Product catalog

AI hardware, mapped to the work.

Compare local AI systems and turnkey private AI deployments using published specifications, dated pricing notes, and an explicit distinction between bare hardware listings and approved partner offers.

Quick comparison

Start with the configuration, not the headline.

Prices, availability, and specifications can change. Use this table to choose what to verify in a current vendor quote.

Selected products from the AI Compute Scout catalog
Specification	DGX Spark	DGX Station	DGX B200	ThinkStation PGX	Ascent GX10	Dell Pro Max with GB10
Manufacturer	NVIDIA	NVIDIA	NVIDIA	Lenovo	ASUS	Dell
Product family	GB10 Grace Blackwell	GB300 / Blackwell Ultra reference platform	Blackwell DGX	GB10 Grace Blackwell	GB10 Grace Blackwell	GB10 Grace Blackwell
Processor	20-core Arm CPU in GB10 Grace Blackwell Superchip	Grace CPU; exact configuration varies	2x Intel Xeon Platinum 8570	20-core Arm CPU	20-core Arm CPU in GB10 Superchip	20-core Arm CPU in GB10 Superchip
GPU	NVIDIA Blackwell GPU	Blackwell Ultra GPU; exact configuration varies	8x NVIDIA Blackwell GPU	NVIDIA GB10 Grace Blackwell GPU	NVIDIA GB10 Grace Blackwell GPU	NVIDIA GB10 Grace Blackwell GPU
Memory	64 GB or 128 GB coherent unified memory; 64 GB configurations are scheduled for 2026-10-23 and will be available only through participating OEM partners	748 GB coherent memory (252 GB HBM3e GPU + 496 GB LPDDR5X CPU)	1,440 GB total GPU memory; 2–4 TB system memory	128 GB unified memory	128 GB unified memory	128 GB LPDDR5X unified memory
Storage	Up to 4 TB NVMe M.2; the current 128 GB NVIDIA Marketplace US configuration lists 4 TB, while OEM configurations vary	4x M.2 Gen 5 slots	2x 1.9 TB NVMe M.2 OS; 8x 3.84 TB NVMe U.2 data cache	1 TB or 4 TB self-encrypting NVMe M.2, depending on SKU	1 TB, 2 TB, or 4 TB NVMe depending on SKU	1 TB, 2 TB, or selected 4 TB M.2 NVMe configurations
Operating system	NVIDIA DGX OS	Current reference configuration: Ubuntu with NVIDIA AI Developer Tools; Windows edition announced for Q4 2026	NVIDIA DGX OS / Ubuntu-based validated stack	NVIDIA DGX OS / Ubuntu Linux Pro with NVIDIA Base OS; verify SKU	ASUS pages label the supported image as Ubuntu Linux / NVIDIA DGX OS (Ubuntu-based); verify SKU and release	NVIDIA DGX OS 7
Listed price	$6,950.00 (observed 2026-10-06)	Quote dependent	Quote dependent	Quote dependent	Quote dependent	Quote dependent
Link	Affiliate partner offer ↗	Affiliate partner offer ↗	Affiliate partner offer ↗	Official listing ↗	Official listing ↗	Official listing ↗

The table includes the core local AI systems currently tracked by this publication. Aradia's three published turnkey tiers are shown on the Aradia private AI systems comparison. A partner link appears only when an approved affiliate URL exists; otherwise the link goes to the manufacturer's official listing.

Full catalog

Systems worth investigating.

Every card keeps unknown values explicit and links to a current official source or an approved partner destination. Aradia cards show the dated turnkey price and service scope separately from bare hardware pricing. In plain English, turnkey means paying for the first setup and validation sprint as well as the machine.

NVIDIAPersonal AI Supercomputer

DGX Spark

Local LLM inference, development, selected fine-tuning

Memory: 64 GB or 128 GB coherent unified memory; 64 GB configurations are scheduled for 2026-10-23 and will be available only through participating OEM partners
Storage: Up to 4 TB NVMe M.2; the current 128 GB NVIDIA Marketplace US configuration lists 4 TB, while OEM configurations vary
Price: $6,950.00 (observed 2026-10-06)
Aradia turnkey: $15,125.00

What turnkey means Hardware plus the first deployment sprint: the agreed setup, configuration, and validation path—not a bare-hardware checkout.

Published scope 30-day estimated deployment; Hardware, staging, model configuration, agent compilation, benchmarking, INT4-AutoRound, configured vLLM continuous batching, hardened Linux + Docker, a zero-inbound network posture by default, and a 14-day re-staging guarantee when configuration drift is detected. $1,500.00/mo optional. Checked 2026-10-06.

Availability The 128 GB / 4 TB NVIDIA Marketplace US listing was out of stock as of 2026-10-06; 64 GB OEM configurations are announced for 2026-10-23, with model and region dependent availability

Affiliate partner offer: Explore Aradia turnkey DGX Spark deployment

NVIDIAAI Workstation

DGX Station

Shared local development and inference

Memory: 748 GB coherent memory (252 GB HBM3e GPU + 496 GB LPDDR5X CPU)
Storage: 4x M.2 Gen 5 slots
Price: Quote or configuration dependent
Aradia turnkey: $194,093.00

What turnkey means Hardware plus the first deployment sprint: the agreed setup, configuration, and validation path—not a bare-hardware checkout.

Published scope 60-day estimated deployment; Hardware, staging, model configuration, agent compilation, benchmarking, AWQ-INT4, configured vLLM continuous batching, hardened Linux + Docker, a vendor-described zero-inbound network posture by default, priority SLA, and a 14-day re-staging guarantee when configuration drift is detected. $5,000.00/mo optional. Checked 2026-10-06.

Availability Contact NVIDIA or an OEM partner; region and edition dependent

Affiliate partner offer: Explore Aradia turnkey DGX Station deployment

NVIDIAData Center AI System

DGX B200

Enterprise multi-GPU training and high-throughput inference

Memory: 1,440 GB total GPU memory; 2–4 TB system memory
Storage: 2x 1.9 TB NVMe M.2 OS; 8x 3.84 TB NVMe U.2 data cache
Price: Quote or configuration dependent
Aradia turnkey: $505,500.00

What turnkey means Hardware plus the first deployment sprint: the agreed setup, configuration, and validation path—not a bare-hardware checkout.

Published scope 90-day estimated deployment; Hardware, 48-hour staging burn-in, benchmarking, FP8/INT4 hybrid quantization, multi-GPU vLLM continuous batching via NVLink, hardened Linux + Docker, a vendor-described zero-inbound network posture by default, and a 14-day re-staging guarantee when configuration drift is detected. $12,500.00/mo optional. Checked 2026-10-06.

Availability Quote and OEM dependent

Affiliate partner offer: Explore Aradia turnkey DGX B200 deployment

LenovoPersonal AI Supercomputer

ThinkStation PGX

CUDA local inference and development

Memory: 128 GB unified memory
Storage: 1 TB or 4 TB self-encrypting NVMe M.2, depending on SKU
Price: Quote or configuration dependent

Availability Available to order; configuration and region dependent

View official listing

ASUSPersonal AI Supercomputer

Ascent GX10

Local LLM inference and AI development

Memory: 128 GB unified memory
Storage: 1 TB, 2 TB, or 4 TB NVMe depending on SKU
Price: Quote or configuration dependent

Availability Verify current SKU, configuration, channel, and stock

View official listing

DellPersonal AI Supercomputer

Dell Pro Max with GB10

Local AI development and inference

Memory: 128 GB LPDDR5X unified memory
Storage: 1 TB, 2 TB, or selected 4 TB M.2 NVMe configurations
Price: Quote or configuration dependent

Availability Listed for purchase in the US; configuration and region dependent

View official listing

HPPersonal AI Supercomputer

ZGX Nano AI Station

Local LLM development and inference

Memory: 128 GB unified memory
Storage: Official US materials conflict: the product page, current QuickSpecs, and store product tiles show 2 TB or 4 TB; store explanatory copy still says 1 TB or 4 TB self-encrypted NVMe M.2 options
Price: Quote or configuration dependent

Availability Region and configuration dependent

View official listing

AcerPersonal AI Supercomputer

Veriton GN100 AI Mini Workstation

Local AI development and inference

Memory: 64 GB or 128 GB unified memory depending on region and SKU; checked UK part DT.R6LEH.002 lists 128 GB
Storage: Up to 4 TB NVMe M.2 by family; checked UK part DT.R6LEH.002 lists 4 TB PCIe 4.0
Price: Quote or configuration dependent

Availability Region and configuration dependent

View official listing

Affiliate disclosure

Some NVIDIA products in this catalog use a labeled Aradia affiliate partner link. Eligible purchases may compensate AI Compute Scout. Unapproved merchants remain official links only. Read the full disclosure.