CEO, DigitalOcean
Most AI clouds were built around capacity. We built ours around software.
Today, I’m at the TomorrowX Summit talking about our AI-Native Cloud. We think the winners of the inference era will put tokens, agents, and the core cloud underneath them on one platform, run it on software that makes every megawatt more valuable, and stay open to whatever model a customer wants to use.
The results say our strategy is working. We grew revenue 40% year over year in Q3, token volume grew roughly 10x quarter over quarter, and our outlook for 2027 is growth of 60% or more.
Here is what that looks like.
Customers don’t want tokens, agents and cloud from three vendors. An agent that calls a model also needs a database, storage, networking, and somewhere to run. So we put them together. 90% of our AI customer revenue comes from customers who also use our core cloud.
Inference is the business. More than 90% of our AI customer ARR is inference services and core cloud, not bare metal GPUs. Customers running production workloads stay, grow, and adopt more of the platform. Accounts using four or more products drive about 60% of our revenue.
Agents are now buyers. About half of the visitors to our site are agents, not people, and more than 37,000 customers have deployed on DigitalOcean without a human in the loop. Agents don’t sit through a sales cycle. They read the docs, call the API, and pick the platform that works on the first try.
We have spent more than a decade making the cloud simple for developers. It turns out that is exactly what agents need too.
We obsess over what each megawatt earns. In Q3, across our entire fleet, we generated $18M of ARR per active megawatt. A GPU landlord rents the chip by the hour to one tenant. We sell what runs on it: tokens, agents, and the compute, storage, and databases around them. And we have levers to push that higher: pricing that moves with the market, more token-efficient GPU generations, higher utilization from elastic compute, and core cloud attach.
And we built it on a different balance sheet. We generated $220M in adjusted free cash flow over the last twelve months, we are GAAP profitable, and we ended the quarter with $1.2B in liquidity and 1.2x net leverage. A stable, profitable core cloud business of about $900M in ARR, growing double-digits, supports a differentiated credit profile as we grow and scale our AI platform.
No single customer defines us. We have more than 700,000 paid customers. Our top 25 are only 27% of total revenue, and no customer is even a high single-digit percentage. We allocate capacity every day to the highest-quality customers, the ones adopting the breadth of the platform. Demand far exceeds supply, so capacity that frees up is immediately repurposed and does not sit idle.
We have minimal geo-political risk. Given our customer diversity, we have minimal concentration in any one country outside of the U.S.
Open models give customers control, and a better cost. As I wrote in September, intelligence should stay yours. Our catalog now has 85 models, including the leading open source models like Deepseek, Kimi, and Nemotron, as well as open weight models from the US, UK. and France. Many of our customers already route across 20 or more of them. If any one model becomes unavailable or restricted, the rest of the catalog is already there to fill the gap.
Most of the market focused on scale first. They raced to secure GPUs and power, and now they are trying to get the software right. We went the other way. We got the software right first, and now we are scaling it prudently, with all three of our 2026 data centers launched on time or early and another wave of capacity coming online in 2027.
A platform that makes every megawatt more valuable has to be built, and that is the head start we have. That is what different by design means.
As Chief Executive Officer, Paddy Srinivasan drives the strategic direction for DigitalOcean. With over 25 years of experience in technology leadership and a proven track record of delivering customer-centric solutions, Srinivasan brings invaluable expertise to further DigitalOcean's mission of simplifying cloud computing.

