Australian sovereign neocloud · Power to silicon to tokens

GPU capacity,
on time, at scale.

A sovereign Australian AI cloud, built beside the renewable generation that powers it. One operator from the substation to the API.

We hold the renewable generation beside each campus, build the hall to NVIDIA's reference design, fill it with GB300 NVL72 racks and run every node ourselves. One contract covers the building, the machines, the compute and the platform.

Australian owned, built and run here · First phase in design · Capacity contracted before construction
25MW IT
Per block
One block, one building, then a campus. Each block built to NVIDIA's reference design, liquid to chip.
72GPUs
Per rack
GB300 NVL72 in the first phase, then Vera Rubin as it arrives. Whole racks, single tenant.
100%
Australian
Owned, operated and resident. Engineered for IRAP assessment at OFFICIAL: Sensitive and PROTECTED.
5yrs
Wholesale commitment
A rate agreed before signature and held for the term, with the power behind the meter.
The AI factory

Four products. One operator. One site.

Take one layer or all four, on one contract. The company that builds the factory is the company that runs it.

01 · The building Physical layer

Powered shell beside the generation

Behind the meter beside wind or solar we hold, firmed by battery storage. 25 MW IT per block.

02 · The machines Compute layer

Whole racks, single tenant

GB300 NVL72 in the first phase, then Vera Rubin as it arrives, at a rate fixed for the term.

03 · The compute Service layer

Three tiers, one of them sovereign

Time on the same machines by the hour, on a committed term, or on the sovereign tier.

04 · The platform Managed layer

Managed services on top

Everything between a bare GPU and a model in production, run by us, with nothing to pay to move your data out.

01 · The building

Powered shell beside the generation, behind the meter.

We build each factory in Australia beside wind or solar generation we hold, firmed by battery storage, with a large grid connection brought in over time. You take the hall. Power supply is our problem.

25 MW IT per blockOne block, one building, then a campus
Liquid to chipBuilt to the NVIDIA reference design
Go-live dateAgreed before you commit
Battery firmedGrid connection brought in over time
02 · The machines

Whole racks, single tenant, at a rate fixed for the term.

GB300 NVL72 in the first phase, then Vera Rubin as it arrives. The power sits behind the meter with the machines, so the cost behind your rate is known before you sign.

72 Blackwell Ultra per rackOne rack, one workload
Vera Rubin NVL72Taken as it arrives
Five-year commitmentThe wholesale term
Rate held for the termSet before signature
03 · The compute

Time on the same machines, in three tiers, one of them sovereign.

You buy the capacity, watch what you consume and pay for it in one place. Move between the tiers as your demand settles, with one supplier.

StandardBy the hour, on demand
ReservedCommitted term, priority on the queue
SovereignAustralian-owned and operated, engineered for IRAP assessment at OFFICIAL: Sensitive and PROTECTED
Egress freeLeaving costs you nothing
04 · The platform

Managed services on top, and nothing to pay to move your data out.

We operate everything between a bare GPU and a model in production, on 10G and 100G links, with a dedicated government connection on the sovereign tier.

Storage50 PB, fast tier and capacity tier
InferenceManaged Kubernetes, kept up by us
Retrieval, streaming, databasesYour own documents, managed Kafka, managed SQL searchable by meaning
Data preparationManaged Spark
Power

Most operators wait on a grid queue. We start behind the meter.

Electricity loses value with every kilometre it travels. Tokens move anywhere in the world in an instant. Ekim AI exists to convert power into tokens, so we take the factory to the power.

Your rateAgreed before signature and fixed for the term
GenerationWind or solar behind the meter, firmed by battery storage, with a large grid connection brought in over time
ControlThe generation is held by us, owned outright or in joint venture with the operator that owns it
Go-liveThe date your capacity goes live, agreed before you commit
OperationsThe company that built the hall is the company that runs it
Scale25 MW IT per block: one block, one building, then a campus
Reserve capacity
Sovereignty

Australian owned, Australian operated, Australian resident.

01

Jurisdiction

Australian-owned and Australian-incorporated, inside a Five Eyes jurisdiction, with the ownership chain ending here.

02

Assessment

Engineered for assessment under IRAP, the Commonwealth security programme, at OFFICIAL: Sensitive and PROTECTED.

03

Operations

Australian staff operate every node. One operator runs the factory it built, on the ground, in this time zone.

04

Residency

Your data stays in Australia, and egress is free, so leaving costs you nothing.

05

Evidence

Residency, operating jurisdiction and the assessment position are written into the contract and evidenced on request, in the form your own auditor asks for.

06

Continuity

Ownership, operations and control of the power sit with one Australian counterparty, which is what makes the position hold for the length of the contract.

Operated

You never touch a node.

Standing up a GPU estate of any size means a network team, a cooling contract, a firmware programme and a support roster that covers three in the morning. Taking the capacity operated moves all of it onto our payroll and inside the service level you sign. Before signature you see it work.

Stage What you see before you sign
Proof of concept A short run on the real hardware, with your workload and your data, measured against the criteria you set rather than the ones we would choose.
Fault injection Your operations team injects the failures. We show what the platform does, what we do, and how long each takes.
Escalation A live escalation to third-line support, demonstrated on the call, with the engineer who would take it in production.
Service levels Availability, the credit schedule and the escalation path are negotiated before signature and written into the contract.
Handover Capacity arrives built, tested and integrated. From that point the operating burden is ours for the length of the term.
The bench

Built by the leading names in AI infrastructure.

NVIDIA's reference architecture makes the end-of-life refresh a rack swap: one hall, ten years, three generations of silicon.

NVIDIA Dell Supermicro Cisco VAST Data Vertiv WWT
Partner What they bring
NVIDIAThe GPU systems. GB300 NVL72 in the first phase, then Vera Rubin as it arrives, on NVIDIA's own reference design.
Dell TechnologiesServer platforms and rack-scale integration, so the machines arrive built and tested.
SupermicroLiquid-cooled server platforms, held as a second source on the compute line so no single supplier sets our delivery date.
CiscoNetworking inside the factory and the links out of it, at 10G and 100G, with a dedicated government connection on the sovereign tier.
VAST DataThe storage platform. Training data and models held inside the factory, delivered to the GPUs at the rate they draw.
VertivPower distribution and liquid cooling, on the prefabricated OneCore platform the 25 MW IT block is built on.
World Wide TechnologyGlobal-scale integration and operations. Design validation, build, integration testing and handover, on the same footing they deliver for the world's largest operators.
The commitment

Four ways to contract the same capacity.

We contract the capacity before construction starts, on a term and a rate agreed before signature.

Shape Who it suits Term What it fixes
Sovereign Reserved A workload under an Australian residency obligation or a security classification. 5 years Reserved capacity on the sovereign tier, with the assessment and residency position written in.
Wholesale Capacity You sell compute under your own brand and need Australian capacity from an operator who stays out of your market. 5 years Whole racks, single tenant, at a rate agreed before signature and held for the term.
Enterprise Committed Production load with a residency requirement and a consumption pattern that moves month to month. 3 years Reserved capacity with consumption on top of it, on a single bill and a single supplier.
Premium Inference API Inference in production, where the cost of a token sets the gross margin of the product. 2 to 3 years A rate per million tokens, served in region, on an OpenAI-compatible interface.
Price mechanismAgreed before signature and held for the term. Where an index applies, the index is stated in the contract.
Minimum termTwo years is the shortest commitment we contract.
GrowthReserved capacity you can grow into, with the platform services attaching to the same contract.
DetailRates, service levels and the full specification follow once an NDA is in place.
Standards

What is true of every factory we build.

Standard What it means for you
01 · One operator, end to endThe company that builds the factory is the company that runs it, so there is one number to call and one party accountable.
02 · Hold the generationThe power sits behind the meter, owned outright or in joint venture with the operator, so your rate holds however the market moves.
03 · Build to the reference designEvery hall is built to NVIDIA's published architecture, so the next generation of silicon drops into the same power and cooling.
04 · Single tenant where it countsWholesale racks are yours alone, with the network and the cooling sized for one workload.
05 · Sovereign by constructionOwnership, operations and residency are all in country from the first drawing.
Reserve capacity

Reserve capacity in the first phase.

Power decides what gets built. We control ours, owned outright or with the operator, so we can commit to a date. The contract fixes that date, a rate that holds for every year of the term, and an Australian position that survives a change of shareholder.

Tell us the power you need, the workload you plan to run on it and the date you need it live. We come back with what is available, the contract shape that fits, and the full detail once an NDA is in place.

Your request goes straight to our team. We reply within two business days.

Australian sovereign neocloud. GPU capacity, on time, at scale, with the power behind the meter. Reserve capacity