Skip to main content

Add the compute capability the work requires.

INFRASTRUCTURE / FUSION: Core FUSION: Core is a family of compute appliances for the Fusion Hardware Fabric. Models can combine CPU, GPU, NPU, FPGA, DSP, media, AI, and compression acceleration with local memory and cache, while FusionCompute places authorized Programmable Nodes without coupling applications to one machine or processor vendor.
Product overviewAll products

How it connects

Portfolio advantageShared identity, storage, intelligence, automation, and security capabilities connect FUSION: Core with FUSION: External Appliances, FUSION: Matrix, and FUSION: Vault without forcing them into one oversized application.
FUSION: Core, Matrix, and Vault modular hardware fabric appliance families

What FUSION: Core brings together.

WHAT IT ENABLES A family of topology-aware compute appliances adds authorized CPU, GPU, NPU, FPGA, DSP, media, AI, and compression capabilities where work needs them.

FUSION: Core Solo

01A compact entry point for adding focused acceleration, media, AI, compression, or programmable-node capacity to an individual workstation or edge environment.

FUSION: Core Pro

02A higher-performance multi-accelerator system for creators, developers, engineers, production workstations, and demanding local compute workflows.

FUSION: Core Business

03A serviceable business and studio appliance for shared compute, concurrent jobs, governed teams, local infrastructure, and rack deployment.

FUSION: Core Cloud

04A modular rack-scale compute system for private cloud, datacenter, managed capacity, distributed execution, and horizontally scaled provider pools.

FusionCompute coordination

05FusionCompute selects compatible authorized providers, verifies inputs and outputs, records provenance, and retries failed attempts without binding callers to a named appliance.

Heterogeneous acceleration

06Combine general-purpose and specialized processors according to the model, deployment, and workloads it is intended to serve.

Capability-first placement

07Applications request vision, inference, encoding, tensor compute, compression, or another capability; FusionRuntime selects a suitable local, Core, workstation, server, or cloud provider.

Topology-aware scheduling

08Health, trust, utilization, queue depth, data locality, bandwidth, latency, congestion, and workload partitionability remain visible to placement decisions.

Horizontal scale

09Add Core appliances for parallel execution while preserving hardware compatibility and avoiding misleading linear aggregation of unlike compute ratings.

Programmable Nodes

10Run bounded, portable workloads with explicit capability, memory, input, output, timeout, retry, integrity, and placement requirements.

Trusted execution

11Device identity, short-lived grants, signed updates, isolation, quotas, hashes, and execution provenance keep physical attachment separate from authorization.

One product experience. A platform underneath.

THE SHARED FOUNDATION

FUSION: Core

FOCUSED PRODUCT EXPERIENCE

Shared capabilities

IDENTITY · INTELLIGENCE · STORAGE · COMPRESSION · DEDUPLICATION · MEDIA · SECURITY

Fusion runtime

Fusionscript · FSML · FSS · CAPABILITY ABI · PROVIDERS

Infrastructure

LOCAL · CLOUD · WINDOWS · LINUX · CONNECTED SERVICES

Explore the connected Fusion portfolio.

Explore the complete Fusion portfolio across creation, operations, communications, development, delivery, storage, and protection.Explore every product