Upwind raises $400 million to build out security for cloud systems and AI agents

Upwind has fresh capital to build security tools that follow AI activity across company devices, code and cloud infrastructure. The company closed a $400 million Series C on October 28, 2026, at a $3.8 billion valuation. Bessemer Venture Partners led the round, which will support product development and expansion of its sales efforts.

More than $800 million behind the platform

The financing brings Upwind’s total funding to more than $800 million, according to CEO and co-founder Amiram Shachar. Canapi Ventures joined the round alongside existing investors, including Cyberstarts, Greylock Partners, Craft Ventures, Salesforce Ventures and TCV. Shachar said the money would fuel product and engineering growth for the cloud and AI security platform.

Upwind also said its revenue quadrupled in 2026. Its announcement does not disclose an absolute revenue figure, leaving the size of that business unspecified. The company names Siemens, Roku, Carvana, Nubank, Supabase and ClickUp among hundreds of organizations using its platform. Those are company-reported indicators of commercial adoption, not independent measurements of security effectiveness.

The round’s size needs one clarification: PYMNTS reported $450 million. Both Upwind’s own blog post and its company-issued financing release specify $400 million; that is the figure used here.

Watching what runs, then deciding what to fix

Upwind’s approach centers on runtime intelligence: information about what software is doing while it operates, rather than only what a periodic scan finds. The company says its platform combines signals from cloud workloads, software interfaces and AI agents to identify critical risks, determine which exposures attackers can reach and detect attacks as they happen.

Shachar’s argument is that disconnected security alerts lack the context needed to understand an AI agent’s intent or whether a weakness is reachable. Upwind wants security, development and operations teams working from a shared view of those risks. Its three categories of security agents divide the work between investigation, testing and finding fixes:

• Blue investigates logs and other data to assemble a single account of an attack. Shachar describes it as connecting evidence across sources rather than leaving teams with separate signals.
• Red tests whether exposed systems and vulnerabilities can actually be exploited in the customer’s environment. Upwind says those tests help teams distinguish real exposure from findings that do not warrant the same attention.
• Green identifies places where a change could address many security findings and affected resources. The company positions it as a way to prioritize fixes across the environment, rather than treating each finding separately.

The distinction is visible in a customer account included in the announcement. HedgeServ security chief Michael Cannizzo said seeing what was actually running, rather than only scan results, changed which risks his team spent time on. That describes a shift in priorities, rather than a quantified reduction in attacks.

Devices, agent scanners and longer-term research

In his funding post, Shachar said Upwind is building AI sensors for endpoints—devices within a company’s environment—and scanners for agents and their skills. It is also reworking its inventory graph, the connected map of that environment. He framed the scope as extending from devices through code to cloud infrastructure, rather than stopping at code and cloud.

The post also describes AI Labs for Cybersecurity as part of that strategy. Its four research areas are AI security; agentic security, focused on AI agents; frontier models, systems and benchmarks; and advanced AI systems. Shachar said the lab is intended to give researchers infrastructure and freedom to work on problems that could shape future security needs.

Upwind already supports Amazon Web Services, Microsoft Azure, Google Cloud Platform and Oracle Cloud Infrastructure. Its AWS Security Hub integration lets customers access Upwind through the AWS console, with onboarding and billing through AWS. Alongside deeper platform investment, Shachar said the company intends to make the product simpler to use and broaden its work around identity, data and AI.