Broader access to AI
Extend intelligence from centralized infrastructure to private environments and field devices, with the right form factor for every scale.
HOLYCORESCOMPUTE AT EVERY SCALE
From cloud to edge.
AETHER, NOVA, and SPARK share one open compute foundation.
WHY HOLYCORES
Bring advanced AI wherever it matters.
Useful compute lives beyond benchmarks. It must enter enterprises, cities, devices, and real workflows—creating a better balance across performance, economics, privacy, and sustainability.Closer to data and decisions
Reduce unnecessary data movement and run latency-sensitive models where results are needed.
Clearer data boundaries
Support local and private deployment for sensitive information, regulated industries, and critical operations.
Controllable system economics
Evaluate utilization, capacity, energy, and operations around real workloads—not isolated peak numbers.
More responsible compute
Use dataflow, hardware-software co-design, and layered deployment to reduce waste across the system.
ONE FOUNDATION / THREE SCALES
AETHERCloud scaleNOVADatacenter accelerationSPARKLocal edge intelligenceOne platform that expands with the workload.
HARDWARE PORTFOLIO
From development, to scaled deployment.
Real hardware makes product boundaries tangible. HOLYCORES carries silicon and software into installable, coolable, and interconnected form factors.
IN-HOUSE SILICON / AI COMPUTE
HELIOS HCX E1
Designed in-house, from the silicon up. A shared compute, memory, fabric and software foundation for datacenter, edge and advanced packaging.
- In-house programmable compute architecture
- Hierarchical on-chip memory and dataflow fabric
- HOLYFLOW hardware-software co-design
CLOUD / RACK-SCALE SYSTEM
AETHER A72
An integrated rack bringing compute, fabric, liquid cooling, telemetry and management into one production system.
- Up to 72-accelerator planning scale
- Rack-wide HC-Fabric
- Integrated liquid-cooling plan
DATACENTER / SINGLE ACCELERATOR
NOVA N100
A single-chip PCIe accelerator for developer workstations, private AI, and general inference servers.
- PCIe 5.0 ×16 host interface
- 24 GB target memory
- Dual-slot active cooling plan
DATACENTER / SCALE-OUT
NOVA N200
A dual-chip platform with dedicated card fabric for throughput, multi-card scaling, and dense servers.
- Dual-chip compute board
- 48 GB target memory
- Active and passive variants
EDGE / EMBEDDED MODULE
SPARK S20
A compact, low-power local inference module for robotics, vision, and intelligent devices.
- Compact M.2 form factor
- Local model and sensor path
- 15 W-class power target
COMPARE NOVA MODELS
One architecture. Different system boundaries.
The following targets support product planning and joint validation. They are not final production or certified specifications.| Planning dimension | NOVA N100A | NOVA N200A | NOVA N200P |
|---|---|---|---|
| Form factor | Single-chip active card | Dual-chip active card | Dual-chip passive card |
| Relative compute scale | 1× baseline | Approx. 2× N100A | Approx. 2× N100A |
| Target memory | 24 GB | 48 GB | 48 GB |
| Host interface | PCIe 5.0 ×16 | PCIe 5.0 ×16 | PCIe 5.0 ×16 |
| Card fabric | Host PCIe | 2× HC-Fabric | 2× HC-Fabric |
| Target board power | Up to 225 W | Up to 350 W | Up to 300 W |
| Cooling | Dual-slot active | 2.5-slot active | Dual-slot passive |
| Target deployment | Developer workstations, inference servers | High-throughput inference, multi-card nodes | High-density rack servers |
Planning targets may change following silicon, board, thermal, and system validation. Final availability, compatibility, and specifications will be defined in future product materials.
One product path from cloud to edge
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.AETHER / cluster-scale systems
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
NOVA / datacenter acceleration
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
SPARK / real-time edge inference
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
WHAT THIS PAGE ANSWERS
Portfolio selection
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
01Portfolio selection
02System boundaries
03Adoption path
04Measured next steps
DEEPER VIEW
System boundaries
Portfolio selection
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
Explore →System boundaries
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
Explore →Adoption path
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
Explore →DECISION FRAME
Adoption path
Workload
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
Environment
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
Operations
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
Evidence
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
NEXT STEPS
Move from context to evidence.
Define
Create a clear output and decision gate for this stage.
Validate
Create a clear output and decision gate for this stage.
Operationalize
Create a clear output and decision gate for this stage.
PRODUCT EVALUATION
Bring a workload—not only a specification sheet.
Portfolio selection
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
System boundaries
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
Adoption path
Choose by deployment location, model scale, latency, power, and operations—not a single peak metric.
FAQ
Bring a workload—not only a specification sheet.
What should I do next?+
Share a non-confidential workload, environment, current baseline, and decision goal.
Are all capabilities available today?+
Availability, compatibility, and validated specifications will be published by product maturity.
01 / AETHER
Cloud AI systems built to scale
Production infrastructure combining dense compute, high-bandwidth fabric, and unified operations.
02 / NOVA
Open datacenter acceleration
A programmable platform for training, fine-tuning, and inference.
03 / SPARK
Intelligence at the edge
Low-power processors for robotics, vision, and real-time AI.
READY TO BUILD?