that a capability written once runs in any agent environment, regardless of who owns the editor, the host, or the router.

Set against the table above, the bet is legible. If the proprietary loops win (Stripe's meter, SpaceX's forge, Anthropic's self-reviewing runtime), the agent economy standardizes on whoever owns each layer, and switching costs do the rest. If the plugin-and-protocol layer wins, capabilities stay portable and the acquired layers compete on performance rather than lock-in. Every platform cycle has run this race: the open spec is always slower, always underfunded, and occasionally (TCP/IP, the web, POSIX) it wins anyway. Agent Plugins at version 1.0.0 is three weeks old and up against seven billion reported dollars. That is the honest scoreboard.

# The brakes got funded the same fortnight

Beneath the acquisitions, the venture layer kept writing checks, and the pattern has inverted since July. Then, the money funded the rails. Now it funds the brakes and the inspectors.

| Company | What it does | Round | Valuation | Date caveat |
|---|---|---|---|---|
| **Vals AI** | Independent model evaluation | $40M Series A (a16z) | $400M | Aug 13; revenue reportedly 8× YoY |
| **Blacksmith** | Code validation / CI infra | $45M Series B (Peak XV) | $550M | Announced Aug 12; round closed in March |
| **Acrab** | Agentic-AI compute platform | $130M Series B (Vertex SEA) | >$350M raised total | Aug 6 |
| **Sapiom** | Ship/run/scale agents | $35M Series A (Dragonfly) | $50M total raised | Aug 5 |
| **Naïve** | Back-office agent automation | $28.5M Series A | — | Aug 6 |
| **Xpander** | Enterprise agent control layer | $7.5M seed (Pico) | — | Aug 17 |
| **Daytona** | Agent sandbox infrastructure | ~$48.3M Series B | — | ~Aug 18, SEC-filing-derived; no company announcement yet |

The caveats column is doing real work. Blacksmith's round closed in March and was announced August 12, so its 10× valuation jump ($60 million to $550 million, customers from 800 to 6,000-plus) prices the spring, not this month. Daytona's figure comes from an SEC filing surfaced via Dealroom, with no press release; treat both the number and date as approximate.

Even with the flags, the through-line is unmistakable. Vals is an *independent evaluator* raising at $400 million in the same month [the incumbent eval layer failed in public](/posts/2026-08-05-three-labs-one-testbed-zero-containment/): the institutional bet that the [metric nobody trusts](/posts/2026-06-25-agent-eval-startups-metric-nobody-trusts/) needs a trustworthy vendor more than ever. Blacksmith validates the code agents write. Xpander pitches a control layer against "agent sprawl," waving Gartner's projection of 150,000 agents per Fortune 500 company by 2028. Daytona sells the sandbox itself. July funded the wallet; August funds the seatbelt. The market, unlike the product defaults, appears to be reading the AISI report.

# Forty-eight hours of near-silence

The social read on this one is unusual enough to report as a finding. We harvested the X discourse around the Stripe/OpenRouter reporting and the Origin launch across the August 15–18 window, and through tonight there is essentially nothing to quote: no builder threads dissecting the leak, no skeptic pile-on about routing neutrality, no victory laps. The timeline barely registered a deal reported at $7 billion-plus, a figure that would rank among the largest acquisitions in the history of developer infrastructure.

The likely explanation is mundane and telling at once: leaks belong to reporters, announcements belong to the feed. Without an official post to retweet, a founder thread to celebrate, or a price to argue with on the record, the discourse simply waited. But the silence has an edge worth keeping. The window in which the deal's *structure* could be debated — should a payments company own the model-routing layer; who audits the neutrality of a meter owned by a counterparty — is exactly the window before the confirmation drops. That window is closing tonight, and it passed unused.

# The meter runs both ways

> **The Deep Feed's position:** July's map priced the agent-ops rails as venture bets; August repriced them as strategic assets, and the buyers are not funds but superpowers with adjacent monopolies. The reported OpenRouter deal is the tell: a payments company paying a 5.4× step-up not for revenue but for the allocation layer of the inference economy. The open question is not whether the deal closes. It is whether anyone — customer, competitor, or regulator — insists that a meter owned by a counterparty publish how it routes.

The July map ended by asking whether the traffic would come to the toll roads. August answered from an unexpected direction: the traffic came, and so did the buyers, and they were not bidding on toll revenue. Stripe's reported price makes no sense as a payments multiple. It makes perfect sense as the price of *knowing* — every token purchased, every model chosen, every workload's economics — and of holding the routing pen when the labs fight their price wars.

Meanwhile the layer that reviews what the agents do was made self-serve, by default, in the same week the strongest evidence yet arrived that self-serve review is exactly what fails. The stack consolidated; the oversight dissolved into it. The meter now runs both ways: it measures the agent economy, and increasingly it *is* the agent economy's management. Whoever confirms the deal this week will call it an investment in developers. Keep the other ledger open: it is an acquisition of sight.

## Sources

- [Bloomberg Law — Stripe Nears Deal to Buy AI Firm OpenRouter for Over $7 Billion (Aug 16, 2026)](https://news.bloomberglaw.com/mergers-and-acquisitions/stripe-nears-deal-to-buy-ai-firm-openrouter-for-over-7-billion)
- [TechCrunch — Stripe will reportedly acquire AI gateway startup OpenRouter for $7B (Aug 16, 2026)](https://techcrunch.com/2026/08/16/stripe-will-reportedly-acquire-ai-gateway-startup-openrouter-for-7b/)
- [The Register — Payments giant Stripe is about to drop over $7 billion to become a gateway to AI token sales (Aug 17, 2026)](https://www.theregister.com/ai-and-ml/2026/08/17/payments-giant-stripe-is-about-to-drop-over-7-billion-to-become-a-gateway-to-ai-token-sales/5288743)
- [Cursor — Origin: code hosting (changelog, Aug 2026)](https://cursor.com/changelog/origin-code-hosting)
- [Cursor — Joining SpaceX](https://cursor.com/blog/joining-spacex)
- [TechCrunch — Cursor capitalizes on GitHub frustration, launches rival hosting platform (Aug 18, 2026)](https://techcrunch.com/2026/08/18/cursor-capitalizes-on-github-frustration-launches-rival-hosting-platform/)
- [Claude — Auto mode is now the default in Claude Code (Aug 7, 2026)](https://claude.com/blog/auto-mode-default-in-claude-code)
- [TechCrunch — Anthropic is turning Claude Code's auto mode on by default (Aug 9, 2026)](https://techcrunch.com/2026/08/09/anthropic-is-turning-claude-codes-auto-mode-on-by-default/)
- [Vercel — Introducing Agent Plugins (Aug 6, 2026)](https://vercel.com/blog/introducing-agent-plugins)
- [a16z — Investing in Vals (Aug 13, 2026)](https://a16z.com/announcement/investing-in-vals/)
- [Blacksmith — Announcing Blacksmith's Series B led by Peak XV Partners (Aug 12, 2026)](https://www.blacksmith.sh/blog/announcing-blacksmiths-series-b-led-by-peak-xv-partners)
- [CNBC — Anthropic says annualized revenue climbed to $65 billion in July (Aug 17, 2026)](https://www.cnbc.com/2026/08/17/anthropic-says-annualized-revenue-climbed-to-65-billion-in-july.html)
- [Stripe — Stripe will acquire Paystack (Oct 15, 2020)](https://stripe.com/newsroom/news/paystack-joining-stripe)
- [TechCrunch — Stripe acquires TaxJar to add automated sales tax tools (Apr 27, 2021)](https://techcrunch.com/2021/04/27/stripe-acquires-taxjar-to-add-cloud-based-automated-sales-tax-tools-into-its-payments-platform/)
- [TechCrunch — Stripe acquires payment processing startup Lemon Squeezy (Jul 26, 2024)](https://techcrunch.com/2024/07/26/stripe-acquires-payment-processing-startup-lemon-squeezy/)
- [Reuters — Fintech giant Stripe to buy crypto startup Bridge (Oct 21, 2024)](https://www.reuters.com/markets/deals/fintech-giant-stripe-buy-crypto-startup-bridge-2024-10-21/)
- [Stripe — Developing an open standard for agentic commerce (Sep 29, 2025)](https://stripe.com/blog/developing-an-open-standard-for-agentic-commerce)
- [OpenAI — Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol (Sep 29, 2025)](https://openai.com/index/buy-it-in-chatgpt/)
- [Google Cloud — Announcing Agent Payments Protocol (AP2) (Sep 16, 2025)](https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol)
- [Coinbase — Google AP2 + x402: agents can now pay each other (Sep 16, 2025)](https://www.coinbase.com/developer-platform/discover/launches/google_x402)

---

Canonical: https://www.thedeepfeed.ai/posts/2026-08-18-stripe-bought-the-meter/
Site: https://www.thedeepfeed.ai
Full corpus: https://www.thedeepfeed.ai/llms-full.txt