Modal Labs valued at $4.65 billion after $355m Series C raise

Modal Labs, a serverless cloud platform that lets developers build and run AI applications without managing infrastructure, has raised US$355 million ($530 million) in Series C funding at a US$4.65 billion valuation, CEO Erik Bernhardsson told Reuters exclusively. The raise more than quadruples the company’s US$1.1 billion valuation from September 2025, and its annualised revenue run rate has climbed to US$300 million from US$60 million over the same eight-month period.

The numbers are eye-catching even by the standards of an AI infrastructure market that has been among the hottest startup categories of 2025 and 2026. But the Modal raise is more than another big-ticket funding round. It crystallises a structural shift in how enterprises buy AI compute — and it arrives just days after Google and Blackstone announced their own US$5 billion ($7.5 billion) joint venture to build an AI cloud, a move that sent shares of neocloud competitor CoreWeave lower.

“Coding for the last six months has been driving everything,” Bernhardsson told Reuters. The remark points to the engine under Modal’s growth: AI-assisted software development tools are consuming GPU capacity at a rate that few forecasters anticipated. Morgan Stanley now estimates that 70 per cent of GPU cloud spend will go to inference rather than training by late 2026, a sharp reversal from the training-heavy pattern of 2023 and 2024.

Modal’s serverless architecture is purpose-built for this pivot. Unlike traditional GPU cloud providers that rent dedicated machines by the hour, Modal lets developers run Python functions on GPUs that scale to zero when idle and cold-start in under a second. The company’s own cost analysis, cited by independent researcher Sean Kim, claims serverless GPU pricing beats reserved instances by roughly 50 per cent for workloads that do not run continuously.

Unless your GPU runs consistently for three years, serverless beats reserved pricing.
— Modal Labs cost analysis

Yet the valuation invites scrutiny. Modal went from US$60 million to US$300 million in annualised revenue inside eight months — a pace that raises the question of how much of the surge is a temporary spike from AI coding hype versus repeatable enterprise demand. The round was structured in two tranches, with the first closing at a US$2.5 billion valuation and the second at US$4.65 billion within months as revenue accelerated. Two-tranche milestone financings are becoming more common in AI infrastructure, but they also signal that investors are repricing on quarterly proof points rather than committing to a single number — a structure that leaves less room for error if GPU supply eases and customers migrate to cheaper alternatives.

The competitive landscape is fragmenting rather than consolidating. Modal occupies a distinct niche — serverless inference for Python developers — that differs sharply from CoreWeave’s large-scale Kubernetes clusters built for training, and from Lambda Labs’ straightforward GPU virtual machines aimed at researchers. A benchmark comparison published in March priced Modal’s H100 instances at US$2.89 per hour with per-second billing, against CoreWeave’s US$2.23 per hour with a 10-minute minimum. Each platform serves a different workload profile and customer type, and the market appears to be supporting all of them simultaneously.

Then there is the Google-Blackstone venture. Announced on 19 May, the US$5 billion partnership will build a US-based AI cloud powered by Google’s custom TPU chips and Blackstone’s capital.

The small fry are getting squeezed out.
— Equities portfolio manager, Business Insider

The deal is an explicit threat to independent neoclouds: a hyperscaler-backed competitor with preferential access to chips and near-unlimited balance-sheet capacity could compress margins across the sector.

Modal’s response, intentional or not, is visible in its supplier count. The company now works with 13 cloud compute providers, up from five a year ago, Bernhardsson said — including some he had never heard of before the current supply squeeze. That multi-supplier strategy is both a hedge and a signal: when a US$4.65 billion company is scouring the market for GPU capacity, it means compute remains the binding constraint on AI growth, and the platforms that can secure supply will set the terms.

For Australian enterprises, the neocloud wave matters. Local organisations building AI applications currently route most inference workloads through US hyperscalers, where data egress fees and latency can become visible line items as usage scales. Modal’s multi-cloud model — aggregating capacity across 13 suppliers and presenting it as a single serverless surface — points to a future where AI compute is bought more like electricity than like a managed service. Australian providers such as Sharon AI and Firmus, which have positioned themselves as sovereign compute alternatives, stand to benefit from the same dynamics that have propelled Modal’s valuation, provided they can secure GPU supply.

The Modal raise does not settle the question of who wins the AI cloud. It shows the market is big enough — and differentiated enough — for multiple architectures to coexist, and that the companies solving the hardest supply-chain problems are the ones investors are betting on.