AI NewsGo-to-marketReported
Corma raises $125M at a $1B valuation, three months out of stealth
Corma, a defensive cybersecurity AI startup, raised a $125 million Series A led by NEA at a $1 billion valuation.

Image: TechCrunch
Why it mattersA security buyer now picks between in-house teams, the big AI labs, and a startup with $185 million. Corma did it in three months.
A cybersecurity company that came out of stealth in August is worth a billion dollars three months later. On 16 November 2026, Corma announced the round on its own site. The home page says the company is announcing a $125 million Series A. Corma says the money will fund its work "toward super intelligence for defensive cybersecurity."
The round was led by NEA. TechCrunch, which broke the story, reports a valuation of $1 billion. The total across all rounds is $185 million.
The round and the investors
The Series A follows an earlier, unpublished round. Corma's home page lists four backers. They are NEA, Coatue, Sequoia and Khosla Ventures. TechCrunch's Julie Bort reports the stage as Series A at a $1 billion valuation. She reports the total across all rounds at $185 million.
Bort writes that AI cybersecurity startup Corma has raised "an $125 million Series A round" led by NEA.
Corma's site does not state the valuation on its public pages. The company has not named its founders in its own copy. TechCrunch does not name them either.
What Corma says it sells
Corma trains its own AI models. The product sits next to a security team. Corma calls it an "AI Workforce". The home page says Corma is training foundation models "toward superintelligence for defensive cybersecurity."
It describes the product as "a superintelligent security workforce."
Corma argues that offence and defence are different problems. Its home page says offence and defence "are not two sides of the same problem."
It adds that "cybersecurity is not symmetric." That is Corma's own framing. It explains why it thinks a specialist model is worth paying for.
The benchmark Corma published
Corma's page carries a chart it calls DefenseBench V5. Corma says the chart shows that Corma v0.4 leads "on defensive task performance and execution speed."
The rivals named on the chart are Claude Fable 5.2, Gemini 4 Carbon, GPT-6 Astra, and four others. Also named are GPT-6.1 Sol, GPT-6.1 Terra, Kimi K3.1 Pro and Grok 4.7.
This is Corma's own benchmark. No third party has published figures for the same tasks. A buyer who wants to compare the model to Claude or GPT should ask for the raw task set. They should also ask for the scoring rules.
A buyer picking a defensive security AI today now has three kinds of seller. The in-house security team, as before. The large AI labs, with general models the buyer scores on its own data. And a growing group of startups trained for the job, of which Corma is the newest expensive one. $185 million buys enough compute to train more rounds of its own models. NEA leading at a $1 billion valuation sets a floor on what the next round must beat. The price pressure moves onto the general-purpose models. Their sellers now have to answer why a dedicated defender is not the safer buy.
Source
Corma, home page announcement. TechCrunch also reported the round with the $1 billion valuation figure.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.


