US cloud-based software giant Salesforce has invested several tens of millions of dollars in Israeli cloud infrastructure security company Upwind Security at a company valuation of $1.6 billion, a source familiar with the matter has told "Globes." This is an especially high valuation for a startup that had annual recurring revenue (ARR) of over $20 million in 2025.
In making the investment, Salesforce has joined the financing round announced by Upwind Security in January in which it raised $250 million at a company valuation of $1.5 billion. One reason investors are enthusiastic about the Israeli cybersecurity company is that it is seen by the cloud industry as one of the growing alternatives to Israeli rival Wiz, which was sold earlier this month to Google's cloud service for $32 billion. Despite promises from Wiz executives and Google's cloud service leaders about neutral continuity, Google's competitors, Amazon and Microsoft, are seeking alternatives.
Upwind, along with Israeli companies such as Echo and Sweet Security, are seen as Wiz’s main competitors today in the field of cloud data security, with Upwind leading the market in terms of revenue and integration agreements with the cloud giants.
Upwind recently signed an enhanced integration agreement with Amazon, so that salespeople for its cloud service, AWS, receive a commission on the sale of the Israeli system as if it were an internal Amazon system. Microsoft is also negotiating to upgrade its integration agreement with Upwind.
The strong position that Upwind has created for itself has brought it new investors such as Bessemer, who led the previous investment round of $250 million last January, with the participation of Picture Capital, a group of Israeli cybersecurity executives, and existing investors such as CyberStarts - the first investor in the company, along with Greylock, TCV, Kraft Ventures and Leaders Fund.
Meanwhile, Upwind today unveiled at the RSAC Conference in San Francisco a collaboration with Nvidia that allows it to identify in real time malicious commands designed to damage language models, a new and emerging type of threat in the field of AI clouds. At the conference, Upwind demonstrated how it is possible to identify malicious commands in real time that are directed at language models without, it claims harming the system's speed or increasing costs - using Nvidia's Nemotron technology.
The new attacks against language models
Unlike traditional cyberattacks, which try to exploit weaknesses in code or corrupted files, the new attack method works through the language itself - in an attempt to change meaning, divert intent and cause the system to perform actions that were not planned in advance, through malicious queries (prompts). Among the most prominent risks are command injection, breaking into model limitations, data leakage and fraud methods based on persuasion and deception. Upwind says traditional security tools are not well built to deal with this type of threat.
Upwind VP research and innovation says, "Language models not only accept input but also try to understand intent. Therefore, the entire security model is changing. Organizations are no longer just dealing with malicious code, but with attempts to distort language and influence systems through sophisticated wording. Research done with Nvidia shows that such attempts can be stopped even in an active production environment, without slowing down the system and without hiking costs."
Upwind Security was founded in 2022 by CEO Amiram Shachar, Liran Polak, Tal Zur and Lavi Ferdman who previously founded cloud security management company Spot.io, which was sold to NetApp for $450 million in 2020.
Published by Globes, Israel business news - en.globes.co.il - on March 24, 2026.
© Copyright of Globes Publisher Itonut (1983) Ltd., 2026.