100%
Customer data on Australian soil
N+N
Redundancy at every layer
Hidden
Facility-location posture
Infrastructure
Australian sovereign cloud infrastructure.
Icarus owns and operates an Australian datacenter footprint engineered for three constraints at once: enterprise-grade reliability, regulator-acceptable physical security, and carbon-negative operation. Solar-primary, distributed-by-design, hidden-facility. Locations are intentionally not public.
Network architecture
Peering, edge, and AZ-pair topology.
Sydney AZ-1 active. Cross-region peering and Cloudflare edge form the path between availability zones and the public internet.
Australian datacenter footprint
Sovereign — Australian-owned, Australian-operated, Australian-hosted.
Every byte of customer data sits inside an Australian facility unless the customer explicitly elects otherwise (and at launch, no such election is offered — sovereignty is the default).
We disclose footprint scope to authenticated accounts and to customers under a security review. We do not publish facility addresses, regional names, or operator photographs. The hidden-facilities posture is a deliberate compliance and resilience choice — locations that are not public are harder to social-engineer, harder to physically target, and easier to defend in a regulator audit.
The footprint is sized for our launch cohort and engineered for incremental capacity addition without compromising the climate-negative architecture. Detailed capacity, latency, and regional information is shared under NDA on request — see Talk to engineering.
N+N redundancy architecture
Fault-tolerant by design at every layer.
The result is an SLA-defensible platform — uptime target 99.99%, automated failover, and a public status page reporting against it.
Power
Redundant solar generation paths feed each rack, with grid as a secondary source and battery storage providing peak-shifting and short-cycle hold-up. Loss of any single source is invisible to the workload.
Network
Dual fabric paths per rack, dual upstream peering, and automated path selection. No single fibre cut, switch failure, or peering session collapse takes a customer's network plane down.
Hypervisor and platform
Apache CloudStack-managed compute with live migration capability and N+N hypervisor redundancy. A failed host moves its workloads automatically.
Cooling and physical plant
N+N cooling, with environmental monitoring at the rack and aisle level.
Network architecture
Designed for low-latency Australian access first.
Deliberate peering choices keep traffic on-shore where possible.
Peering
Multi-provider peering at major Australian internet exchanges. Egress paths are diversified to avoid single-carrier concentration risk.
Latency
Optimised for sub-region latency between availability zones; figures shared under NDA on request.
Availability zones
Multiple AZs within the footprint, isolated power and network domains, customer-configurable workload distribution.
Power architecture — solar primary, grid secondary
The energy story is the foundation of the climate-negative claim.
Solar generation is the primary power source — sized to feed rack load directly during daylight hours. Grid acts as a secondary source when solar generation is below load, and battery storage covers the gap between solar peak and load peak (peak-shifting), as well as short-cycle hold-up during source transitions.
This stack matters for two reasons. First, it lowers the embodied carbon per teraflop-hour without buying a single offset. Second, the offsets we then purchase sit on top of a low-emission base — that's the architectural lever that lets the network claim negative, not neutral. Full energy and offset accounting on the methodology page.
The three layers
- 01 Solar primary. Sized to feed rack load directly during daylight hours.
- 02 Battery storage. Peak-shifting between solar peak and load peak; short-cycle hold-up during source transitions.
- 03 Grid secondary. Conventional grid as fallback when solar generation is below load.
Physical security — what hidden means
"Hidden facilities" is a compliance posture, not a marketing flourish.
- Facility addresses, exterior photographs, and regional names are not published.
- Vendor and supplier relationships are scoped on a need-to-know basis.
- Physical access is biometric, audited, and limited to a vetted access list.
- The audit trail is built to satisfy ISO 27001 controls, SOC 2 Type II, and APRA CPS 234-style supervisor review.
For compliance buyers, the hidden-facility posture is the difference between a vendor a CISO can sign off and a vendor a CISO has to defend.
From rack to climate-negative
A short, deliberate path.
Solar generation feeds the rack. The hardware draws less per unit of compute because it is distributed-by-design rather than stacked-for-redundancy in a single hyperscale building. The verified offset stack sits above the already-reduced base, and the result lands in negative territory — methodology and verification roadmap on /methodology.
Architecture under the hood
Detailed footprint, latency, and regional info — under NDA, on request.
Most teams should just create an account. The detailed brief exists for the cases where self-service isn't enough on its own.