Fly.io Sprites Offer Fast-Booting VMs with Persistent Disk and Auto-Sleep

Fly.io's new Sprites product introduces Linux virtual machines designed for instant creation, persistent storage, and cost-effective auto-sleep, diverging from their container-based Fly Machines.

The Answer Up Front

Sprites are for developers who need ephemeral, stateful compute environments that boot in seconds, offer root access, and come with durable storage, all while minimizing costs during inactivity. If your workflow involves frequent, short-lived tasks, interactive shells, or build environments where a traditional VM is overkill and a serverless function is too restrictive, Sprites are a compelling option. Developers already using Fly.io for other services might find Sprites a natural extension. Skip Sprites if your primary need is long-running, always-on services with predictable, high-concurrency traffic, where the overhead of even a fast cold-start is unacceptable. The bottom line is that Sprites carve out a new niche for disposable, persistent compute, bridging the gap between serverless functions and full-fledged virtual machines.

Methodology

This v0 review draws on the founder's published claims at the Fly.io blog post, "The Design & Implementation of Sprites," accessed on 2026-07-07. Independent benchmarks are pending. Update cadence: re-tested when claims diverge from observed behavior.

This review covers Sprites, a newly launched product from Fly.io. The source signal details the design philosophy and the first of three core technical decisions behind Sprites. Specifically, we analyze the founder's claims regarding instant creation, persistent disk, auto-sleep, and the strategic shift away from container images for this specific offering. What is not covered in this review includes independent performance benchmarks, long-term workflow integration, cost analysis under various load patterns, or edge case behaviors. Our assessment is based solely on the architectural details and stated benefits presented in the initial announcement.

What It Does

Fly.io Sprites are Linux virtual machines designed as "ball-point disposable computers." The core proposition is to provide compute resources that are available within "a second or two" of creation, mimicking the experience of SSH'ing into an already running machine. The founder highlights four key characteristics:

Instant Creation

Unlike traditional Fly Machines, which can take over a minute to create due to their OCI container-based architecture, Sprites claim near-instantaneous creation. This speed is central to their utility as disposable compute, allowing developers to spin up environments on demand without significant wait times.

Persistent Disk and Auto-Sleep

Each Sprite comes with a 100GB durable root filesystem. This persistent storage differentiates Sprites from many ephemeral serverless offerings. When inactive, Sprites automatically put themselves to sleep, entering a state that the founder claims costs "practically nothing." This combination of persistence and cost-effective inactivity aims to make them suitable for tasks that require state but are not continuously running.

No More Container Images

One of the "three big decisions" in Sprites' design is the departure from container images. Fly Machines are built on OCI containers repackaged as KVM micro-VMs, offering Docker-like ergonomics with VM isolation. For Sprites, this approach was deemed unsuitable for achieving the desired instant creation speed. The blog post indicates a new orchestration stack was developed, undoing some core decisions made for Fly Machines, to facilitate this faster, more scalable, and manageable VM creation process.

What's Interesting / What's Not

The most interesting aspect of Sprites is their positioning as a new primitive in cloud computing, explicitly designed to fill a gap between traditional VMs and serverless functions. The combination of instant creation, root access, persistent disk, and automatic, cheap sleep is genuinely novel. This allows for workflows that are difficult or expensive to achieve with existing tools, such as truly disposable development environments, interactive build agents, or short-lived data processing tasks that need to retain state. The "ball-point disposable computers" analogy effectively communicates the intent: resources that are always at hand and easily discarded.

The strategic decision to move away from OCI container images for Sprites, despite Fly.io's success with Fly Machines, is also notable. It signals a recognition that a one-size-fits-all compute primitive is insufficient for all use cases, and that optimizing for specific developer experiences (like instant creation for ephemeral tasks) may require fundamentally different architectural choices. The founder claims this new orchestration stack makes Sprites "drastically easier for us to scale and manage," which is a significant internal benefit that translates to a better external product.

What's less clear from this initial signal is the concrete performance characteristics beyond "a second or two" for creation. While the concept is strong, the specific latency for cold starts under various loads, the actual cost savings during sleep, and the performance of the persistent disk are all details that would require independent verification. The blog post outlines the design decisions but offers limited empirical data or specific use case examples beyond generic tasks. A deeper dive into the new orchestration stack's technical details would also be valuable, as the source only touches on the first of three big decisions.

Pricing

The founder claims Sprites "cost practically nothing while asleep." No specific pricing tiers, per-second rates, or free-tier limits are enumerated in the source material. This pricing snapshot is based on information available on 2026-07-07.

Verdict

Fly.io Sprites are a significant new offering for developers seeking highly ephemeral, stateful compute. They are best suited for interactive development, CI/CD steps, or any task requiring a Linux VM with root access and persistent storage that can spin up quickly and then cost-effectively go idle. The explicit departure from container images for this product underscores a commitment to optimizing for specific developer needs, rather than shoehorning all workloads into a single compute model. While the claims of instant creation and near-zero sleep costs are compelling, independent verification of these performance and cost metrics will be crucial for a full assessment. For those in need of disposable, persistent VMs, Sprites represent a strong contender.

What We'd Test Next

Our next steps would focus on empirical validation and deeper use case exploration. We would benchmark the actual creation times for Sprites across different regions and under varying network conditions, comparing them against both Fly Machines and other cloud VM/serverless offerings. We would also measure the latency of waking a Sprite from its auto-sleep state. A detailed cost analysis for various usage patterns, including frequent short bursts and extended periods of inactivity, would be essential to verify the "practically nothing" claim. Furthermore, we would test the performance and durability of the 100GB persistent root filesystem under different I/O loads. Finally, we would explore complex developer workflows, such as using Sprites for full-stack development environments, to assess the overall developer experience and integration capabilities beyond simple shell access.

The investor read

Fly.io's launch of Sprites signals a continued trend towards specialized compute primitives beyond generic containers. While containers offered a universal packaging format, the market is segmenting into highly optimized solutions for specific workloads: serverless functions for event-driven, stateless tasks; and now, Sprites for ephemeral, stateful, fast-booting VMs. This move by an established player like Fly.io validates the demand for a middle ground between restrictive serverless functions and slower, more expensive traditional VMs. For investors, this highlights a potential market for companies that can deliver novel virtualization or orchestration technologies to achieve such performance and cost profiles. A new entrant would need to demonstrate a clear technical advantage in boot times, idle cost efficiency, and developer experience to compete with Fly.io's existing infrastructure and brand. This is a strategic product expansion for Fly.io, targeting the long-tail of developer tasks and interactive environments, potentially increasing platform stickiness and total addressable market.