What is the Bid Pack?

Send us the RFP and in ten working days you receive a complete AI factory architecture package aligned to NVIDIA's reference architectures, and you keep the compiler that produced it. The package includes the architecture model, topology diagrams, a logical bill of materials, port and connectivity budgets, the power and rack envelope, a requirements traceability matrix, an architecture decision record and an evidence manifest.

Can I see it running before I talk to anyone?

Yes. We host a demo instance and give you access, with nothing to install and no IT ticket. You put the parameters in and walk a minimum of 48 AI factory designs yourself by varying the inputs. The demo covers one GPU platform, the B300; the licensed build covers the full range. It is a throwaway environment, so do not put a live customer RFP into it.

How much does it cost?

There are three ways in. A design review is 5,000 US dollars and returns a scored compliance report in three working days. A perpetual licence to the compiler alone is 50,000 US dollars. The Bid Pack, where we do the work with you, is 75,000 US dollars, and that price rises with every Bid Pack delivered.

What is the guarantee?

Miss ten working days and you pay nothing. You keep the package and compiler licence. We fix missing evidence and work disqualifying gaps found by your reviewer at no charge.

NeoCloud and GPU cloud operators planning roughly 500 to 10,000 GPUs with repeated customer designs, a live RFP and a small architecture team. Also systems integrators delivering five or more substantial designs a year.

Is Neon a data centre design tool?

No. Neon designs the AI factory's IT infrastructure and calculates its facility requirements. Building design, power distribution, cooling plants and BIM remain with specialist engineers.

Is the architecture generated by AI?

No. The architecture is produced by a deterministic compiler, not a language model. It applies design rules and records the evidence behind decisions. Outputs still require review against the selected architecture; the worked example still needs detailed engineering and vendor validation.

AI factory architecture compiler, aligned to NVIDIA's reference architectures

Know exactly what you must deliver.

Your next AI factory bid has a deadline. Your architect already has a backlog. Start with a $5,000 design review: identify the evidence, gaps and decisions before they become customer commitments.

Explore the $5,000 design reviewThree working days from the agreed start date. One decision meeting.

NVIDIA reference architectures
Deterministic compiler
Ten working days
You keep the compiler

See the work

See the product. Inspect the output.

Start with 30 seconds.

Watch the short introduction, then follow the longer workbench recording when you want the detail.

Watch the 30-second introduction ↗

A 1,024-GPU worked example.

Inspect the eight-page proposal, 16 security findings and 12 delivery gates. The Asteria pack shows both the output and the engineering work still open.

See the example and its conditions →

Request access to the hosted demo →

For operators with a bid in flight

You run five hundred to ten thousand GPUs. You have a live RFP, a customer waiting, and one solutions architect who is already the bottleneck. The bid is due in weeks and the design is the part that has not started.

Built for repeated customer designs. Also for systems integrators handling five or more substantial designs a year. Start with a $5,000 design review on the bid in front of you.

Illustrative review finding

Know the gap before you commit.

This illustrates the structure of a review finding. It is not a customer result or a completed engineering assessment.

Requirement: The customer needs 8 PiB of usable storage.
Evidence gap: The capacity calculation exists. Vendor confirmation of usable capacity, workload throughput and the orderable configuration is still missing.
Consequence: The bid could commit to capacity or performance that the proposed configuration cannot deliver.
Decision and owner: Assign the storage architect to obtain vendor sizing evidence. The bid owner must approve the scope or record the exception before committing.

Compare with the open decisions in Asteria →

Ways in

Start with the design review.

Establish the gaps for $5,000. If you need help producing the full package, move to the $75,000 Bid Pack. Choose the standalone licence when your team will do the work.

Design review

$5,000

Send us a design. Three working days later you have a scored compliance report naming every gap. Pro rated against whatever you buy next.

Licence only

$50,000

The compiler, perpetual. You do the work.

Bid Pack

$75,000

Send the RFP. Ten working days later you have the package and the compiler.

Before you buy

What you are buying. What needs agreement.

What does the review fee cover?

A $5,000 review, delivered in three working days from the agreed start date, with an evidence register, gaps and a decision meeting. The fee is pro rated against what you buy next. The next order records the credit; refund and repeat-review terms are agreed in writing.

What do we keep?

The $50,000 standalone licence is perpetual. The $75,000 Bid Pack includes the compiler licence and a package produced with you in ten working days. Knowledge-pack subscriptions and additional services are priced separately.

Where does our data go?

Before you share a confidential RFP, we agree processing location, permitted access, retention and deletion. The hosted demo is a disposable evaluation environment. Use sample inputs there.

What about deployment, packs, updates and support?

Your order names the deployment arrangement, included knowledge packs, update entitlement and support coverage. We confirm these against your environment before you buy. The public demo does not establish your production deployment terms.

Who will we work with?

Matt Hertel is your first contact and replies within two working days. The architecture team has designed compliant architectures for leading hyperscalers and model builders. We agree the reviewer and decision meeting as part of the engagement.

Does repeated bid work justify it?

Illustration only: five designs a year × ten architect-days saved per design × $1,000 per day = $50,000 of architecture capacity. Those are assumptions, not measured Neon savings. Replace them with your bid volume, effort and internal cost. Include pack subscriptions, support and implementation costs in your comparison.

The immediate review decision is narrower: is $5,000 worthwhile to identify the gaps and owners on this bid before you commit?

Discuss your design

The artefact

See what goes into an architecture package.

Start with an eight-page proposal, then follow the requirements, architecture and delivery plan. The Asteria worked example plans 1,024 HGX B300 GPUs with Spectrum-X and WEKA.

1,024planned populated GPUs

128planned compute nodes

6example PDFs

Storage, facilities and security need detailed engineering. This example records high-level architecture approval with conditions.

Read the example and design notes

The 20,736-GPU illustration used elsewhere describes a separate, larger design. It is not the downloadable example or evidence of an installed system.

Compiled, not written

The architecture is produced by a deterministic compiler, not by a language model. The same inputs produce the same design, every time, and every decision traces to the reference architecture rule behind it.

The same inputs produce the same design. Every time.
Ask a language model the same low level design question twice and you get two architectures. Ask a different model and you get a third. You do not want to risk a fifty million dollar contract on a hallucination.
Compare options against the selected design.
Review the selected architecture, compatible implementations and alternatives that require a new design. Each change needs its own evidence and validation before it belongs in a bid.
Alignment is what keeps your NVIDIA support entitlement.
Deviate on Spectrum-X, InfiniBand or certified storage, and when the GPUs underperform the answer you get is that you used the wrong kit. This is a reason to align that has nothing to do with winning the bid.
Maker and checker are separated.
An immutable candidate, a decision log and an evidence package, for when someone asks how you arrived at this.
The knowledge does not leave with your architect.
It is in the system.

Who built this

We have built for the leading GPU manufacturer in the world.

The people who built this have designed compliant architectures for leading hyperscalers and leading model builders.

We cannot name them. In GPU as a service, nobody can.

Delivery

Ten working days

Four deliveries. The gaps between them are to scale. Six months to the first requirement shows where bids wait.

Up to 10xfaster to submission

Our AI Architect's own figure, from doing this work by hand and then with a language model. It is a comparison of effort, not a promise about the outcome of your bid.

01
Requirements assessment and compliance matrix. Every requirement in your RFP extracted, traced and rated.
03
Two or three candidate architectures with the tradeoffs made explicit, plus the risk register and the open questions list.
07
The full package. Architecture model, topology diagrams, logical BOM, port and connectivity budget, power and rack envelope, requirements traceability matrix, decision record, evidence manifest.
10
Handover. The licensed compiler, your configuration already loaded, and a walkthrough so your architect can regenerate and re-cost it without us.

Day one and day three exist on purpose. Visible progress against a bid clock is worth as much as speed.

Our risk

Ten working days, or you pay nothing and keep the package

You cannot risk your bid clock on a vendor you have not used. So we carry that risk instead.

Miss the ten working days and you pay nothing.
You keep the architecture package we delivered, and you keep the compiler licence. There is no clause underneath that sentence.
Every decision traces to a rule and its evidence.
Any that does not, we fix free.
If your reviewer finds a disqualifying gap in what we delivered,
we work it until it closes, at no charge.

Request a call about your bid

Architecture for the IT stack.

Compute, fabrics, storage and the evidence behind the design. Neon calculates facility-facing requirements. Building, electrical, cooling and installation engineering remain with your specialists.

Read the scope and approval boundaries →

Start with the $5,000 design review

Start with the design you need to submit.

Share your deadline and the decision you need to make. Matt replies within two working days. We agree the scope and start date before you send confidential files.

Request the $5,000 review

3
working days: $5,000
design review: 1
decision meeting