The majority of Wall Street is not keeping a close eye on the quiet uprising taking place in the AI infrastructure trade. While the Magnificent Seven—Apple, Microsoft, Alphabet, Amazon, Nvidia, Meta, and Tesla—continue to make headlines, a much smaller company has been accomplishing something that none of them have in 2026. Since January, the stock of DigitalOcean, a $10 billion cloud company with its headquarters in New York, has doubled. Only three of the seven mega-caps are outperforming the S&P 500 this year, in contrast. The contrast is difficult to ignore, and the math is awkward.
The narrative does not claim that DigitalOcean created something that Google Cloud or Microsoft Azure did not. The company chose a market niche that the industry titans don’t particularly want to compete for. Enterprise whales are contracts with dedicated solutions architects and quarterly business reviews that are sought after by Amazon Web Services, Azure, and Google Cloud. Startups, independent developers, and small-to-medium enterprises that merely require a GPU, a tidy dashboard, and consistent monthly pricing are the opposite end of the spectrum that DigitalOcean built its business around. Fifteen years after its founding in a Brooklyn co-working space in 2011, the company has managed to rent Nvidia and AMD compute capacity to thousands of AI builders.
Unexpected benefits are coming from that model. DigitalOcean’s AI products generated $120 million in run-rate revenue in the fourth quarter, an increase of about 150% over the previous year. In 2025, total revenue increased by 15% to $901 million. Obviously, those aren’t Nvidia numbers. However, these are the kinds of numbers that imply a genuine activity is taking place beneath the surface. Just 21,000 so-called “digital native enterprises” out of over 650,000 clients generate about 62% of annualized revenue. Growth was nearly triple digits among the group’s larger spenders, those who paid half a million or a million a year. In other words, demand is concentrated and growing.
As this develops, it seems like the AI infrastructure story is finally taking off, as some analysts had predicted last fall. Around September, Wall Street strategists started promoting the “Magnificent Ten” reframe, attempting to include Palantir, Oracle, and Broadcom. Since DigitalOcean is too tiny, too unknown, and, to be honest, too unattractive in comparison to the trillion-dollar names, it isn’t currently on most of those lists. However, while everyone else is fighting over the front-page story, that’s precisely the kind of business that tends to compound quietly.
The part that continues to stick with me is the pricing argument. According to DigitalOcean, renting its data center compute can be up to 75% less expensive than what hyperscalers charge for comparable infrastructure. A significant portion of the AI economy will continue to gravitate toward this type of provider if that is even directionally true, as the company’s customer growth indicates. A Microsoft sales representative is not necessary for a lone developer in São Paulo creating a chatbot. They require a single GPU that is paid for on an hourly basis using a credit card. The gap is that.
Naturally, none of this indicates that DigitalOcean is now a better option than Microsoft. It isn’t. The stock is trading at about 11 times sales, which is significantly higher than its long-term average of 8. In order to build more data centers, the company raised $800 million in March. This kind of capital expenditure commitment could backfire if demand declines. And there’s always the question of whether the hyperscalers will ultimately determine that the SMB market is worthwhile. Their budgets are sufficient to make life challenging.
It’s difficult to ignore the asymmetry, though. While Amazon is aiming for $200 billion and Meta is being asked to defend $135 billion in capital expenditures, DigitalOcean is accomplishing more, proportionately, with a fraction of the resources. It’s already the most intriguing AI story that hardly anyone is adding to their portfolio, regardless of whether it scales for an additional two years or hits a wall.
