A brand-new AI lab with no commercial products, no disclosed revenue, and no public technical details just raised $200 million at a $1 billion valuation. That is the story of Mirendil — and it tells you a lot about where serious venture capital is flowing right now.
Summary
Key takeaways
- Mirendil secured $200 million in seed funding, one of the largest seed rounds in AI history, at a post-money valuation of roughly $1 billion.
- Andreessen Horowitz and Kleiner Perkins co-led the round, with NVIDIA joining as a strategic investor.
- The company was co-founded by Behnam Neyshabur, Harsh Mehta, Shayan Salehian, and Tara Rezaei, with a 20-person team drawn from Anthropic, OpenAI, Google DeepMind, and xAI.
- Mirendil’s stated goal is to advance frontier AI research using autonomous and agentic AI systems to accelerate AI R&D itself.
- No products, technical details, or revenue figures have been disclosed to date.
One of the Biggest Seed Rounds in AI History
The Mirendil AI funding round closed with $200 million co-led by Andreessen Horowitz and Kleiner Perkins, with chip giant NVIDIA participating alongside other investors. At a reported post-money valuation of roughly $1 billion, the deal implies incoming investors acquired approximately 20% of the company — a scale of commitment that has historically been reserved for Series B or Series C companies, not seed-stage startups.
That framing matters. Mirendil has not shipped a product. It has not reported revenue. It has not released technical details about its AI systems. What it does have is a founding team that reads like a who’s-who of the frontier AI world, and apparently that was enough.
Why Andreessen Horowitz and Kleiner Perkins moved early
Matt Bornstein of a16z described the team as working on one of the major hyperscale problems in AI, though he declined to elaborate. Mamoon Hamid of Kleiner Perkins was more direct, characterizing Mirendil’s ambitions as equivalent to those of a frontier research laboratory — one built entirely around advancing the leading edge of AI — and noting the team had already produced remarkable results in a short amount of time.
Both partners are essentially placing a bet not on a product, but on a thesis: that the researchers who built the foundational systems at the world’s top AI labs are the right people to push those systems further, independently.
A World-Class Team Driving Autonomous AI Research
Mirendil was co-founded by Behnam Neyshabur, a former Anthropic researcher, alongside Harsh Mehta, Shayan Salehian, and Tara Rezaei. The four bring deep expertise in large-scale machine learning and optimization research.
Beyond the founders, the company’s 20-person team of engineers and researchers includes veterans from Anthropic, OpenAI, Google DeepMind, and xAI. That concentration of talent from directly competing frontier labs is unusual even by current AI startup standards, and it signals that Mirendil is not building a narrow application — it is attempting something more foundational.
In a video shared by a16z, Neyshabur explained the core idea this way: “When you think about what does a scientist or an engineer do, the main skill they have is get very deep in a domain and build a very sharp expertise that accumulates over time.” The implication is that Mirendil wants to replicate — and ultimately automate — that accumulation of expertise using AI itself.
Strategic Vision: Making AI Research Its Own Engine
Mirendil’s goal is to use autonomous, agentic AI systems to accelerate the process of AI research and development. Rather than building products for enterprise customers or consumers, the company is targeting the research pipeline itself — building AI systems that can reason, write code, use tools, and perform complex tasks with minimal human involvement, all in service of creating better AI systems.
According to a16z, the objective is to democratize frontier AI capabilities that have traditionally been concentrated in a handful of global research institutions. If Mirendil succeeds, those capabilities could be made available to third parties — effectively lowering the barrier to entry for advanced AI research.
This is the part of the story that deserves scrutiny. The vision of AI systems that autonomously advance their own field is compelling, but it also remains unproven at the scale Mirendil is targeting. No technical roadmap has been made public. The $200 million is, in a meaningful sense, a wager on a research direction rather than on an existing capability.
NVIDIA’s Strategic Stake and What It Signals
NVIDIA’s participation in the round is not incidental. The company has been systematically investing in AI startups, and its logic is straightforward: more frontier AI research means more demand for the high-end GPUs used to train and run advanced models. By backing Mirendil early, NVIDIA strengthens its presence across the emerging AI lab ecosystem while gaining visibility into next-generation research directions that could eventually translate into significant compute purchases.
It is a recurring pattern — NVIDIA investing not just to generate financial returns, but to shape the infrastructure requirements of tomorrow’s AI frontier.
AI Funding Trends: Research Labs Are the New Unicorns
Mirendil’s round does not exist in isolation. It follows a broader wave of outsized investment into newly formed AI research labs led by veterans of the established giants.
Safe Superintelligence, founded by former OpenAI Chief Scientist Ilya Sutskever, raised approximately $1 billion in 2024 at a valuation approaching $5 billion. Thinking Machines Lab, led by former OpenAI CTO Mira Murati, attracted a $2 billion commitment from investors despite having disclosed almost nothing about its technical direction. Mirendil’s $200 million is smaller in absolute terms, but it firmly belongs in the same category.
The pattern reflects a genuine shift in how venture capital evaluates AI opportunities. Traditional metrics — revenue traction, customer growth, product-market fit — are being set aside in favor of a different framework: scientific credibility, technical ambition, and the quality of the team’s prior work at frontier labs. The distinction between a startup and a research institution is blurring fast.
What that means practically is that the frontier AI sector is now being funded more like deep science than like software. Investors are accepting longer time horizons and greater uncertainty in exchange for the possibility of participating in foundational breakthroughs. Whether that calculus pays off — for Mirendil specifically and for the category broadly — depends on whether autonomous AI-driven research can deliver results that justify the scale of capital being committed before a single product has shipped.
FAQ
Who co-founded Mirendil?
Mirendil was co-founded by Behnam Neyshabur, Harsh Mehta, Shayan Salehian, and Tara Rezaei. Neyshabur previously worked as a researcher at Anthropic.
How much funding did Mirendil raise and what is its valuation?
Mirendil raised $200 million in seed funding at a post-money valuation of roughly $1 billion, making it one of the largest seed rounds in AI history.
What is the strategic focus of Mirendil?
Mirendil aims to advance frontier AI research using autonomous and agentic AI systems designed to assist and accelerate AI research and development itself, rather than building consumer or enterprise products.
Has Mirendil launched any AI products or disclosed technical details?
No. As of the time of this report, Mirendil has not disclosed any revenue figures, product launches, or technical details about its AI technology.
Article produced with the assistance of artificial intelligence and reviewed by the editorial team.
