For years, cloud computing quietly rewired how companies build products, scale infrastructure, and manage costs. And the widespread sudden adoption of AI has accelerated it. In doing so, it exposed a far more dangerous problem than rising bills: how little most enterprises actually understand about what they’re spending and why.
This week, I spoke to a company that understands how the financial stakes have changed. “We’re spending all that money on something that we’re not really sure what the ROI is,” Roi Ravhon, Co-founder and CEO of cloud cost management platform Finout, told me. What once lived in innovation budgets is now embedded directly in business fundamentals. “Now it’s part of our gross margin… It’s part of what we’re building as a service.”
Founded in 2021, Finout helps enterprises monitor, allocate, and forecast cloud spending across providers, including AWS, Google Cloud, Azure, Datadog, Kubernetes, and Snowflake. The company has raised $85 million to date and works with customers such as Lyft, The New York Times, SiriusXM, Wiz, and Tenable.
Not only is one problem that is AI inherently expensive, but businesses are also adopting it and transforming their practices without clarity. “AI is a lot more expensive than what we thought it would be,” Ravhon explained. “We’re not really sure how predictable it’s going to be. We’re not really sure if we’re using it effectively or not. Just buying and buying and buying more AI services.”
Cloud costs were already complex before AI arrived. But today, AI workloads are priced by tokens, usage, and models that can quickly change, making it even harder to ignore. The result is cost waste that hides in plain sight. “There are so many dumb ways, it’s amazing,” Ravhon says when asked how those may materialize.
The dynamic is familiar, even if the scale is not. Just as individuals may lose track of unused subscriptions, businesses can accumulate cloud services that persist simply because no one is sure what would happen if they disappeared.
In some cases, the scale is staggering. Ravhon recalls working with enterprises that had “tens of millions of dollars of ‘shadow IT’”, meaning services running in the cloud that no one fully understood and no one wanted to turn off. Teams hesitate to shut anything down because it might break something, or might do nothing at all.
I asked if there was a tension between the engineering teams, who are incentivized to move fast and build reliably, and the finance teams, who are accountable for budgets and forecasts. Turns out there is - and in practice, engineering usually wins. “I'm an engineer, we tend to be very defensive,” Ravhon says. “Finance sets a budget, engineering depletes it.” He adds, “It’s very easy to pick the most expensive model to sleep better at night.”
Ravhon, who first spotted these kinds of gaps when he was Director of Core Engineering at Logz.io, a cloud observability company, argues the conversation needs to shift away from simply cost-cutting and toward control. “Cloud is not spend, it’s an investment,” he says. “The best way to overcome this is with data.”
When asked during our quickfire round what leaders should remember from the AI cost reckoning now underway, Ravhon doesn’t hesitate. His answer is a single word: “Allocate.”
In the AI era, ignorance can be an existential issue if not managed from the start.
