A nine-month-old startup with no product in market just raised $94 million at a $650 million valuation. OpenAI is among the investors. The founders are 23 years old.
Meet Isara — and the bet that the next AI breakthrough isn’t about making models bigger, but about making thousands of agents work together.
What Isara Is Building
The core idea: multi-agent coordination at scale. While most AI applications involve one model responding to one prompt, Isara’s architecture coordinates hundreds or thousands of specialized agents that communicate, divide tasks, align on goals, and produce combined output.
Their demonstration so far: approximately 2,000 agents working in concert to forecast the price of gold.
The initial commercial targets are investment firms paying for predictive modeling. Biotech and geopolitical analysis are secondary markets. The longer vision: agent swarms that track geopolitical shifts and forecast economic trends.
The Founders
Eddie Zhang, a former OpenAI AI safety researcher, left to start Isara in June 2025 alongside Henry Gasztowtt, a computer science student at Oxford. They co-authored a paper at ICML 2024 on how AI systems could cooperate to improve policymaking — the intellectual foundation for the company.
They’ve since hired roughly a dozen researchers from Google, Meta, and OpenAI itself.
The Investor List
The round, first reported by The Wall Street Journal, includes:
- OpenAI — strategic bet on multi-agent coordination
- Amity Ventures
- Michael Ovitz — former CAA chairman, early Uber backer
- Stanley Druckenmiller — billionaire hedge fund manager
Smart money, concentrated bets, and an OpenAI co-sign.
Why This Matters: The Multi-Agent Coordination Problem
Here’s the gap between the demo and the reality:
Getting a single AI agent to perform reliably on a complex task is already hard. Getting thousands to communicate without cascading errors, conflicting objectives, or compounding hallucinations is a problem the academic literature has barely begun to address at scale.
Multi-agent frameworks exist — LangChain, CrewAI, AutoGen — but they typically coordinate small numbers of agents on structured tasks. OpenClaw users running multi-agent setups know this firsthand: even a 5-agent team requires careful orchestration to avoid duplicated work, conflicting state, and runaway loops.
Isara’s ambition to orchestrate thousands on open-ended analytical problems is a different order of difficulty entirely.
The Neolab Phenomenon
Isara is part of a pattern The Information has labeled “neolabs”: research-heavy AI startups founded by alumni of OpenAI, DeepMind, Anthropic, and Google Brain that operate more like privately funded research institutions than traditional companies.
The numbers are staggering:
- $2.5 billion into five such startups in just over a month
- $10+ billion total into the neolab category
- Valuations in the hundreds of millions before revenue
The investor thesis: foundational research capability, not a specific product, is the scarce asset worth funding. In a market where Cognition (Devin) reached a $10.2 billion valuation with $73M ARR, the logic is that a genuine breakthrough in multi-agent coordination could be worth far more.
The risk: foundational research is, by definition, uncertain. The gap between a demo forecasting gold prices and a production system investment firms rely on for real capital allocation is large enough to swallow a $94M funding round.
Why OpenAI Invested in Its Own Alumnus
The question is familiar: why does a frontier lab back a startup founded by its own former employee to work on adjacent problems?
Strategic optionality. If multi-agent coordination becomes critical, OpenAI wants exposure to approaches developing outside its walls. At $94M on a $650M valuation, it’s cheap insurance relative to OpenAI’s $300B valuation.
Talent retention through investment. The AI industry’s scarcest resource isn’t compute — it’s the researchers who know how to use it. By investing in Isara, OpenAI maintains a relationship with Zhang’s team rather than losing them to competitors. The same dynamic drives Google, Microsoft, and Amazon investments into smaller labs.
What This Means for OpenClaw Users
Multi-Agent Is Going Mainstream
The agentic AI market is projected to grow from $7.8B (2025) to $52.6B (2030). Every major platform is shipping multi-agent features. The question isn’t whether multi-agent coordination matters — it’s who solves it at scale.
The Coordination Problem Is Real
If you’ve run even a small multi-agent team on OpenClaw, you know the pain: agents stepping on each other’s work, state conflicts, duplicated effort, runaway execution. Isara’s thesis validates that this isn’t a UI problem — it’s a fundamental architectural challenge that needs new approaches.
The “Neolab” Wave Creates Opportunities
As research labs commercialize multi-agent coordination, the tooling and frameworks that emerge will flow downstream to platforms like OpenClaw. Better orchestration primitives, better inter-agent communication protocols, better conflict resolution — all of this benefits the broader ecosystem.
But Also: Hype Risk
A $650M valuation for a 9-month-old company with no product is a signal of either extraordinary potential or extraordinary froth. The honest answer is that nobody knows which yet. The ICML paper is promising. The 2,000-agent demo is impressive. But production reliability at that scale, on tasks where accuracy matters for real capital allocation? That’s unproven.
The Bottom Line
Isara represents the most ambitious bet in the multi-agent space to date. If they succeed in making thousands of agents coordinate reliably on complex analytical tasks, it changes what’s possible for every agent platform — including OpenClaw. If they don’t, it becomes another cautionary tale about funding research dreams at production valuations.
Either way, the signal is clear: the industry believes the next breakthrough in AI isn’t about individual model capability but about collective agent intelligence. That’s a thesis worth watching.
Sources: The Next Web, TechFundingNews, Wall Street Journal (original report)