Extend your neocloud with a storage line of your own
We fund, install and run S3-compatible object storage, including inside your own data center. You sell it to your GPU customers under your own brand, as part of your own product.
No hardware to buy · No egress fees · S3-compatible
- S3-compatible API
- No egress fees, no request charges
- Storage installed in your data center
- No hardware to buy, no team to hire
Your customers store more than they compute
Training data, checkpoints, model artifacts and job output all have to live somewhere. Right now most of it lives somewhere you do not bill for.
Your fast tier holds finished work
The tier you bought for the working set ends up holding completed checkpoints and last quarter's datasets, because there is nowhere cheaper to put them that is still close by.
Everything else leaves the building
Overflow goes to a general-purpose cloud somewhere else, across the public internet, with a meter running on every read back.
The storage line is not yours
The colder the data, the more of it there is. That is a growing line on your customer's bill, and today none of it appears on your invoice.
An S3 tier built to be resold
Object storage is a commodity. One that sits next to your compute and is sold by you is not.
In your data center
We can install and run object storage inside your own facility, next to your GPU nodes, instead of a region away. Talk to us about what a site needs.
S3 as your customers know it
A standard S3-compatible API with path-style addressing, multipart upload and presigned URLs. Any tool with an S3-compatible target points at it. Nothing to rewrite.
Multi-tenant from the start
Organization roles for Owner, Admin, Member and ReadOnly, with access keys revoked automatically when someone is demoted.
Keys scoped per bucket
Access keys are scoped to a single bucket and region, with read, write, list and delete granularity and an optional expiry date.
Retention your customers can rely on
Object lock with governance or compliance retention, alongside versioning. Both are set when the bucket is created.
An audit log you can export
Account activity is recorded in an audit log that exports to CSV. Sign-in is protected with MFA and passkeys.
Start on our regions, then move it into your building
On our regions
Europe (France) and US East (Michigan) today, with more regions on the way. A bucket's region is fixed when you create it, and an access key is scoped to one region. Start here while we talk about your site.
In your data center
We install and operate the storage inside your own facility, cross-connected to your GPU nodes, so the bytes never take a trip across the public internet.
You sell it, you price it
Storage becomes a product on your line card, under your own brand, without a purchase order and without a hire.
One metered line
Storage is the only thing metered. No egress charges and no per-request charges. Our public rate is $4.99/TB/month; partner terms are agreed with our team.
We fund and run the hardware
Our capex, our racks, our operations. You add a storage product to your line card without a purchase order and without hiring a storage team.
White label, end to end
Your brand on the console, the docs and the S3 endpoint. Your customers buy storage from you, on your paper, alongside the compute they already rent. Fil One stays behind the curtain unless you want us in front of it.
Metering that feeds your billing
Per-bucket usage is exported to you, so the storage line lands on your own invoice next to the GPU hours, priced the way you choose.
One workload, one site, one conversation
Tell us about the site
Where your compute sits, what your customers store today, and which workload annoys you most: checkpoint sprawl, dataset reloads or a backup target you overpay for.
Start on our regions
Point one cluster at a Fil One region and run it for real, while we work through the practicalities of putting storage in your building.
Agree the commercials
We settle the rate, the terms and how your customers are billed, so the storage line is one you can quote and sell with confidence.
Sell it as your own
Storage becomes a line on your invoice next to the GPU hours, with no hardware for you to buy and no product for you to build.
Frequently asked questions
Fil One speaks the S3 API, so you keep the SDK and CLI you already use. Point it at your region’s endpoint, switch on path-style addressing, and authenticate with a Fil One access key. Everyday object work behaves the way your code expects: put, get, head, list, delete, presigned URLs, and multipart upload from your own SDK.
Buckets are private by default. For the operation-by-operation detail, including the calls that differ between regions, see the S3 compatibility reference in the docs.
Read Fil One docs, access the app to get started with no code required, or talk to someone on our team to get started.
Objects are encrypted at rest by the storage gateway using AES256 (SSE-S3), and every request runs over TLS. It is on by default with nothing to configure. Keys are held by the regional storage operator that fulfils your bucket’s region, so if your policy requires you to hold the keys yourself, encrypt client-side before upload with the standard S3 tooling and Fil One stores the ciphertext. Our services are delivered through top-tier data centers that are certified to ISO 27001, SOC 2, and PCI DSS standards. Reach out to security@fil.one for compliance documentation or any other security questions.
Egress is any data transferred out of your bucket: to the internet, to another cloud, or to your own servers. On a paid plan it is free at any scale, and there are no per-request charges either. The 30-day trial includes 2 TB of egress.
Billing is $4.99 per TB stored per month, with no fees for egress or API operations, subject to a $4.99 monthly minimum. Store under 1 TB and you pay the minimum; store more and you pay per TB for what you use.
Yes. Teams with predictable storage needs can talk to us about multi-year committed capacity. Contact sales and we will put a quote together.
You choose when you create the bucket: Europe (France) or US East (Michigan), with more regions on the way. A bucket's region is fixed at creation, so data written there stays there. Each region has its own endpoint, and an access key is scoped to a single region.
Running a GPU cloud? We can also install storage directly inside your own data center. See our partner program.
Put a storage line next to your GPU hours
Tell us where your compute sits and which workload you would start with. We will take it from there.