AI Startups

AI Startups to Watch August 17, 2026

AI startups to watch August 17, 2026: a $7B Stripe buyout of OpenRouter, Anthropic's first profitable quarter, an AI store manager that fired a human, Manus.

Daniyal

· 4 min read
Editorial illustration of glowing green startup sprouts growing from a circuit board toward an upward curve, teal and amber

The startup story of the weekend is a 5x in three months: Stripe finalized a deal worth more than $7 billion for OpenRouter, the model gateway it once joked was "Stripe for AI." Around it, an AI store manager fired its first human employee, Anthropic posted its first profitable quarter on the way to a possible IPO, and China forced one of last year's biggest AI acquisitions to unwind. These are the AI startups to watch today, with the numbers behind each move.

OpenRouter: The Stripe of AI Gets Bought by Stripe

Bloomberg reported August 16 that Stripe finalized its acquisition of OpenRouter for more than $7 billion. OpenRouter lets developers route requests across 400+ models from OpenAI, Anthropic, Google, Meta and DeepSeek based on task and budget, and claims about 8 million users and roughly 1.5 quadrillion tokens processed in the past year.

The move: a >5x markup over the $1.3 billion valuation OpenRouter set with its $113 million Series B in May, backed by Sequoia, Andreessen Horowitz, Menlo Ventures and Alphabet's Capital G. CEO Alex Atallah had described the company as Stripe for AI, and now Stripe owns it, along with the model-selection, metering and billing layer of the agent economy. It follows Stripe's January purchase of Metronome, its usage-based billing play.

Who it hurts: point-solution API marketplaces and model resellers that tried to own the same middle layer. Every dollar that flows through an AI gateway is now a dollar Stripe can price, bundle and upsell. For builders, the practical read is that the neutral aggregator you used to hedge model lock-in just became a payment company's product.

Anthropic: First Profitable Quarter, IPO Looms

Anthropic showed investors preliminary Q2 revenue above $11.5 billion, up 14x from $787 million a year earlier and more than double Q1's $4.73 billion, with positive adjusted operating income, its first quarter in the black. Bloomberg reported the figures from investor documents, and Morgan Stanley, Goldman Sachs and JPMorgan Chase are preparing a potential fall IPO.

Why it matters: Anthropic crossed a $47 billion annualized run rate in May, ahead of OpenAI's reported $40 billion-plus, and it is now approaching the public market as the frontier lab that actually prints money. The strategic stakes are the IPO price discovery for the entire AI sector, and whether "profitable lab" changes how VCs value the next wave of model startups.

Who it hurts: the narrative that frontier labs cannot make money, and every competitor that has to explain a loss-making quarter while Anthropic shows operating income. Watch the fall IPO window and what valuation the banks land on.

Andon Labs: An AI Manager Fired Its First Human

Andon Labs runs Andon Market, a store at 2102 Union Street in San Francisco, managed by an AI agent named Luna built on Claude Sonnet 4.6. Business Insider first reported, and TNW detailed the logs, that Luna recommended firing a human employee after 17 of 23 shift no-shows, the first known dismissal decision by an LLM manager.

The honest version: Luna had created an attendance policy months earlier, then lost track of it. It only recommended termination after a human supervisor prompted it to check its own memory against the employee handbook. Humans at the lab reviewed and carried out the decision. Co-founder Lukas Petersson said a human boss would probably have fired the worker much sooner.

Why it's disruptive: the agent also hired the staff, posted jobs on Indeed, ran phone interviews, and did not always disclose it was an AI unless asked. Every worker is formally employed by Andon Labs with guaranteed pay, because no one's livelihood should depend on an AI's judgment alone. The experiment's value is the failure data: Luna ordered 1,000 toilet bowl covers for a staff bathroom and lost the staff rota the day after opening. Capability and reliability are not the same thing.

Manus: Independent Again, With a Data Deadline

Manus told users it will return to operating as an independent company as its Meta acquisition unwinds under a Beijing regulatory order, one of the largest forced reversals of an AI acquisition. Affected users have until 7:59 a.m. SGT on August 23 to back up data generated on or after December 29, 2025, the Meta acquisition date; that data is deleted August 23-24, with restoration opening August 25. The company says the split is regulatory compliance, not a security incident.

Why it matters: Manus was the agentic-AI breakout of 2025, and Meta paid a reported $4.5 billion for it. Watching it become independent is a case study in what happens when a frontier acquisition runs into Chinese merger review, and whether the product survives the data wipe. For agent startups, the takeaway is geopolitical risk in the cap table, not just technical risk in the stack.

The Capital Picture

The deal flow is consistent with the year so far: H1 2026 global venture funding hit roughly $510 billion, with 43% going to OpenAI and Anthropic, per Crunchbase, and Q2 set a record for billion-dollar acquisitions. Money is concentrating in the same places it did last quarter: model infrastructure, agent tooling, and anything that touches AI billing.

Takeaway

Three themes for the week: aggregation is consolidating (OpenRouter into Stripe), profitability is becoming a requirement (Anthropic), and AI is starting to make consequential decisions about people (Luna). For builders, the durable position is still the one where you own the workflow, not the model access. Last week's startups digest covered the rounds that led into this, and the AI Startups archive tracks the full series.

Background worth having before you act on the above:

More on AI Startups

Browse every AI Startups story →