Artificial intelligence startup Mirendil has raised US$1 billion in Series A funding to accelerate the development of AI-powered research tools, marking one of the largest early-stage financing rounds in the rapidly expanding artificial intelligence industry.
Announced on 12 November 2026, the investment was co-led by venture capital firm Kleiner Perkins and private equity firm Clearlake Capital. Other participating investors included Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC and S32, alongside additional new and existing investors.
The San Francisco-based company is developing artificial intelligence systems designed to automate parts of the research and development process, allowing scientists and engineers to undertake complex technical projects with smaller teams.
Founded by former Anthropic researchers Behnam Neyshabur and Harsh Mehta, Mirendil is pursuing an approach in which AI agents can write code, conduct experiments, analyse results and support the development of increasingly capable AI models.
The company’s ambition is to make advanced AI research accessible beyond the small number of organisations currently possessing the financial resources, computing infrastructure and specialist talent required to develop frontier models.
The latest investment follows a a US$200 million seed funding round announced in June 2026, bringing Mirendil’s publicly reported funding to approximately US$1.2 billion.
Although the company has not officially disclosed its valuation, reports have placed the latest financing at approximately US$5 billion.
The substantial investment reflects growing interest in artificial intelligence systems capable of assisting with the development of other AI technologies, an emerging area commonly associated with automated AI research and self-improving systems.
Mirendil recently demonstrated its approach through the development of Mirendil-Base, a coding-focused AI model that the company says was pretrained by one researcher working with AI agents over three and a half months.
According to the company, the process required less than US$10 million in computing resources, illustrating its ambition to reduce the human resources traditionally required for advanced model development.
The company subsequently developed Fern, an AI model supporting a context window of up to 10 million tokens. This capacity is intended to enable extended tasks requiring large amounts of information and sustained processing.
Mirendil has also introduced Lor, an AI research and development platform currently available in private beta.
Lor allows hundreds of AI agents to undertake coding and experimental tasks over extended periods, supporting workflows that would otherwise require substantial human coordination.
The company uses the platform internally to support its own research activities and intends to make similar capabilities available to external researchers and engineering teams.
The development of such systems could have implications beyond the artificial intelligence industry.
Automated research tools may eventually support scientific discovery in areas including healthcare, biotechnology, materials science and engineering, where complex experiments and large volumes of data can make research expensive and time-consuming.
However, the effectiveness of Mirendil’s approach across different scientific disciplines has yet to be independently established at scale.
The company has also invested in computing infrastructure to support its development plans.
In August 2026, Mirendil entered into a multiyear agreement with Google Cloud valued at more than US$100 million, providing access to computing resources including Google’s Tensor Processing Units and Nvidia graphics processing units.
The arrangement gives Mirendil access to different types of AI hardware, allowing it to optimise computing workloads and support increasingly demanding research operations.
Google’s participation in the latest funding round further strengthens the relationship between the technology company and the emerging AI research startup.
The US$1 billion Series A also highlights a broader shift in venture capital investment, with substantial funding increasingly concentrated among companies developing advanced AI models, infrastructure and autonomous agent technologies.
While traditional early-stage funding rounds typically provide startups with capital to establish products and expand operations, the computing requirements of frontier AI development have contributed to unusually large financing rounds for relatively young businesses.
For investors, the potential commercial opportunity lies in enabling organisations to conduct sophisticated research without building large internal AI laboratories.
If Mirendil’s technology proves effective, it could allow smaller research teams, universities and businesses to undertake projects that currently require significant technical resources.
At the same time, the development of increasingly autonomous AI research systems raises questions about reliability, oversight and the ability to evaluate technologies that may contribute to their own improvement.
The latest financing provides Mirendil with substantial resources to continue developing its models, expand its research infrastructure and explore commercial applications.
Its progress will depend on demonstrating that automated AI research can consistently deliver reliable results while reducing the cost and complexity of developing advanced systems.
The investment represents another significant development in the global AI industry, where competition is increasingly focused not only on creating more capable models but also on developing technologies that can accelerate the research process itself.
