Upwind Security Reaches $3.8 Billion Valuation With $300 Million Bet on Cloud and AI Runtime Defense

Cybersecurity startup Upwind Security has raised approximately $300 million in new financing at a valuation of roughly $3.8 billion, illustrating how quickly investor attention is shifting toward the security layer underneath cloud and artificial-intelligence infrastructure.

The financing was led by Bessemer Venture Partners and TCV, with participation reported from Salesforce Ventures, Greylock, Craft Ventures and other investors. The valuation represents a substantial increase from Upwind’s earlier 2026 financing, when the company raised $250 million at roughly a $1.5 billion valuation.

The timing reflects a change in how enterprises think about security. Cloud environments now change continuously: containers start and disappear, APIs are deployed constantly, machine identities receive credentials automatically and AI agents connect to applications and data stores.

Upwind’s product strategy is built around watching that environment while it is actually running. The company describes its architecture as runtime-first cloud and AI security, combining configuration analysis with sensors that observe workloads, processes, network activity and identities in production.

That runtime context can help answer a question traditional scanning often cannot: not merely whether a vulnerability exists somewhere in a software package, but whether the affected code is active, reachable from the internet and connected to sensitive data.

The distinction matters because large cloud environments can contain thousands of known vulnerabilities and policy warnings. Treating every finding as equally urgent overwhelms security teams. Runtime evidence can help prioritize the issues that actually create an attack path.

Upwind is extending the same approach to AI infrastructure. Enterprise AI deployments can include model APIs, vector databases, orchestration frameworks, agents, MCP servers and cloud credentials. Those components do not replace older cybersecurity risks; they add new ones.

A prompt attack can manipulate an agent. A stolen model-provider key can consume expensive inference capacity or expose private services. A compromised orchestration server can execute commands. A misconfigured machine identity can give an autonomous system more access than intended.

That is why the cloud-security and AI-security markets are beginning to converge. The same runtime sensors used to understand a container or Kubernetes workload can also help show which AI services are running, what they can reach and what identities they are using.

Upwind says its live inventory refreshes on a seconds-scale and that its platform now spans cloud security, API protection, identity analysis, AI security and agentic security. The company says hundreds of enterprises use its technology.

The broader competitive environment is formidable. Palo Alto Networks, CrowdStrike, Microsoft, Google Cloud, Check Point and numerous specialized vendors are all expanding across cloud and AI security. Google’s acquisition of Wiz also removed one of the largest independent cloud-security companies from the standalone vendor market, potentially creating room for younger rivals while increasing competitive pressure.

Upwind’s $3.8 billion valuation is therefore a significant execution bet. The company now has to turn rapid fundraising and product expansion into durable enterprise revenue.

But the market signal is clear. AI investment is pulling cybersecurity deeper into the infrastructure stack. The first wave of AI capital funded models. The second funded GPUs, data centers and power. A parallel wave is now funding the companies expected to secure all of that infrastructure.

Upwind’s $300 million round is one of the clearest private-market examples yet.