Finn
DOCN Cloud infrastructure · AI cloud · Developer tools · Consumption software · Thesis updated November 15, 2026

Hyper-growth returns on surging AI token volume

01 Running thesis

A hyper-growth AI story with rising capital needs

DigitalOcean used to be known mainly as the simpler cloud for developers and smaller tech companies. The core business still matters, but the new thesis is much bigger. Management wants the company to become the easy cloud for AI inference and agentic applications, which means running AI models after they are trained.

Growth is accelerating rapidly. In the third quarter of 2026, the company posted its fifth straight quarter of acceleration at 40% year over year. Management raised 2027 revenue guidance from 50% to 60% or more. This massive upgrade is fueled by explosive traction in token inference, where volume grew 10 times over the prior quarter.

The company is proving it is more than just a low margin hardware rental business. It generates $18 million in annual recurring revenue per average active megawatt across its fleet. Furthermore, 90% of AI customer revenue now includes core cloud services, proving that AI acts as a strong wedge to drive traditional cloud usage.

The bear case centers on capital requirements and margins. To support scaling to 180 megawatts of committed capacity by 2028, the company secured a $785 million debt facility and a $1 billion stock sale program. Gross margins compressed to around 50% in the third quarter due to heavy depreciation from these new data centers. The central question is whether the company can maintain returns on capital as hardware costs rise.

Nov 2026▲Q3 2026 earnings triggered a massive thesis upgrade. Management raised 2027 growth guidance to 60% or more, driven by a 10-fold quarterly increase in token volume.
Aug 2026▲Q2 2026 earnings showed accelerating momentum. Management raised 2026 revenue growth guidance to about 30%, driven by explosive growth in inference services and a shift toward open weight models.
May 2026▲DigitalOcean launched its AI Native Cloud and raised its outlook. Management expected about 26% revenue growth in 2026 and 50% or more in 2027, supported by 60 megawatts of new capacity.
May 2026→Q1 2026 filings showed ARR of $1.032 billion and AI Customer ARR of $170 million, but gross margin fell to 56% from 61% a year earlier. The update made the growth story stronger and the margin question more important.
Feb 2026▲The 2025 10-K recast the company around agentic inference cloud and introduced the DNE customer framework. ARR reached $970 million at year-end 2025, and net dollar retention improved to 100%.
Nov 2025▲Q3 2025 showed better ARPU and improved net dollar retention at 99%. The data supported the view that higher-spend customers were becoming more important.
May 2025▲Q1 2025 improved net dollar retention to 100%, which eased a key concern about customer churn and contraction.
Nov 2024→Q3 2024 showed stronger ARPU and more revenue from larger customer cohorts, but net dollar retention stayed below 100%. Leadership turnover also added execution risk.
02 Business model

Simple cloud, paid by usage

DigitalOcean makes money when customers use its cloud platform. Pricing is mainly consumption based and billed monthly, so revenue rises when customers run more apps, store more data, or use more compute. This model is easy for small teams to start with, but it can also move down if customers cut usage.

The company focuses heavily on Digital Native Enterprise customers. These users spend more than $500 in a month. Product density is accelerating in this group. In the third quarter of 2026, 60% of total revenue was driven by customers using four or more products.

DigitalOcean still uses a self service sales motion. Developers can sign up without a large sales process. The company adds targeted sales help for larger AI native accounts. This keeps the model simpler than the hyperscalers, but it means the company must prove it can land larger customers without losing its core identity.

The main break point is capital intensity. AI workloads require massive data center capacity and expensive GPUs. The company avoids the pure GPU rental model by focusing on token flow and agentic execution, pulling through higher margin core cloud services to offset hardware costs.

03 Product portfolio

Five layers for AI builders

Cash cow

Foundational cloud infrastructure

This includes global infrastructure, Kubernetes, CPU and GPU Droplets, networking, and storage. It is the base layer customers use to run their applications.

Growth engine

Inference Engine

This layer provides serverless and dedicated endpoints for AI models, batch processing, and an intelligent inference router. It handles massive token flows across open models.

Steady

Data and Learning Layer

This includes managed MySQL and PostgreSQL plus vector database support. It helps AI applications store normal data and manage context and memory.

Option

Managed Agents Platform

This platform executes autonomous agents at scale. It features secure sandboxes and action gateways providing over 20,000 tools for agents.

Growth engine

AI Middleware

This intelligent data plane routes workloads across models and regions based on real-time trade-offs, improving performance and cost efficiency.

04 Business segments

Bigger customers drive the mix

Digital Native Enterprise customers67%growing fast
Developers33%declining

DigitalOcean does not report formal operating segments. Based on Q2 2026 disclosures, Digital Native Enterprise customers were 67% of revenue, and Developers were the remaining 33%.

05 Risk factors

What could go wrong

180 megawatts of capacity execution

High impact · Medium odds

The company is scaling from 75 megawatts currently active to 180 megawatts committed by 2028. Flawless hardware procurement and partner execution are required to prevent stranded capital expenses.

We watchCapacity turn-ups and management updates on data center construction timelines.

Gross margin compresses further

High impact · Medium odds

GAAP gross margins fell to around 50% in the third quarter due to depreciation and lease hits as new data centers turn up. If hardware cycle times shorten, this pressure could become permanent.

We watchQuarterly gross margin trends and detailed 2027 profitability guidance expected in February.

Next generation hardware costs

Medium impact · Medium odds

Transitioning to GB300 or Vera Rubin architecture next year will push capital expenditures to $35 million or more per megawatt. Software-driven revenue must rise to match this to keep returns intact.

We watchReturn on invested capital and the annual recurring revenue generated per megawatt.

High expectations meet a high price

High impact · High odds

The story has improved dramatically, but the market is already giving credit for flawless hyper-growth execution. If 2027 growth slips below 60%, the stock could react sharply to the downside.

We watchAny changes to 2027 revenue guidance and the pace of inference volume growth.
06 Quick answers

In one breath

What does DigitalOcean do?

DigitalOcean sells cloud infrastructure and developer tools. Customers use it to run apps, databases, websites, Kubernetes, storage, GPUs, and now AI inference and agent workloads.

Why is AI important for DOCN?

AI is now the primary growth engine. Inference services and token volumes are growing exponentially, leading management to guide for 60% or more total revenue growth in 2027.

How is DigitalOcean different from AWS or Azure?

DigitalOcean aims to be simpler and more predictable for developers and AI-native teams. It also offers open-source options at every layer of its AI stack to prevent vendor lock-in.

What is the biggest risk for DigitalOcean stock?

The biggest risk is execution on massive data center builds. The company must bring 180 megawatts of capacity online by 2028 and fill it with high margin software services to justify the capital spend.

07 Research standards

Sources and research notes

This page combines Finn's company research with public filings and other cited materials. The thesis is reviewed when material company information changes; Finn Scores use the latest available scoring data.

Thesis reviewed
November 15, 2026
Score data
November 15, 2026
Reviewed by
Shivam Bharuka
  1. DigitalOcean Q3 2026 earnings call transcript
  2. DigitalOcean Q3 2026 Form 10-Q
  3. DigitalOcean Q2 2026 earnings call transcript
  4. DigitalOcean Q2 2026 Form 10-Q
08 Explore the industry

Comparable Software - Infrastructure companies

Companies near DigitalOcean Holdings, Inc. in Finn's Software - Infrastructure industry ranking.

Get started with Finn today