
Cyber startup Corma goes from stealth to unicorn in just three months with $125 million Series A
Founded in 2025, the Israeli startup has raised $185 million in total funding, including a $60 million Seed round announced in August. NEA led the latest financing, which values the company at $1 billion.
Cybersecurity company Corma has raised $125 million in a Series A funding round at a valuation of $1 billion. The financing comes just three months after the company announced its $60 million Seed round in August, bringing its total funding to $185 million. The Series A was led by NEA, with participation from existing investors Sequoia Capital, Khosla Ventures and Coatue.
AI-powered attacks have become one of the most pressing threats facing businesses and governments worldwide. Against this backdrop, the world's largest enterprises are racing to integrate Corma's foundation model into their cybersecurity operations. Since emerging from stealth in August, the company says demand has surged among Fortune 100, Fortune 500 and Fortune Global 500 companies. To meet that demand, it has expanded its server capacity 24-fold.
Corma works with large enterprises across healthcare, financial services, energy and critical infrastructure, retail, and telecommunications. These organizations are increasingly delegating security tasks to Corma's autonomous AI workforce, powered by the company's proprietary foundation model.
"We began back in August, and we are in a very fast-paced race. Every day, we see further evidence that we are in a race between attackers and defenders. With current models, attackers hold a massive advantage. We must advance as quickly as possible to give defenders the ability to win," Alon Pluda, Corma's co-founder and CEO, told Calcalist.
"We are training foundation models specifically for cybersecurity, models that are an order of magnitude better. We are training them for defense, and our mission is to keep advancing all the way toward 'superintelligence' for defensive cybersecurity. The models currently on the market are excellent at coding, but they lack the specific capabilities required for defensive cybersecurity. As they improve their coding proficiency, they also become better at identifying vulnerabilities. Yet defensive cybersecurity demands an entirely different set of skills from offensive cybersecurity.
"No existing model was built for defense. While they are highly capable when it comes to attacks, they lack the capacity to keep pace with the current rate of evolution. We don't offer a standard security product like other companies. Rather, we provide an AI-powered workforce, driven by our foundation model, capable of handling security tasks across the entire spectrum while integrating with a company's existing security stack. Our agents work with all these products, and the entire process is fully autonomous, with no human intervention involved," Pluda said.
"We are leading the race. We must move forward as quickly as possible. Investors look at Corma and clearly understand that the cybersecurity field requires the kind of intelligence we offer. We have the best investors in the world, and some of the most significant companies are already using our technology. Our revenue is growing very rapidly, to the point that anything we say today might already be outdated," he added.
Corma is developing an end-to-end autonomous workforce for cybersecurity, powered by a proprietary foundation model designed specifically for defensive security. Its agents operate alongside security teams, using an organization's existing tools to investigate threats, carry out proactive security measures and handle a broad range of defensive tasks.
The company is training its models with the goal of achieving what it calls "superintelligence" for defensive cybersecurity, enabling security teams to operate at the speed and scale needed to counter advances in AI-driven attacks.
Founded in 2025 by Pluda, a graduate of the Psagot program and an alumnus of Unit 8200, Corma employs approximately 25 people, with operations centered in Tel Aviv and San Francisco. Its team combines expertise in frontier AI research, offensive security and defensive cybersecurity. It includes researchers from Google and DeepMind, alongside cybersecurity specialists from Israel's Unit 8200. The company says this combination of expertise in foundation-model development and hands-on cyber operations enables it to train and evaluate its models and deploy them in the complex security environments of large organizations.
When Corma emerged from stealth in August, it warned of a widening gap between AI-driven offensive capabilities, which are advancing rapidly, and defensive capabilities, which have struggled to keep pace. In the months since, the company says, this imbalance has become increasingly apparent. October saw a wave of critical vulnerabilities and breaches affecting some of the world's most widely used software systems, an episode dubbed "Black October." The incidents prompted urgent responses from security teams and warnings from governments, including in Washington and Tokyo.
The gap between offensive and defensive capabilities is partly rooted in the direction of the broader AI race. Coding has become a central focus of AI development, with leading research labs devoting substantial resources to building models that can write and improve software. But an AI system that learns to write the software powering the digital world can also learn how to exploit it.
Security vulnerabilities are, in essence, another type of software bug. As models become better at writing code, they also become more capable of identifying vulnerabilities. When combined with AI agents, these models can potentially exploit the weaknesses they uncover and execute attacks from beginning to end, as the incidents associated with "Black October" illustrated.
Cyber defense, however, requires a different set of capabilities. Much of the work extends beyond code security and involves analyzing vast quantities of information, including audit logs, security events and network traffic. Security teams must identify subtle indicators across different systems over extended periods, filter out enormous amounts of background noise and maintain consistency across thousands of sequential decisions.
Models trained primarily for coding were not designed to handle these tasks. As a result, the same advances that make AI increasingly effective at identifying and exploiting vulnerabilities do not necessarily translate into comparable improvements in defensive cybersecurity.
Corma was built to address precisely this challenge. It trains foundation models specifically for cyber defense and deploys them as an AI-driven workforce that operates alongside security teams, uses organizations' existing tools and carries out a wide range of defensive security tasks.
"Cybersecurity is now a battle of intelligence against intelligence, and Black October made that impossible to ignore," Pluda said in a statement accompanying the funding announcement. "Since we came out of stealth, demand from the world's largest organizations has been overwhelming, and one of our hardest problems has been securing capacity fast enough to keep up with the demand."














