Modal: Two Olympiad Gold Medalists Building Cloud Infrastructure That Feels Like Magic

Key Facts

• Modal Labs is a New York-based serverless cloud infrastructure company, founded in January 2021 by Erik Bernhardsson and Akshat Bubna, that quadrupled its valuation from $1.1 billion in September 2025 to $4.65 billion by May 2026.
• Who founded Modal Labs: Erik Bernhardsson, former Spotify machine learning lead and Better.com CTO, and Akshat Bubna, a former Scale AI engineer — both gold medalists at the International Olympiad in Informatics.
• How fast has its valuation grown: From $1.1 billion in September 2025, to a reported $2.5 billion in talks by February 2026, to $4.65 billion in a confirmed Series C by May 2026.
• What does Modal actually do: A serverless platform that lets developers run Python functions in the cloud with GPU acceleration by simply adding a decorator to their code, with sub-second cold starts and usage-based, per-second billing.
• What is Modal built on: The company built its own infrastructure from scratch in Rust, including a custom file system, container runtime, scheduler, and GPU memory snapshotting technology.

From Spotify's Recommendation Engine to Reinventing Cloud Infrastructure

Modal's founding story traces directly back to a specific, deeply felt frustration its CEO carried across nearly a decade of prior engineering work. Erik Bernhardsson spent seven years at Spotify, where he built the company's music recommendation system and created Luigi, a widely adopted open-source workflow scheduler, before later serving as CTO at Better.com. According to his own written account published in December 2022, Bernhardsson had grown convinced that data teams across the industry were spending far too much time fighting infrastructure friction rather than doing actual data work: "I have been working on a new tool called Modal that makes it a lot easier for data teams to deal with infrastructure," he wrote, describing his goal as building something that "takes code on a user's computer and launches it in the cloud within a second".

Bernhardsson founded Modal Labs in January 2021, with Akshat Bubna — a former Scale AI engineer — joining as co-founder and CTO that August. Silicon Valley Invest Club's profile of the company adds a striking shared credential between the two founders: both are International Olympiad in Informatics gold medalists, Bernhardsson representing Sweden in 2003 and Bubna representing India in 2014 — an unusually rare pairing of elite competitive programming pedigree at the very top of a company now building some of the most technically demanding infrastructure in the AI industry.

The Product: Turning Any Python Function Into Scalable Cloud Compute

Modal's core technical innovation is deceptively simple to describe but genuinely difficult to execute well: developers import the Modal library, wrap an ordinary Python function with a decorator like @app.function(), and call .remote() to instantly execute that code in the cloud, with automatic GPU acceleration and autoscaling, all without manually provisioning servers, configuring containers, or managing scaling infrastructure themselves. Bernhardsson's own framing from that original 2022 post makes the underlying ambition clear: "the foundational building block in Modal is a decorator that takes any Python function and moves its execution to the cloud," describing the resulting system as "a general purpose compute layer" usable for data pipelines, cron jobs, large-scale simulations, and countless other workloads beyond just AI.

Startup Intros' technical summary of the platform's capabilities highlights "sub-second cold starts, instant autoscaling, and elastic GPU scaling," describing the approach as letting developers manage infrastructure entirely "via code," which the company claims "significantly outperform[s] traditional methods". That "infrastructure-as-code" philosophy directly targets a specific, widely felt pain point: Sacra's analysis notes traditional cloud providers "require developers to manage containers, scaling, and resource allocation manually, creating friction in the development cycle," while Modal "eliminates this overhead" and charges "customers by the second for actual compute usage, eliminating idle resource costs that plague traditional cloud providers".

Built From Scratch, Not Bolted on Top of Existing Cloud Tools

A crucial technical detail separating Modal from many infrastructure companies making similar developer-experience promises: rather than building a thinner, more convenient layer on top of existing cloud provider primitives, Modal's own company page states directly that "we went deep and built our own infrastructure — including our own custom file system, container runtime, scheduler, container image builder, and much more". Silicon Valley Invest Club's profile adds that this entire stack was "built... from scratch in Rust," including "GPU memory snapshotting that delivers sub-second cold starts" — a genuinely deep, foundational engineering investment rather than a superficial wrapper, and one that likely explains much of why Modal's specific performance characteristics have proven difficult for competitors to replicate quickly.

Modal's own company page lists its primary customer use cases as spanning "Generative AI inference, LLM fine-tuning, computational biotech, and media processing" — a genuinely broad range of compute-intensive workloads unified less by a specific industry vertical than by a shared underlying need for fast, reliable, elastically scalable compute without the operational overhead of managing that infrastructure directly.

An Astonishing Valuation Curve: 4x in Eight Months

Modal's fundraising trajectory accelerated with striking speed through late 2025 and into 2026. The company achieved unicorn status in September 2025, closing an $87 million Series B led by Lux Capital, at a $1.1 billion valuation, bringing total funding at that point to $111 million. SiliconANGLE's coverage of that round framed Modal's ambition explicitly: to "become the infrastructure platform of choice for every stage of artificial intelligence development".

By February 2026, reporting indicated Modal was already in talks for a new round that would roughly double that valuation to approximately $2.5 billion, with the company disclosed to have surpassed its $200 million annual recurring revenue target for 2025, ultimately hitting $240 million. That underlying revenue growth accelerated further still: Sacra's analysis estimates Modal Labs hit $300 million in annualized revenue by April 2026, up sharply from roughly $119 million just at the end of 2025 — a genuinely rare pace of revenue expansion, more than doubling in under four months.

That revenue trajectory culminated in a confirmed Series C round reported by Reuters in May 2026: Modal raised $355 million, valuing the company at $4.65 billion — more than four times its valuation from the prior fall. CEO Bernhardsson told Reuters directly that the round reflected "AI coding" taking off as a major growth driver, with the company providing AI firms both direct chip access for running inference and "a sandbox product that enables developers to experiment with their code — often generated by AI — prior to integrating it into their applications". Silicon Valley Invest Club specifically named General Catalyst and Redpoint Ventures as leading that round, with additional participation from Accel, Menlo Ventures, and Bain Capital Ventures.

The "Sandbox" Angle: Betting on AI-Generated Code Needing a Safe Place to Run

Bernhardsson's own framing of the Series C round to Reuters points to a specific, forward-looking bet embedded in Modal's growth strategy: as AI coding assistants and autonomous coding agents generate an increasing share of the world's software, that AI-generated code needs somewhere safe and fast to actually execute and be tested before it's trusted and integrated into real production applications. Modal's sandbox product is positioned to be exactly that execution environment — a detail that connects directly to its customer roster, which specifically includes Cognition, the company behind the AI coding agent Devin, suggesting Modal has identified and is actively capturing a distinct, fast-growing niche within the broader AI infrastructure market: providing the safe execution layer underneath the rapidly expanding category of AI systems that themselves write and run code.

A Customer List That Reads Like a Who's Who of Frontier AI

Modal's enterprise customer base by mid-2026 included some of the most prominent names across the AI industry: Anthropic, Meta, DoorDash, Cognition (maker of Devin), Suno, Physical Intelligence, and Chai Discovery — a roster spanning frontier model labs, established technology giants, AI coding agents, generative music, and robotics. Silicon Valley Invest Club's profile notes that over 1 billion sandboxes have been launched cumulatively on Modal's platform, a scale metric reflecting genuinely massive, sustained production usage rather than early-stage experimentation.

Why "Magic" Is the Right Word for What Modal Is Selling

A 2024 talk title asked directly, "Can AI Infrastructure Work Like Magic?" — and that framing captures precisely what Modal's rapid growth reflects about the broader AI infrastructure market in 2025 and 2026. As more of the world's most consequential AI companies — from Anthropic's frontier models to Suno's generative music to Cognition's autonomous coding agents — need to run massive, unpredictable, GPU-intensive compute workloads reliably and instantly, the specific value of infrastructure that genuinely removes operational friction, rather than merely making that friction slightly more bearable, compounds dramatically. Modal's willingness to build its entire technical stack from the ground up in Rust, rather than accepting the constraints of existing cloud provider primitives, appears to be the foundational engineering decision that let the company deliver on that "magic" promise convincingly enough to quadruple its valuation in under a year.

FAQ

Source: CrackTheDeck Research.