Gryf: The First-ever Suitcase-sized AI Supercomputer

Gryf: The First-ever Carry On Suitcase-sized AI Supercomputer

Gryf™ brings uncompromised AI inference power anywhere you can wheel it — fully independent, field-serviceable, and built for where cloud dependency is not an option.

Talk to an Expert

Compare Other Options

Gryf: The first Suitcase-sized AI Supercomputer co-designed by GigaIO and SourceCode

AMD EPYC™ Processors 16 or 64 cores per slot

2,500W Dual hot-swap power supplies

Up to 5PB storage Up to 1PB per sled, five sleds

What is GigaIO’s Gryf?

Gryf is the only portable edge AI platform capable of the full range of data processing needs, from on-device AI inferencing to machine learning and ISR analytics, all in a ruggedized form factor that travels anywhere you can wheel it.

Gryf was co-designed by GigaIO and SourceCode, marrying GigaIO’s deep expertise in interconnects and composable infrastructure solutions with SourceCode’s ruggedized system design and co-design manufacturing.

Gryf provides true datacenter-class computing at the far edge, delivering meaningful AI inferencing wherever data is created and where answers are actually needed. GPU / Accelerator, Compute, Storage, and Network sleds provide the ability to customize resources for each workload by simply swapping sleds.

Why Does Field Intelligence Still Wait on a Datacenter?

Prior to Gryf, there were two options for generating on-site results at the far edge. Data was sent to a datacenter for processing and analysis, requiring connectivity you might not have and time you couldn’t spare. Or data was limited to workstation-grade compute, generating field answers that were worse than the ones waiting back at base.

Now, Gryf provides server-class GPUs, petabyte-scale storage, and full AI inference, all running in a case you can wheel in yourself. Gryf puts meaningful AI inference power directly at the source of the data, allowing you to quickly transform the vast amounts of sensor data collected at the edge into immediately actionable solutions.

AI Inferencing Wherever You Need It

Portability & Mobility

The compute can move wherever the mission requires.

Take your data and processing with you wherever you go
Bring datacenter-class compute to places with poor or expensive connectivity
Travel in challenging environments with a ruggedized case

Field Configurability

No new hardware purchases, no re-engineering, just rearrange sleds.

Six different sleds for GPU, compute, storage, and networking
Share storage or GPUs across multiple server sleds or VMs
Reconfigure resource allocation in real time, without going offline

Field Serviceability

Dramatically improve operational availability without on-site support.

If something fails, simply swap sleds in minutes
Completely self-service, no technical expertise necessary
No downtime, even if technicians are days or weeks away

No Vendor Lock-in

Never be locked into a single vendor’s roadmap, pricing, or stack.

Open platform built for any accelerator (GPUs, ASICs, inference cards)
Runs Linux, Windows, KVM, and VMware out of the box
Upgrade individual sleds without replacing the entire system

No Data Value Decay

Act on data the moment it’s captured, before the insight loses value.

Get real-time insights, not perishable insights
Process ISR, simulation, or inference workloads on-site
No round trip to a datacenter before your data is relevant

Scalability

One Gryf not enough? Stack your Gryfs into a larger single system.

Stack up to five Gryf units for a total of 30 total sleds
Any server can access any resource across the fabric shared by all five Gryfs
Runs on GigaIO’s patented PCIe-based open fabric technology

Explore the Sleds:

Compute Sled

Network Sled

Storage Sled

GPU / Accelerator Sled

Where Gryf Goes to Work

Defense & Intelligence

Learn More

Energy and
Oil & Gas

Learn More

Media & Broadcasting

Learn More

Sports Analytics

Learn More

Mobile Fly-away Kit

Gyrf,: AI Supercomputer suitcase-sized form factor

Gryf: AI Supercomputer. View of customizable sleds

Gryf Sample AI/ML Configuration

Stack Your Gryfs

This diagram shows five interconnected Gryfs connected by FabreX, GigaIO’s patented PCIe-based open fabric technology that allows any server to access any other server or resource device within the fabric. Up to five Gryfs can be stacked, and each can be reconfigured in real time to meet mission requirements.

Deploy multi-Gryf arrays for even more compute power, all at the edge!

Modular Edge Platforms for All Settings

GigaIO offers Edge solutions in two form factors: Manticore for stationary compute at the near edge, and Gryf for portable, field-ready operation and far-edge locations. Both share the same architecture and are powered by FabreX, GigaIO’s patented PCIe-based open fabric technology.

Put Manticore in your factory’s server closet. Bring Gryf along with you to your oil rig. Both run the same demanding AI workloads at full performance — datacenter-class performance without needing datacenter infrastructure.

Gryf is Best For:	Manticore is Best For:
Far Edge	Near Edge
Mobile Situations	Fixed Locations
Smaller Spaces	Larger Spaces
Ruggedized Situations	Stable Locations
Several GPUs	Up to 16 GPUs
More Flexible Solution	Lower-cost Solution

Gryf FAQs

What is Gryf used for?

Gryf is a portable, self-contained AI inference platform built to run server-class compute anywhere you can carry it. It’s used for on-device AI inferencing, machine learning, and ISR analytics in defense, energy, media, and sports environments where the mission moves and the cloud can’t follow.
Who is Gryf a good fit for?

Gryf fits teams that need datacenter-class AI inference in the field, not back at base. That includes forward defense operations, remote industrial sites, mobile broadcast crews, and any team where connectivity is limited, costly, or denied, but the mission still needs real-time answers.
Does Gryf require an internet or cloud connection to run?
How portable is Gryf, really?
Which GPUs and accelerators does Gryf support?

Gryf supports up to five GPUs per unit, including NVIDIA L40S and H100 NVL, and is built on an open platform that accepts GPUs, ASICs, and inference cards from multiple vendors. Sleds are field-swappable, so hardware upgrades don’t require replacing the system.
What operating systems and hypervisors does Gryf support?
Can I reconfigure Gryf after I’ve deployed it?
How is Gryf different from Manticore?

Gryf is portable and deploys to the field in a transportable, carry-on form factor; Manticore is a fixed-site half-rack that brings heavier, stationary AI compute to a permanent edge location. They share the same architecture and software but solve edge compute for two different deployment profiles, mobile versus fixed.