Pooled Unified Memory
Aggregate up to 2.0 TB of memory across nodes to host massive local LLMs without cloud egress fees or external API exposure.
Scale On-Device AI on Apple Silicon.
Zero Cloud Fees.
Turn high-density Mac mini and Mac Studio hardware into a managed, enterprise-ready rack platform equipped with 10 GbE networking, Thunderbolt 5 RDMA, and KVM control.
14 / 14 Nodes Active
Power Draw: 1,680 W| Live
2.0 TB
Maximum Pooled Unified Memory
320 Cores
Apple GPU Core Array
10 GbE
High-Throughput Network Backplane
80 Gbps
High Speed, Low Latency Thunderbolt 5
Single 13A Socket
Standard Power Draw, Air-Cooled & Quiet (<35dB)
Battery Powered
Runs On 1800W Portable Power Station
Hardware engineered for the rack.
Enterprise 10 GbE Networking
Integrated high-bandwidth network switching providing fast, redundant local network communication.
Thunderbolt 5 Interconnect
High-speed point-to-point interconnect option leveraging Apple Remote Direct Memory Access (RDMA) for distributed local AI inference across nodes.
Sharding
Split a single Large Language Model (LLM) or neural network across multiple Apple Silicon.
Remote Power Control: Live Demo
Hardwired remote power button capabilities and KVM integrations for complete lights-out cluster management. Remotely controlled actuator arms physically press the power button on Mac Mini and Mac Studio.
Turn the sound on to hear the actuator stroke and Mac startup chime.
3U vs 5U drawer
Custom-engineered modular drawers built specifically for rapid access, maintenance, and high-density mounting of Mac Mini and Mac Studio. Compare the Mac mini and Mac Studio drawer assemblies, slide travel, airflow and mounting dimensions.
6-node Mac mini drawer
High-density drawer carrying six Mac mini in 3U, with per-node power toggles and full-travel ball-bearing slides for front-of-rack service.
- Capacity
- 6 × Mac mini
- Rack space
- 3U · 129.5 mm
- Width / depth
- 482.6 mm / 308.0 mm
- DRAWER TRAVEL
- 213 mm
- Drawer weight
- 4.3 kg
- Density
- 2 nodes per U
- One mechanical power toggle per node
- Rear brace kit for deep-rack stability
- Vented side walls for front-to-back airflow
All drawers mount to standard EIA 482.6 mm (19-inch standard) rails with 10-32 hardware.
Native device management with Jamf for Mac.
The Mac Hive ships Jamf for Mac-ready: Mac Auto Advance zero-touch provisioning, remote device commands and automated local model service deployment out of the box.
Mac Auto Advance
Zero-touch setup and skip all Setup Assistant with no human intervention.
Device Management
Jamf is the gold standard in Apple device management and security.
Automatically Deploy LLM
Deploy Large Language Model (LLM) runtimes with Jamf.
AI Governance
Coming soon…
Built for industries that can't send data out.
Private models on private data.
Private high-speed document processing, quantitative analysis, and secure internal LLM sandboxing.
- On-prem document and claims processing
- Quantitative research workloads
- Isolated per-desk model sandboxes
Let’s Connect
Chat with a specialist to evaluate your cluster requirements.