Railway, a San Francisco-based cloud platform that has grown to two million developers largely by word of mouth, has raised $100 million in Series B funding. The round, led by TQ Ventures with participation from FPV Ventures, Redpoint, and Unusual Ventures, positions the five-year-old company as one of the more aggressive attempts to rethink cloud infrastructure for an era where AI systems, not just humans, write and ship code.
The company’s pitch to cloud engineers and infrastructure leaders is stark: cloud primitives designed for minutes-long deployment cycles and idle virtual machines are misaligned with AI coding assistants that can generate and revise applications in seconds. Railway argues that the next wave of infrastructure must be built to operate at what its founder and CEO Jake Cooper calls “agentic speed” — fast enough and cheap enough for continuous, automated deployment by AI agents.
With more than 10 million deployments a month and over a trillion requests flowing through its edge network, Railway claims usage metrics that rival better-funded competitors, despite having raised only $24 million prior to this round. The new capital is meant less to keep the business alive — Cooper says the company is “default alive” — and more to push its AI-native thesis into the mainstream enterprise cloud conversation.
From AI Coding Assistants to Infrastructure Bottlenecks
Railway’s core critique of the current cloud stack starts with deployment latency. The dominant infrastructure-as-code tools and workflows — Terraform being the canonical example — typically run build-and-deploy cycles in the two- to three-minute range. That cadence made sense when humans wrote and changed code at human pace.
In a world where tools like Claude, ChatGPT, Cursor, GitHub Copilot, and similar assistants can emit working code in seconds, those same pipelines become the limiting factor. Cooper describes the resulting mismatch bluntly: when “godly intelligence is on tap and can solve any problem in three seconds,” the deployment machinery becomes the bottleneck instead of the developer.
Railway claims sub‑second deployments, positioning its platform not just as a productivity boost but as an enabling layer for AI agents that can iteratively ship and refine services in tight loops. The company reports customers seeing a tenfold increase in developer velocity and up to 65% cost savings versus traditional cloud providers. While those are headline numbers, some enterprise references provide more concrete detail.
G2X, a platform serving roughly 100,000 federal contractors, migrated its infrastructure to Railway and reported deployment speedups of 7x and an 87% reduction in infrastructure spend. According to CTO Daniel Lobaton, that translated into cutting his monthly bill from about $15,000 to roughly $1,000. More importantly for engineering leaders, the velocity change altered his team’s operating rhythm: work that previously took a week on the prior setup could be completed in a day on Railway, and spinning up six services to test new architectures became a matter of minutes instead of protracted provisioning cycles.
For organizations already experimenting with AI-assisted or agent-driven development workflows, Railway is positioning this kind of deployment speed as a prerequisite rather than a nice-to-have. The claim is not merely that developers are happier, but that existing cloud pipelines are structurally misaligned with the emerging pattern of constant, automated code generation and redeployment.
Vertical Integration: Why Railway Walked Away from Google Cloud
Where many newer platforms abstract over AWS or Google Cloud, Railway chose the opposite path. In 2024, the company exited Google Cloud entirely and began building and operating its own data centers. It is a strikingly vertically integrated approach for a young startup, echoing Alan Kay’s line that those serious about software should build their own hardware.
Cooper’s rationale is that full-stack control over network, compute, and storage is the only way to consistently deliver the sub-second build-and-deploy loops he wants to target. By designing its hardware and infrastructure layout specifically for density and low-latency deployment workflows, Railway aims to create what Cooper describes as “agentic speed” while still presenting a smooth experience to engineers.
This vertical control has also had operational side effects. During recent widespread outages that disrupted major cloud providers globally, Railway reports that its infrastructure remained online. While the company has not published a detailed postmortem or SRE-level accounting of why its stack rode out those events, the incident underscores a practical implication of its strategy: by not depending on a hyperscaler’s control plane or shared network fabric, Railway’s failure modes can diverge from those of the major clouds.
The model also underpins Railway’s unit economics and pricing. By squeezing more workloads onto its hardware and charging only for active compute, it asserts that it can undercut hyperscalers by around 50% and newer cloud startups by three to four times. For engineers and architects used to balancing performance, reliability, and cost against the gravity of AWS or GCP, Railway’s willingness to own the entire stack represents both a risk and a potential advantage: fewer external dependencies, but more reliance on a single emerging vendor’s execution.
Pricing, Performance, and the Appeal of Usage-Only Compute
Railway’s pricing is built around fine-grained, per-second metering of actual resource use, rather than billing for provisioned but idle capacity. The published rates are:
- $0.00000386 per gigabyte-second of memory
- $0.00000772 per vCPU-second
- $0.00000006 per gigabyte-second of storage
There is no charge for idle virtual machines or reserved capacity. For teams that habitually overprovision instances or maintain long-running, lightly used services, this model can materially change cost profiles. Cooper contrasts this directly with the traditional pattern on AWS and other hyperscalers, where customers pay for full VM allocations even if they routinely use only a fraction of that capacity.
The claim is that a hardware and runtime stack optimized for density — plus per-second metering — produces enough margin to support lower headline prices while still being sustainable. That is a familiar argument conceptually, but Railway’s early customer anecdotes, like G2X’s 87% cost reduction, suggest that its particular combination of pricing and resource management can translate into tangible savings in practice.
For cloud engineers, the more interesting question is not simply raw price but how this usage-only model interacts with AI-era workloads. As code generation accelerates and more short-lived services or experiments get deployed, the overhead of constantly creating, scaling, and tearing down infrastructure becomes central. Per-second charging aligned with near-instant deployment allows teams — and eventually AI agents — to treat infrastructure as a tightly coupled part of the development loop rather than a static backdrop that must be manually tuned to avoid waste.
Lean Operations, High Revenue per Engineer
Underpinning Railway’s technology story is an unusual operating profile. The platform runs with about 30 employees, yet is generating what the company describes as tens of millions in annual revenue. Revenue grew 3.5x last year and is currently expanding at roughly 15% month-over-month, according to Railway.
That revenue-per-employee ratio would be notable for a mature SaaS company, and it stands out further for an infrastructure provider that now owns data centers. It also helps explain why Cooper emphasizes that the Series B was strategic rather than existential: he characterizes the business as “default alive,” suggesting that it could sustain itself without additional capital but raised to accelerate its plans.
Railway’s go-to-market motion has also been atypical. The company hired its first salesperson only last year and has just two solutions engineers. Nearly all of its two million users found the platform through word of mouth — developers telling peers that the product “actually works,” as Cooper puts it. In practice, that has meant five years of building largely for, and with, a grassroots developer base rather than marketing to enterprise CIOs.
For engineering leaders evaluating the platform, this history cuts both ways. On one hand, it signals a strong product-led growth engine and a community that is willing to adopt new infrastructure without heavy sales pressure. On the other, it means Railway is only now beginning to invest in the kind of enterprise-facing processes, documentation, and account management functions that large organizations typically expect from core infrastructure vendors.
Early Enterprise Traction and Compliance Posture
Despite its developer-first origin, Railway reports that 31% of Fortune 500 companies now use the platform in some capacity. These deployments range from small team projects to broader infrastructure footprints, rather than being uniformly company-wide migrations.
Named customers include Bilt, Intuit subsidiary GoCo, TripAdvisor’s Cruise Critic, and MGM Resorts. One representative example of how Railway is being used in a more infrastructure-intensive context comes from Kernel, a Y Combinator-backed startup that provides AI infrastructure to over 1,000 companies. Kernel runs its entire customer-facing system on Railway for $444 per month. CTO Rafael Garcia contrasts this with his prior experience at Clever, where he had six full-time engineers just to manage AWS; at Kernel, his six engineers all focus on product, not cloud plumbing. He describes Railway as the tool he wishes he had in 2012.
For larger organizations, compliance and integration with existing security practices are often gating factors. Railway offers SOC 2 Type 2 compliance and HIPAA readiness, including business associate agreements on request. The platform supports single sign-on, provides detailed audit logs, and can be deployed in a “bring your own cloud” model, allowing enterprises to host workloads within their existing cloud environments while still using Railway’s orchestration layer.
Enterprise pricing is customized, with add-on fees for extended log retention ($200 per month), HIPAA BAAs ($1,000), enterprise support with service-level objectives ($2,000), and dedicated virtual machines ($10,000). For cloud architects, these offerings indicate that Railway is actively targeting regulated and high-governance environments, not just hobby projects and early-stage startups — but as with any emerging provider, due diligence around integration depth, support maturity, and long-term vendor risk remains essential.
Competing with Hyperscalers and the New Cloud Guard
Railway is entering a saturated field. On one side are the hyperscalers — AWS, Microsoft Azure, and Google Cloud Platform — whose services underpin much of today’s software. On the other side are a growing cohort of developer-focused platforms such as Vercel, Render, Fly.io, and Heroku, which aim to simplify deployment while often riding on top of the big clouds.
Cooper casts these competitors into two broad camps. In his framing, hyperscalers are constrained by their legacy revenue models: they maintain older, VM-centric systems that still generate substantial profit from customers who provision instances and use only a fraction of their capacity. That financial gravity, he argues, makes it difficult for them to fully commit to a cloud experience optimized around ultra-fast, usage-only workloads designed for AI agents.
In contrast, Railway’s startup peers tend to focus on slices of the stack — containers, or specific runtime environments — rather than the full infrastructure gamut. Railway differentiates itself by offering VM primitives, stateful storage, virtual private networking, automated load balancing, and databases including PostgreSQL, MySQL, MongoDB, and Redis, wrapped in a UI targeted at both human developers and AI agents.
On the capacity side, Railway supports up to 256 terabytes of persistent storage with over 100,000 IOPS, and allows enterprise customers to scale individual services up to 112 vCPUs and 2 terabytes of RAM. Deployment regions currently span four global locations across the U.S., Europe, and Southeast Asia. These specs align Railway with mid- to large-scale application workloads, not just small side projects, though the company does not position itself as a general-purpose supercomputing or GPU-heavy training provider.
For engineering decision-makers, the competitive question is whether Railway’s vertically integrated, AI-native pitch translates into enough practical differentiation to justify adding another core platform to the stack — or even replacing parts of existing AWS or GCP deployments. History is full of infrastructure startups that failed to meaningfully dent hyperscaler dominance; Railway is attempting to avoid that outcome by attacking at the intersection of cost, simplicity, and AI-driven development speed.
AI-Native Loops: Deployments Driven Directly by Agents
The most forward-looking part of Railway’s strategy is its bet that AI systems will increasingly drive the software lifecycle end-to-end. Cooper predicts that the amount of software coming online over the next five years will be “a thousand times” what exists today, driven not by a proportional increase in human developers, but by AI coding tools becoming ubiquitous.
To prepare for that scenario, Railway has already begun integrating directly with AI agents. In August 2025, the company released a Model Context Protocol (MCP) server that allows AI coding assistants to deploy applications and manage infrastructure from within code editors. Cooper describes these as loops where systems like Claude can “hook in, call deployments, and analyze infrastructure automatically.”
The implication for cloud engineers is that the interface to infrastructure could increasingly be an AI agent mediating between developers and the platform, rather than human-authored Terraform or bespoke deployment scripts. Cooper argues that the definition of “developer” itself is changing: individuals no longer need deep engineering expertise to build systems, only the ability to think critically and reason about how components should fit together, with AI handling much of the implementation and deployment.
Railway’s focus on “agentic primitives” — features designed explicitly for machine-driven orchestration — is intended to position the platform as a natural target for these workflows. If AI agents are iterating on code in seconds and using MCP integrations to continuously deploy and inspect services, sub-second deployment and granular per-second billing become operational necessities rather than optimizations.
How broadly and how quickly this agent-driven model will be adopted remains uncertain. Many organizations are still in the early stages of integrating AI assistants into developer workflows, and concerns about safety, governance, and observability for AI-managed production systems are far from resolved. But Railway is clearly orienting its roadmap around the assumption that these concerns will be worked through, and that infrastructure designed for human-speed operations will look increasingly out of step.
Where the $100M Goes: Scaling Data Centers, Team, and Go-To-Market
The newly raised $100 million will fund three main initiatives: expanding Railway’s global data center footprint, growing its 30-person team, and building a more conventional go-to-market operation. For the first five years, the company largely eschewed marketing and sales, relying instead on product-led adoption among developers. Cooper now describes 2026 as the year Railway plans to “play on the world stage.”
On the infrastructure side, more data centers and regions should translate into lower latency and better resiliency for global customers, as well as more capacity for large enterprise workloads. On the organizational side, hiring in sales, support, and solutions engineering will be key to winning and keeping larger accounts, especially those with strict procurement, compliance, and support expectations.
The investor roster underscores how central Railway’s thesis is to the broader developer tooling ecosystem. Angels include figures such as GitHub co-founder Tom Preston-Werner; Vercel CEO Guillermo Rauch; Cockroach Labs CEO Spencer Kimball; Datadog CEO Olivier Pomel; and Linear co-founder Jori Lallo — all leaders of companies that sit adjacent to, or on top of, core infrastructure. Their backing signals a belief that AI-era development patterns will reshape how infrastructure is consumed and that there is room for a new, AI-native provider in that stack.
Whether Railway can convert its strong grassroots traction into durable enterprise adoption is still an open question. The cloud market remains dominated by Amazon, Microsoft, and Google, and many promising challengers have failed to escape niche status. Cooper, who previously worked as a software engineer at Wolfram Alpha, Bloomberg, and Uber before founding Railway in 2020, is explicit about the scale of his ambition: he envisions Railway as “the place where software gets created and evolved, period,” promising instant deployment, effectively infinite scale, and “zero friction.”
For now, the company has proven that a sizeable population of developers will seek out and adopt a faster, cheaper, AI-aware cloud platform without traditional marketing. The coming years will test whether that same value proposition — sub-second deployments, usage-only pricing, vertical integration, and agentic workflows — is compelling enough for risk-conscious enterprises to entrust Railway with a growing share of their production workloads.