DigitalOcean has spent most of its life as the cloud for developers and small teams, billing hundreds of thousands of customers by the hour for servers and databases. A preliminary set of second-quarter results, released on July 7, shows how quickly that is changing. The company said it landed multiple nine-figure annual customer commitments for AI inference and cloud services in the quarter, sending its contracted backlog up more than tenfold and pushing its growth rate back into acceleration. The developer cloud is growing into AI infrastructure, and it wants to be judged as a platform rather than a rack of GPUs.

What the Backlog Jump Shows

The standout figure is the backlog. Remaining performance obligations, or RPO, measure revenue a company has under contract but has not yet booked, and DigitalOcean’s is set to exceed $800 million, more than ten times its level a year earlier. Alongside it:

  • RPO expected to exceed $800 million, up more than tenfold year over year
  • The weighted-average length of those contracts rising from 1.6 years to over three years
  • Second-quarter revenue growth accelerating to about 29 percent, from 14 percent a year earlier
  • An incremental 20 MW of data center capacity secured for late 2027 and early 2028, bringing committed capacity to roughly 155 MW

The company also expects to meet or exceed its prior guidance for adjusted EBITDA margin and non-GAAP earnings, and its shares rose on the update. These are preliminary figures ahead of full results, but the direction is clear: a business that had been growing in the mid-teens is re-accelerating, and it is doing so on large, multi-year AI commitments rather than incremental self-serve signups.

Half a Billion Dollars to Shrink: What Public Filings Say a Hosting Customer Costs in 2026

From Developer Side-Projects to Nine-Figure Contracts

For a company built on customers who spend tens or hundreds of dollars a month, nine-figure annual commitments are a different kind of business. DigitalOcean did roughly $900 million in revenue last year, so a handful of contracts each worth $100 million or more per year reshapes the customer base fast. It also continues a shift the company has signaled for a while, having re-based its own metrics toward higher-spending customers and away from the smallest tiers. Longer contracts and a growing backlog make the revenue more predictable and more enterprise-like, which is what investors reward. They also move DigitalOcean further from the self-serve simplicity that was its original pitch.

“AI-Native Cloud,” Not Raw Compute

The framing DigitalOcean is pushing matters as much as the numbers. Chief executive Paddy Srinivasan said customers “recognize the differentiation of our AI-Native Cloud platform” and that the company is winning what he called the world’s most sophisticated AI customers. The distinction he is drawing is with commodity GPU rental: rather than selling raw compute by the hour, DigitalOcean is selling a software layer on top of it, including an Inference Router that balances price and performance across both proprietary and open-source models. It is the same argument a growing number of providers are making as GPU capacity becomes abundant and undifferentiated, that the margin and the lock-in live in the software around the chips, not the chips themselves. Whether DigitalOcean’s version is sticky enough to hold nine-figure customers against far larger clouds is the open question.

The Risk in a Few Big Deals

A backlog built on a small number of very large commitments is a strength and a vulnerability at once. It gives DigitalOcean visibility and a growth story it did not have a year ago, but it also concentrates the business in a handful of unnamed AI customers whose own demand can shift quickly, and it commits the company to building capacity, 20 MW more on the way, against contracts that run for years. The self-serve base that made DigitalOcean resilient does not disappear, but the growth narrative now rests on a lumpier kind of customer. For a developer cloud that spent more than a decade avoiding the hyperscaler game, the nine-figure AI deal is both the prize and the new risk. The full second-quarter results will show how durable the turn is.