Key takeaways
- Corma raised a $125 million Series A at a $1 billion valuation on November 16, 2026, led by NEA.
- It has raised $185 million in total, three months after a $60 million seed led by Sequoia.
- Corma trains its own models for cyber defense and runs them as AI agents alongside security teams.
Corma, a startup that trains its own AI models to defend companies against cyberattacks, raised a $125 million Series A at a $1 billion valuation on November 16, 2026. NEA led the round, and Sequoia Capital, Khosla Ventures and Coatue joined. Corma has now raised $185 million, including the $60 million seed it announced when it came out of stealth in August.
Three months from seed to a billion-dollar valuation is fast even for AI. The Wall Street Journal reported the round, and Israel's Globes added that the company has about 20 employees in Tel Aviv and San Francisco. The money is following a fear: that AI is making attackers stronger faster than it is making defenders stronger.
Corma trains its own models instead of renting them
Rather than build on models from Anthropic, OpenAI or DeepSeek, Corma trains its own foundation models for defensive security, Globes reports, and runs them as what it calls the Corma Workforce: AI agents that sit alongside a company's security team, use the tools the team already has and take on tasks from start to finish.
The company says the agents get more capable with each new model version and that security teams hand them more work over time. It says it will share its next generation of models soon. It has not published benchmark results, so there is no independent way yet to compare its models with general ones.
Customers and demand
Corma says some of the world's largest companies already use the Workforce and that it has a waitlist. It names none of them. Globes reports the customers include Fortune 100 and Fortune 500 companies in healthcare, finance, energy, retail and telecommunications, and that annual recurring revenue is estimated in the tens of millions of dollars. Corma's own post calls compute its biggest bottleneck, and the new money goes to training bigger models, buying more compute and hiring selectively.
Why investors are paying up now
The round lands on the same day Google's threat intelligence group said it had seen attackers use an AI framework to turn security patches into working exploits with no human involved. Corma's post points to a string of serious vulnerabilities and breaches in October that the industry called "Black October". If finding and exploiting bugs gets cheap, defenders need software that works at the same speed, and that is the bet NEA is making.
What to watch: whether Corma publishes evaluations of its next models, and whether a named customer says what the agents actually caught.
Sources
- Three months inCorma
- Cyber startup Corma raises $125 million for defensive AI agentsThe Wall Street Journal
- Defensive AI agents co Corma raises $125m at $1b valuationGlobes

