From Basement Rigs to Backbone Infrastructure: How Runpod Quietly Built a Profitable Lane in the AI cloud race

Runpod’s rise from a pair of repurposed cryptocurrency mining rigs in New Jersey basements to a global AI app hosting platform reads like a case study in timing, technical intuition, and stubborn bootstrapping.

The startup’s trajectory cuts against much of the prevailing narrative around the AI boom, where scale is often bought with vast amounts of venture capital long before revenues materialize. In just four years, the AI app hosting platform has grown to a $120 million annual revenue run rate, largely by solving a problem developers themselves were complaining about, staying disciplined on costs, and arriving just early enough to benefit from the generative AI explosion that followed.

Founded by Zhen Lu and Pardeep Singh, two former Comcast developers, Runpod did not begin as a grand vision to challenge the hyperscalers. It emerged instead from a failed crypto-mining experiment and a practical need to salvage expensive hardware sitting idle in their homes.

In late 2021, Lu and Singh had built Ethereum mining rigs in their New Jersey basements, investing roughly $50,000 between them. The returns were modest, the work quickly became repetitive, and the looming Ethereum “Merge” meant mining would soon end altogether. More pressing was the domestic reality: they had convinced their wives to support the investment and needed to show it was not money wasted.

Both founders had experience working on machine learning projects in their day jobs, so they decided to repurpose the GPUs for AI workloads. That decision exposed a deeper frustration. Lu recalled that the software stack for running and developing on GPUs was clumsy, brittle, and unfriendly to developers. Configuration was time-consuming, tooling was fragmented, and getting from idea to deployment was far harder than it needed to be.

That frustration became the seed for Runpod. The founders set out to build a platform that prioritized developer experience, offering fast access to GPUs, flexible configurations, and tools that developers already understood. By early 2022, they had assembled a working product with APIs, command-line interfaces, and integrations such as Jupyter notebooks, alongside a serverless option that automated much of the underlying setup.

What they lacked was visibility. As first-time founders, they had no marketing playbook and no sales team. Lu turned to Reddit, posting in AI-focused subreddits with a straightforward pitch: free access to GPU servers in exchange for feedback. The response validated their instincts. Developers signed up, tested the platform, and began paying for it. Within nine months, Runpod had crossed $1 million in revenue, enough for Lu and Singh to leave their jobs.

But growth brought new complications, as early customers were hobbyists and researchers, but businesses soon followed, and they were unwilling to run production workloads on servers hosted in private homes. Rather than immediately raising venture capital, the founders pursued revenue-sharing agreements with data centers to scale capacity. The approach allowed them to grow without dilution, but it required constant vigilance.

Singh said capacity was existential. If developers logged in and found no GPUs available, they would simply move on. The risk intensified after the launch of ChatGPT, which triggered a surge in demand for AI infrastructure and pushed Runpod’s Reddit and Discord communities into rapid expansion.

For nearly two years, Runpod operated without external funding. It never offered a free tier and refused to take on debt, even as other AI cloud providers subsidized growth. Every workload had to pay its way. Lu said that constraint forced discipline early and shaped how the company thought about pricing, reliability, and trust.

Venture capital eventually found them anyway. Radhika Malik, a partner at Dell Technologies Capital, noticed Runpod through Reddit discussions and reached out. Lu admitted he had little understanding of how investors evaluated startups. Malik, he said, helped demystify the process while continuing to monitor the company’s progress.

By May 2024, with AI app development accelerating and Runpod serving around 100,000 developers, the company raised a $20 million seed round co-led by Dell Technologies Capital and Intel Capital. The round included high-profile angels such as Nat Friedman and Hugging Face co-founder Julien Chaumond, who had independently discovered Runpod as a user and contacted the team through customer support.

Since then, Runpod has continued to scale without raising additional capital. The platform now counts roughly 500,000 developers as customers, ranging from individual builders to Fortune 500 enterprises with multimillion-dollar annual contracts. Its infrastructure spans 31 regions globally, and its customer list includes names such as Replit, Cursor, OpenAI, Perplexity, Wix, and Zillow.

The competitive environment is crowded and unforgiving. Hyperscalers like Amazon Web Services, Microsoft, and Google dominate the cloud market, while specialized providers such as CoreWeave and Core Scientific focus heavily on AI workloads. Runpod’s founders do not frame their ambition as replacing those players. Instead, they position the company as a developer-first layer, built by people who felt ignored by existing tools.

Lu argues that software development itself is changing. Rather than disappearing, programmers are becoming operators of AI agents and systems, orchestrating models rather than writing every line of code by hand. Runpod, he said, wants to be the environment that those developers learn from and trust as their needs evolve.

With a $120 million annual revenue run rate, a large global footprint, and a product shaped by years of direct engagement with developers, Runpod is now preparing for a Series A raise from a position few AI infrastructure startups can claim: profitability-driven growth rather than promise-led expansion.