Fly.io closed a $25 million Series D co-led by Dell Technologies Capital and Intel Capital to accelerate its shift toward persistent “computers for agents” , while appointing former Docker CEO Scott Johnston as its new chief executive.
Fly.io’s $25 million Series D, co-led by Dell Technologies Capital and Intel Capital, with participation from Andreessen Horowitz (a16z), EQT, Geodesic, and Y Combinator, marks a strategic pivot and acceleration around “computers for agents”, persistent, stateful infrastructure purpose built for AI agents rather than disposable sandboxes.
The round brings total publicly reported funding to roughly $130–136 million. An SEC Form D showed ~$23.2 million sold (of a ~$25 million target) starting in early July 2026 under a Rule 506(b) private offering. Prior rounds included a ~$12 million Series A (Intel Capital-led, 2021), $25 million Series B (a16z-led, 2022), and $70 million Series C (EQT-led, 2023, at a reported ~$467 million post money valuation). Secondary market data around the new round points to a post money valuation near $475 million.
Martin Casado (a16z general partner) and Daniel Docter (Dell Technologies Capital managing director) joined the board. Founder Kurt Mackey transitioned from CEO to an advisory/board role focused on technical vision; Scott Johnston (former Docker CEO, with prior leadership at Puppet, Loudcloud, and Netscape) became CEO and a board member. Johnston previously scaled Docker through an identity crisis into strong growth (reportedly taking ARR from low tens of millions to well over $100 million).
What is Fly.io?
Fly.io, founded in 2017 (Chicago roots, now San Francisco-associated, remote first), originally built a global edge platform for running full stack apps and databases close to users with simple developer experience, owning hardware in colocations rather than purely renting hyperscaler capacity. Core offerings include Fly Machines (hardware isolated VMs from containers) with multi region support, durable storage, and networking.
The Series D funds a sharpened focus on agent workloads. The company argues that AI agents (coding agents, autonomous systems that build/operate/deploy software) need “real computers”: long running, stateful environments with durable disks, secure connectivity to other systems/APIs, network presence, and the ability to scale to millions of instances, unlike short lived sandboxes. Its key product here is Sprites: semi disposable persistent Linux environments (Firecracker based) with features such as:
- Instant create/checkpoint/restore.
- Durable storage (rebuilt Sprite Block Device stack enabling efficient drive forking/cloning of templates).
- Idle aware metering (compute pauses when inactive).
- Secure “Connectors” for authenticated outbound access without exposing credentials.
Traditional Fly Machines and PaaS features continue, but Sprites and agent infrastructure are now the primary company focus. Customer examples cited include Firecrawl, Kilocode, Plastic Labs, and Phonic.
Traction and Momentum
- 37,000 customers overall; >8,000 agent native.
- Strongest quarter in company history, driven almost entirely by agent workloads.
- Revenue from largest agent native customers grew nearly 12× over the prior 12 months.
- Among largest overall customers, agent native companies represent ~two thirds of revenue (agent workloads were described as a rounding error a year earlier; now ~66% of revenue from the top 100 customers).
This growth occurred while the company operated near self sufficiency after the large 2023 raise; the new capital is explicitly for acceleration amid the agent shift rather than survival.
Recommended: Lendistry Secures $100 Million Credit Facility From East West Bank
Mackey’s candid blog post framed an internal “identity crisis” resolved by betting fully on agents determining how most software is built/shipped (more personalized, smaller audience, adaptive software) and by recruiting Johnston for scaling execution. Johnston’s background in platform shifts (containers, DevOps, PaaS) aligns with defining a new category. Investors (long time backers Dell Tech Capital, Intel Capital, and a16z, plus EQT) emphasize the shift from human centric to agent centric infrastructure and Fly’s hardware ownership + developer friendly primitives as differentiation.
Fly positions itself against both traditional public clouds (AWS/GCP/Azure) optimized for fixed function, high scale apps and pure sandbox providers. Competitors in adjacent spaces include other edge/PaaS players (Render, Vercel, Heroku) and emerging agent infrastructure, but Fly stresses its combination of owned hardware economics, global footprint, durable stateful compute, and simplicity.
The raise is modest relative to 2026 AI mega rounds, reflecting a focused infrastructure bet rather than broad model training. Risks include execution on scaling Sprites/storage/connectors, competition from hyperscalers adding agent features, and the still evolving nature of production agent workloads. Upside lies in becoming the default execution layer as agents proliferate, supported by existing traction, technical differentiation (forkable durable disks, secure connectors), and experienced leadership for the growth phase.
The $25M Series D, leadership transition, and product sharpening signal Fly.io treating the rise of AI agents as a foundational platform shift comparable to earlier ones (containers, cloud, edge), with capital and talent deployed to own the “computers for agents” layer.
Please email us your feedback and news tips at hello(at)superbcrew.com
