SF Compute

We build and operate clusters
that you own, not lease.

You own the hardware and its residual value. We handle the design, build, and ongoing operations. Use the capacity for your customers, sell it through SFC, or do both.

Talk to us See how it works
Owned by you, operated by SFC
A connected GPU cluster: you own the hardware, SFC designs, builds, and operates it.GPU / 01GPU / 02GPU / 03GPU / 04
We built the neoclouds

Your team of supercomputing experts.

From power and cooling to GPU networking and cluster software, our engineers take your infrastructure from the drawing board to daily operations. You get the team behind the machine.

“SFC is the strongest infrastructure team I’ve worked with.”
Dan Wardle
Dan WardleVP Engineering, SFC
Talk to us
A cluster of your own

Built around your reasons to own.

Your capital & assetsYour own clusterYour customers & the market

Platforms

Control the hardware behind your product.

Build your platform as a service on hardware you own. Get deeper access, specialized configurations, and the residual value of the GPUs, with SFC running the infrastructure.

Talk to us

Funds & asset owners

Put capital, power, and land to work.

Design around your target internal rate of return (IRR), or get more from land, power, a data center, or an existing GPU cluster. We bring engineering, operations, and access to demand.

Already have a cluster? Start with our cluster and facility requirements.

Talk to us

Venture firms & nonprofits

Secure capacity for the people you back.

Give your portfolio or community a dedicated source of compute. We deploy and operate the cluster, and sell spare capacity until your teams need it.

Talk to us
Capacity you can resell

Resale added 35.2 percentage points of utilization.

Your demand changes. SFC finds buyers for the capacity you aren’t using. Here’s what that looked like for one customer.

Customer usage alone60.2%
Including SFC resale95.4%
Customer usage and resale, August 12–26From August 19 to 26, customer usage averaged 60.2% of reserved capacity. Resale added 35.2 percentage points, bringing combined usage and resale to 95.4%.
Talk to us
Capacity on your timeline

Sell by the month. Reclaim it when you’re ready.

SFC sells your capacity to other buyers month by month. When you need the GPUs, the capacity returns to you. Your cluster earns until you’re ready to use it.

Month 1Month 2Month 3Month 4
Buyer A
Buyer B
Buyer C
Buyer D
Buyer E
Buyer F
Your needs
Sold to other buyersReclaimed for your needs

Need out of your contract?

SFC puts your compute contract back on the market and sells the remaining capacity to another buyer.

Talk to us
From consultation to operations

You own the hardware.
SFC does the infrastructure work.

01 / Consultation02 / Design & build03 / Ongoing operations
  1. 1

    Consultation

    Start with your workloads, financial goals, location, and timeline. We evaluate what you already own and work through the hardware, power, cooling, and network design with you.

    A plan for your cluster, budget, and deployment.
  2. 2

    Design & build

    We coordinate the facility, equipment procurement, installation, and commissioning. We test the cluster as a complete system, from individual GPUs to the fabric connecting them.

    A commissioned cluster, ready for real workloads.
  3. 3

    Ongoing operations

    We handle provisioning, monitoring, maintenance, and incident response. Use the capacity for your own customers, sell it through the SFC market, or do both.

    Your asset, operated by our infrastructure team.
Talk to us

Deployment experience

We know where deployments go wrong.

“After deploying hundreds of megawatts of capacity, you learn what can go wrong: late equipment, power that isn’t ready, cooling that doesn’t perform, and a network that won’t pass acceptance. We’ve seen just about all of it. The job is to find those problems before they become yours.”
Eric Park
Eric ParkCTO, SFC · Cofounder and former CEO, Voltage Park
OEM relationshipsEquipment on handReady for deployment

Relationships with every major OEM

We work directly with equipment manufacturers to coordinate procurement, integration, and support for your deployment.

Long-lead equipment, already on hand

We stockpile equipment with long procurement lead times and bring it to your project to keep the build moving.

Talk to us
Programmable

Your customers don’t need
to know we’re here.

Think of Stripe for compute infrastructure. Your website is the frontend, your customers stay in your product, and SFC operates underneath. Our platform was designed for this use case.

Your website & brandSFC operates underneathYour images & fabric

Your VM images

Upload your own images and deliver the software environment your customers expect.

Your InfiniBand partitions

Partition the fabric yourself and control how your customers’ workloads share the cluster.

Your hardware requirements

Go deeper than a standard cloud instance. Design around the hardware and access your platform needs.

Talk to us
Independent by design

A fair market for every cluster.

SFC doesn’t compete with its customers. We operate clusters on behalf of partners and treat incoming orders fairly between them.

Talk to us

Tell us what you want to build.

A new cluster, an existing asset, or a capacity plan for your portfolio. Start with a conversation.

Talk to us
Point Mugu, CA