- Railway closed a $100 million Series B led by TQ Ventures, bringing total funding past $130 million while already generating over $10 million in annual recurring revenue.
- The cloud platform now serves more than 2 million developers—31% of the Fortune 500—and is adding nearly 200,000 new users every month.
- Founder Jake Cooper says the next phase targets AI-native infrastructure, arguing today’s cloud was never designed for the workloads that actually matter now.
A cloud startup most people haven’t heard of just raised nine figures to pick a fight with Amazon. Railway, a San Francisco-based deployment platform founded in 2020, closed a $100 million Series B round led by TQ Ventures, with participation from FPV Ventures, Redpoint, and Unusual Ventures. The company is already past $10 million in annual recurring revenue—a milestone most enterprise SaaS companies spend years chasing.
The pitch is simple: developers hate configuring infrastructure, and AI is making that pain worse, not better. Railway offers instant provisioning, built-in CI/CD, observability, and database management from a single code push. No YAML files. No three-hour onboarding docs. The platform now supports over 2 million users and is growing by roughly 200,000 developers per month.
The company’s client roster reads like a cross between a Fortune 500 directory and an AI startup demo day. Intuit’s GoCo, TripAdvisor’s Cruise Critic, and MGM Resorts all run workloads on Railway alongside AI-native teams like Profound and Happy Robot. According to FinSMEs, 31% of Fortune 500 companies now use the platform in some capacity.
Why AI Workloads Are Breaking the Cloud Playbook
The timing isn’t accidental. As AI models get more capable, the infrastructure beneath them gets more strained—and the cracks in legacy cloud architecture keep widening. 40% of US data centers scheduled to come online in 2026 are delayed, according to industry tracking. The hyperscalers are spending hundreds of billions on new capacity, but that expansion doesn’t help a startup that needs to deploy an inference endpoint at 2 AM on a Tuesday.
Railway’s bet is that AI-era development demands a different abstraction layer entirely. Traditional cloud platforms were designed for predictable web traffic—stateless requests, horizontal scaling, predictable load patterns. AI workloads flip that model: GPU-heavy bursts, stateful model serving, vector database queries that spike unpredictably. The tools built for 2015-era web apps are buckling under 2026-era machine learning pipelines.
Cooper has pointed to this mismatch directly, arguing that today’s cloud infrastructure was built for a world that no longer exists. The plan now is to expand Railway’s global data center footprint and build new tools specifically designed for AI systems—essentially rearchitecting the deployment layer from scratch rather than bolting AI support onto legacy plumbing.
The Developer-First Growth Hyperscalers Can’t Copy
What makes Railway interesting isn’t just the product—it’s the distribution model. The company claims it has spent essentially nothing on marketing. Its 2 million users came through word of mouth, developer communities, and the kind of organic adoption that enterprise sales teams spend decades trying to replicate. Growing by 200,000 developers per month without a paid acquisition strategy is the kind of metric that makes VCs write checks before the pitch deck finishes loading.
This is also where the AWS comparison gets nuanced. Amazon Web Services generated over $100 billion in revenue last year. Railway’s $10 million ARR is a rounding error by comparison. But the threat model for hyperscalers was never about direct revenue competition—it was about losing the next generation of developers before they ever touch the legacy console. If Railway captures the cohort of AI-native builders who never learn AWS in the first place, the long-term implications for cloud market share are real.
The funding will go toward global data center expansion, team growth, and building what Railway calls tools for “developers and AI systems”—a deliberately vague phrase that signals the company is planning infrastructure specifically optimized for model training, inference, and the agentic workflows that are reshaping how software gets built.
The company was founded in 2020 by CEO Jake Cooper and has raised over $130 million to date across its seed, Series A, and now Series B rounds.
