What “in development” means for Fabric

Walk past a building site and you often see a hoarding with drawings of the finished building. The drawings tell you what is planned. They do not tell you when it will open, or what it will be like inside.

When we describe Fabric, we label it "in development". This piece explains what Fabric is designed to do, what that label tells you, what it does not, and how we check our own words before they are published.

What Fabric is designed to do

Fabric is Starfield's control plane. A control plane is the part of a system that decides what happens where, rather than doing the work itself. (We describe the idea in AI workload scheduling: a control tower for AI work.)

Fabric is designed to schedule, meter and record every request across many sites. In plain terms, it is being built to do three jobs.

Schedule. Decide where each piece of AI work should run, following the rules set for it.

Meter. Measure what each piece of work used.

Record. Keep an account of what was agreed, where the work was placed and what was delivered. (See AI audit trail: a receipt for AI work.)

Those are design goals. We describe them so readers know what we are working toward, not to suggest they are finished.

What "in development" tells you

The label is a plain statement of status. It tells you three things.

First, the work is real and under way. We are building Fabric, not just describing an idea. Second, the design can still change as we learn. Third, it is not available for customers to use today.

That is why we write that Fabric is "designed to" do things, or that "we're building" it. Those phrases are chosen to match the status, not to soften it.

What it does not tell you

The label is just as useful for what it leaves out. Please do not read any of the following into it.

No date. We do not give release dates for products in development. A label is not a hidden deadline.

No performance numbers. We do not publish speed, capacity or reliability figures for Fabric. Any number you see attributed to it did not come from us.

No guaranteed features. The design goals describe intent. The final product may differ, and we will say so if it does.

No customer claims. The label does not imply that anyone is using Fabric today.

How we check our own words

Status labels only help if they are applied consistently. So we have a simple rule. Every technical line we publish is meant to be checked by our CTO against what is actually built.

The full rules are in What we will and won't claim here. The same habit applies to prices: a figure we publish should carry a date and a source, or not appear at all. (See LLM API pricing: why we publish dated prices.)

How to read us

If you are deciding whether to plan around Fabric, here is a fair way to read our status labels:

1. Treat "in development" as "being built, not available, no date".
2. Treat "designed to" as a description of intent, not a capability you can rely on yet.
3. If you need a firm answer for your own planning, ask us directly and ask for it in writing.

We would rather you plan around what exists than around what we hope to build.

What we are building

Fabric is one part of a wider design. Here is the whole of it, in the words we use everywhere.

Starfield deploys standardized compute close to power, connectivity and demand, and we're building one programmable network to connect it for AI inference. Workloads are designed to be placed by policy, on latency, data residency, model and load, and every workload is designed to generate an auditable record of commitment, placement and delivery.

When Fabric's status changes, the label will change with it, and we will say so here.

If you are planning AI workloads and want to talk about capacity, use Request capacity below.