Corma Raises $125M Series A for Defensive Cybersecurity
Corma has chosen to own the foundation model behind its security workforce. That decision ties the company's ability to serve another enterprise to the research and computing capacity needed to keep the workforce improving, a substantial commitment inside a cybersecurity software business.
On November 16, 2026, Corma announced a $125M Series A led by NEA, with returning investors Sequoia Capital, Khosla Ventures and Coatue. The financing values the company at $1B and brings total funding to $185M, following the $60M disclosed when it emerged from stealth in August. Founder and CEO Alon Pluda is building the company around models trained specifically for defensive cybersecurity.
For enterprise security teams, Corma's proposition reaches into how work gets assigned and completed across existing tools. For its investors, the financing supports a company that must advance its underlying models while supplying enough compute to meet demand. Corma says the proceeds will expand model training, increase computing capacity and support selective hiring, putting infrastructure spending close to the center of this funding story.
What Corma's Series A finances
Corma's November announcement describes demand for its models and a waitlist that the company is working to accommodate. Those are company accounts of commercial interest, without a disclosed customer count or a public revenue figure in the announcement. They give the use of proceeds a practical shape: customers need capacity before a larger research ambition can become useful work inside their organizations.
The distinction between the round and the total also matters. The new investment is $125M; the $185M figure includes earlier funding. The $1B valuation describes the financing's valuation rather than additional cash, and the announcement does not specify whether it is a pre-money or post-money figure. Corma presents the amounts with a dollar sign without identifying the denomination, so no currency conversion is applied here.
Sequoia's August investment account records its role leading the seed round. NEA now leads the Series A alongside the returning firms, extending the financial backing behind the specialized-model approach. The sequence gives Pluda resources to pursue research and deployment together, while leaving the allocation among training, serving customers and hiring undisclosed.
Why defensive AI demands its own training
Corma and its investors argue that defending an enterprise requires capabilities that general-purpose models encounter less often in training. A defender works through security events, logs and configuration state, looking for meaningful activity inside an environment that is mostly behaving normally. The relevant evidence can be scattered across products, with an investigation requiring many connected decisions.
NEA's investment essay, written by Aaron Jacobson, Thomas Joshi and Lila Tretikov, describes Corma's use of reinforcement learning in environments resembling enterprise networks. The investment thesis rests on teaching a model to perform defensive work against outcomes that can be checked. NEA also describes treating security data as a distinct modality and maintaining consistency through the decisions involved in an investigation.
Those descriptions explain the engineering direction; the investor has a financial interest in its success. Corma's aspiration toward superintelligence should be read as a development ambition, while its product claims require evaluation in the setting where a buyer plans to use them. The useful commercial question is how the system handles that organization's tools, evidence and authority, including the moments when an uncertain finding needs a person with local knowledge.
An AI workforce inside the existing security stack
The Corma Workforce is designed to operate through tools a security team already uses. Corma describes software that investigates, connects evidence, takes actions and communicates through channels such as Slack and Teams. Its product presentation includes human review and approval, placing the team inside the workflow as work progresses.
That design makes permissions part of the purchase. An enterprise considering this kind of workforce has to decide which systems it can access, which actions need approval and who receives the request. These are operating implications of the disclosed product design, rather than claims that a particular customer has already settled those questions. Delegation gives the buyer a continuing role in shaping the work the software is allowed to do.
Corma publishes numerical performance claims and anonymous customer accounts on its product page. The company describes deployments across large enterprises, including healthcare, financial services, energy, critical infrastructure and retail; NEA also reports production use. The public material reviewed does not provide enough independently validated detail to turn those accounts into a general performance guarantee. Named reference customers and evaluation results scoped to a buyer's own environment would make the commercial discussion more concrete.
The research organization behind the product
Sequoia's company record dates Corma's founding to 2025 and identifies Alon Pluda. Corma's official company profile places its headquarters in Tel Aviv and San Francisco. The disclosed leadership evidence establishes Pluda's founder and CEO role; it supplies no basis for inventing a CTO or filling out an executive roster from people loosely associated with the company online.
The careers page seeks researchers and engineers across pre-training, post-training, machine-learning infrastructure and security research. That combination shows where the organization is asking for help: developing the model, making it run and understanding the security work it is meant to perform. Corma has not attached a numerical hiring target to this financing.
For security buyers, those research priorities remain connected to everyday service. A better model needs the capacity to handle assigned work, and added capacity needs software that can use the customer's environment responsibly. As Pluda's team expands training and compute, each new deployment brings another set of tools, approval relationships and security decisions into the workforce's working day.
Cybersecurity funding, last 30 days
DevCuration's funding database tracked 14 Cybersecurity rounds totaling $784M in disclosed capital over the past 30 days. Recent deals we covered:
- IBM Acquires AuthMind to Expand AI Agent Identity SecurityNov 14
- Finite State Raises $27M Series C for Device SecuritySeries C · $27M · Nov 12
- Vanta Agrees to Acquire Pillar Security for AI GovernanceM&A · $150M (reported by CTech) · Nov 10
- Ballistic Ventures Targets $490M Across Two New FundsFund Target · $490M target · Nov 9
- Sentra Raises $150M Series C for AI Agent Data SecuritySeries C · $150M · Nov 7
Frequently Asked Questions
What would Corma's AI workforce change for a security team?
Corma describes software that uses existing security tools to carry defensive tasks through investigation and action. People can review findings and approve actions through familiar communication channels, so permissions and escalation remain part of deployment.
Why is Corma developing its own defensive foundation models?
Corma and its investors argue that defensive work depends on security data and tasks underrepresented in general-purpose model training. The company is pursuing specialized training, including reinforcement learning in enterprise-like environments; this is its technical approach, rather than an independent guarantee of better outcomes.
How will the Series A proceeds be used?
Corma says the funding will expand its foundation-model training, increase compute capacity and support selective hiring. It has not disclosed a spending split or numerical hiring target.
How should buyers interpret Corma's reported performance?
The public product materials contain company performance claims and anonymous customer accounts. Buyers should assess results in their own environment, including evidence quality, tool permissions and approval responsibilities; the reviewed sources do not establish a universal performance guarantee.
Where the Money Moved
The intelligence briefing of the innovation economy. Funding, M&A, debt and fund closes, read as market signal rather than deal announcements.
Subscribe to Where the Money Moved