ZeroFS claims major performance wins against S3 for small files

ZeroFS, a new object storage alternative, published benchmarks claiming significant speed and cost advantages over Amazon S3. We analyze the methodology and what it means for founders.

For developers building applications that handle a high volume of small files, like user avatars or configuration data, ZeroFS makes a compelling pitch. Its self-reported benchmarks target a known S3 pain point: latency and per-request pricing on tiny objects. If these claims hold, it’s a service worth evaluating for specific, performance-sensitive workloads. However, teams that depend on the battle-tested reliability, global footprint, and deep ecosystem integrations of AWS S3 should skip it for now. The bottom line is that ZeroFS presents a plausible, yet entirely unverified, case for a specific niche. Independent validation is required before it can be considered for production systems of record.

Methodology

This is a v0 review based exclusively on the founder's technical blog post, published on July 12, 2026. We have not conducted independent performance benchmarks. The analysis covers the claims made in the source, the provided benchmark setup, and the strategic positioning against Amazon S3. What is not covered is independently verified performance, long-term data durability, behavior under real-world network conditions, or the completeness of its S3 API compatibility. All performance figures cited here are claims made by ZeroFS and should be treated as such until reproduced by a third party. Our assessment focuses on the validity of the comparison and the trade-offs a developer would be making.

What It Does

An S3-compatible API

ZeroFS presents itself as a drop-in replacement for S3 for basic object storage operations. It uses an S3-compatible API, meaning existing AWS SDKs and tools should, in theory, work by changing the endpoint URL. This is a critical feature for adoption, as it lowers the switching cost for applications already architected around S3's object storage model.

Optimized for small files

The core technical claim is that ZeroFS is architected to minimize the overhead on small file operations (e.g., files under 1MB). The founder's post argues that S3's architecture, built for massive-scale durability and large objects, imposes a latency penalty on small GET and PUT requests. The published benchmarks claim significantly lower p99 latencies for these operations, which could be meaningful for applications like chat services or IoT data ingestion that generate millions of tiny files.

A simplified pricing model

The service contrasts its pricing with S3's multi-vector model. Where S3 charges for storage, data transfer, and different classes of operations (GET, PUT, LIST), ZeroFS claims to offer a simpler plan. This typically involves a flat per-gigabyte storage cost with a generous or unlimited allowance for API operations. This model is designed to be more predictable for workloads with high request volumes, where S3's per-operation fees can become a surprisingly large part of the bill.

What's Interesting / What's Not

The most interesting aspect of ZeroFS is its explicit focus on a known S3 weakness. The small-file problem is real. For many SaaS applications, S3's per-request costs for PUT and LIST operations can exceed storage costs. A solution that credibly eliminates this trade-off is valuable. The provided benchmarks, while self-reported, appear to use a specific and reproducible setup (e.g., specific instance types, regions, and file sizes), which is a positive sign of transparency.

What's less convincing is the implicit scope of the comparison. The benchmarks likely represent an ideal-case scenario, with the test client located in the same cloud region and perhaps the same VPC as the ZeroFS service. This minimizes network latency, a factor that would heavily influence real-world performance for a global user base. The comparison also conveniently ignores the vast ecosystem and non-core features that make S3 the default choice. There is no mention of data durability guarantees (S3's famous 11 nines), cross-region replication, object lifecycle policies, event notifications, or deep IAM integration. The comparison is purely on speed and cost for a narrow set of operations, which is a very different proposition from being a true S3 replacement.

Pricing

Pricing as of July 2026, based on the blog post.

• ZeroFS Free Tier: Up to 10 GB storage and 1 million API operations per month.
• ZeroFS Pro Tier: $15 per TB per month ($0.015/GB). API operations are included.
• Amazon S3 Standard (for comparison): ~$23 per TB per month ($0.023/GB). This excludes significant additional costs for operations ($0.005 per 1,000 PUT requests) and data transfer.

Verdict

ZeroFS is a promising challenger for a very specific workload: high-volume, small-file I/O where S3's request latency and pricing model are genuine bottlenecks. The founder's claims of superior performance and simpler pricing are compelling on paper. However, these claims remain unverified. For any production system, switching from S3 means trading a decade of proven reliability and an unmatched feature set for a claimed performance gain on a narrow metric. Developers should use ZeroFS for new, non-critical projects where the small-file workload is dominant and the risks of a new, unproven provider are acceptable. For everything else, stick with S3 until independent benchmarks and long-term user reports validate the claims.

What We'd Test Next

In a v2 review, we would first aim to reproduce the founder's benchmark in a controlled, third-party environment to verify the core performance claims. Next, we would design tests to measure real-world performance, specifically by running clients from different geographic regions and over variable public internet connections. We would also evaluate the completeness of the S3 API compatibility by testing a suite of common tools like rclone, s3cmd, and various language SDKs. Finally, we would investigate failure modes and data durability by simulating node outages to understand how the system behaves under stress.

The investor read

ZeroFS signals the continued unbundling of cloud infrastructure. While S3 is the general-purpose default, specialized providers can build profitable businesses by optimizing for a single dimension that incumbents don't prioritize. This follows the pattern of Wasabi (hot storage), Backblaze B2 (cheap archival), and Cloudflare R2 (zero egress). ZeroFS targets the high-volume, small-file niche where S3's request pricing creates an opening. For the company to be venture-investable, it must move beyond self-reported benchmarks to third-party validation. The key risk is go-to-market; displacing a default service like S3 requires a powerful developer-led adoption loop or a compelling open-source offering. Without that, it's more likely a sustainable, bootstrapped business than a breakout venture.

Pull quote: “For any production system, switching from S3 means trading a decade of proven reliability and an unmatched feature set for a claimed performance gain on a narrow metric.”