loading...
Railway built a cloud platform that undercuts AWS by 50%, serves 2M developers, and has never spent a dollar on marketing.
ENTRY ANGLES
Build AI agent orchestration tooling optimized for per-second burstable compute on Railway · Create Railway-native developer tools that leverage per-second billing economics
VERTICALS
CAPABILITIES
Infrastructure engineering, Cloud architecture, Developer tooling, AI agent scheduling
Thirty employees. Two million developers. Zero dollars spent on marketing. Railway’s numbers read like a typo until you understand the economics underneath.
Jake Cooper founded Railway in 2020. By January 2026, when he closed a $100 million Series B led by TQ Ventures, the company had become one of the most capital-efficient cloud platforms ever built. Deploy an application, it scales, you pay by the second for actual usage and nothing when idle. No minimum commitments, no surprise VM charges – pricing that undercuts AWS by roughly 50% and newer cloud alternatives by three to four times.
That differential is not subsidized. Railway owns its infrastructure stack end to end rather than renting capacity from a hyperscaler and marking it up. The 31% Fortune 500 adoption alongside two million individual developers – growing by 200,000 each month – happened without a single paid acquisition campaign. FPV Ventures, Redpoint, and Unusual Ventures also participated in the round.
Per-second billing with no idle charges is not just cheaper – it is structurally different. An AI team shipping ten experiments a week wants to spin up, test, tear down, and move on, not carry the cost of capacity sitting at 3% utilization between runs. The hyperscalers built their pricing for steady-state workloads where idle capacity is something you pay for. AI agents and AI-first applications deploy differently: fast iteration cycles, variable load, short bursts of intense compute followed by extended quiet.
The thirty-employee figure deserves specific attention. Most infrastructure companies at Railway’s scale have hundreds of employees and hundreds of millions in venture backing burned through sales and marketing. Railway’s unit economics are unusual because the company controls its stack end to end and did not pay a hyperscaler to operate it. The 50% price differential is a structural cost advantage from owning what competitors rent, not a subsidy financed by venture capital.
The bottleneck Railway will hit is enterprise procurement. Individual developers and startup teams adopt on product merit; large enterprises buy cloud infrastructure through multi-year vendor relationships and security reviews that Railway does not yet have full coverage for. The $100 million raise is partly about building that motion.
The more interesting angle for builders is what economics become viable on a cheap cloud substrate. AI-first applications that were economically marginal on AWS pricing become viable on Railway’s. The latent demand is not just from teams that want to save money – it is from teams whose products require usage economics that AWS pricing makes unprofitable.
The specific entry angle: agent orchestration tooling built around per-second, burstable compute. The cost visibility and billing granularity Railway offers make it possible to run agents economically at the usage pattern that fits AI workloads – many short tasks, not a few long ones. Tooling that optimizes agent scheduling for this billing model does not exist yet and would be more valuable on Railway than anywhere else.