Maria's AI Hardware Company

Real NVIDIA hardware for running real open-source AI models — from a single workstation card to a full datacenter rack. Every number explained in plain words.

Help me choose the right build →

Hardware We Sell

Six products, from a $2,499 desktop card to a $3,000,000 datacenter rack. Every spec explained so you know what it means for your workload.

Consumer Desktop GPU

NVIDIA GeForce RTX 4090

Price $2,499

This is the entry point to serious AI hardware — roughly the price of a used car. It is the highest-end card you can buy off the shelf for a gaming or workstation PC.

GPU Memory (VRAM) 24 GB

24 GB of video memory means the card can hold AI model weights for models up to about 24 billion parameters. Useful for smaller open-source models, but far too small for the three models listed on this page.

Power Draw 450 W

450 W is about as much as four bright incandescent light bulbs running at once, or roughly 38% of what an average home uses continuously.

Memory Bandwidth 1,008 GB/s
Architecture Ada Lovelace

Best for: Small language models (up to ~24B parameters), image generation, video editing, game development.

Request a Quote
Professional Workstation GPU

NVIDIA RTX 6000 Ada Generation

Price $6,800

Professional tier — used by scientists, architects, and AI researchers who need reliability at a workstation desk. ECC memory automatically catches and corrects memory errors, which matters for long-running AI jobs where a single bad bit could silently corrupt results.

GPU Memory (VRAM) 48 GB

48 GB is double the RTX 4090's capacity. You can fit models with up to about 48 billion parameters in memory at once, making it suitable for medium-sized open-source models.

Power Draw 300 W

At 300 W, this professional card actually uses less power than the consumer RTX 4090 — NVIDIA optimized it for sustained all-day workstation use rather than short gaming bursts.

Memory Bandwidth 960 GB/s
Architecture Ada Lovelace

Best for: Medium language models (up to ~48B parameters), scientific computing, professional visualization.

Request a Quote
Datacenter AI Accelerator

NVIDIA H100 PCIe 80GB

Price $27,500

Entering datacenter territory: one card costs more than most cars. The price reflects purpose-built 24/7 reliability, error-correcting memory, and roughly 3× the AI throughput of consumer cards per watt.

GPU Memory (VRAM) 80 GB

80 GB of HBM2e (High-Bandwidth Memory) — more than triple the RTX 4090. When you connect several H100s together in a cluster, their memories pool together so a model can span multiple cards.

Power Draw 350 W

350 W is similar to the professional RTX 6000 Ada, but this card delivers far more AI performance per watt for the matrix calculations transformer models use.

Memory Bandwidth 2,000 GB/s
Architecture Hopper

Best for: Large language model inference, fine-tuning, building multi-GPU clusters for very large models.

Request a Quote
Datacenter AI Accelerator (NVLink)

NVIDIA H100 SXM5 80GB

Price $38,000

The SXM version costs more because the SXM socket lets multiple GPUs share data via NVLink — a special high-speed highway between cards — at 900 GB/s per GPU. When a model is spread across several cards, that speed cuts the time each card waits for data from the others.

GPU Memory (VRAM) 80 GB

Same 80 GB as the PCIe version, but with HBM3 memory that moves data 67% faster. When a model is split across many cards, this extra speed translates directly into faster responses.

Power Draw 700 W

700 W — double the PCIe version. The SXM form factor requires a special server chassis with high-power delivery and liquid-assisted cooling. It is not for a standard server rack; it lives inside the DGX H100 system below.

Memory Bandwidth 3,350 GB/s
Architecture Hopper

Best for: Large AI model training and inference where maximum inter-GPU bandwidth is critical — housed inside DGX H100 systems.

Request a Quote
Complete AI Server (8 × H100 SXM)

NVIDIA DGX H100

Price $350,000

A complete, ready-to-run AI server — eight H100 SXM GPUs, the chassis, power supplies, cooling, and networking all in one tested box. Roughly what a small house costs in some U.S. cities.

GPU Memory (VRAM) 640 GB

8 × 80 GB = 640 GB total. The eight GPUs are connected via NVLink at 900 GB/s each, so from the software's point of view they act as one large 640 GB memory pool — not eight separate cards.

Power Draw 10,200 W

10,200 W — about 8.5× what an entire average American home draws continuously. It needs a dedicated power circuit and datacenter-grade cooling; it cannot run in a normal office.

Memory Bandwidth 26,800 GB/s
Architecture Hopper (8 GPUs)

Best for: Running or fine-tuning large AI models up to ~640B parameters; on-premise AI infrastructure for serious organizations.

Request a Quote
Full Datacenter Rack (72 Blackwell GPUs)

NVIDIA GB200 NVL72 Rack

Price $3,000,000

Three million dollars for a single rack. This is what the world's leading AI labs and cloud providers buy — the current ceiling of publicly available AI hardware. Companies typically sign orders for dozens of racks at a time.

GPU Memory (VRAM) 13824 GB

72 Blackwell B200 GPUs × 192 GB each = 13,824 GB — nearly 14 terabytes of GPU memory. The largest AI models in the world fit here with room to spare; all three models on this page could run simultaneously.

Power Draw 120,000 W

120,000 W — 120 kilowatts, or about as much as 100 average homes running at once. A single rack requires its own dedicated power feed and liquid cooling infrastructure; it cannot share a building's normal electrical system.

Memory Bandwidth 1,300,000 GB/s
Architecture Blackwell (72 GPUs)

Best for: Frontier AI model training, massive-scale inference, datacenter AI infrastructure for cloud providers and top AI labs.

Request a Quote

Open-Source Models We Support

Three frontier open-source models, each with MIT license — free for commercial use. Below you'll see exactly how much hardware each one needs to run, with the memory math shown step by step.

What "parameters" means: A parameter is one learned number inside the AI model — like a dial that was tuned during training. More parameters means the model has "read" more and can handle more complex tasks. The catch: every parameter needs to live in GPU memory while the model runs.

What "active parameters" means: All three models below use a Mixture of Experts (MoE) architecture — only a fraction of parameters activate for each question. This makes inference more efficient, but all parameters still need to be loaded into GPU memory so the right experts are available on demand.

Memory formula used on this page: Total parameters (billions) × 1 GB + 20% safety buffer. This assumes INT8 quantization (1 byte per parameter), which is the standard way to run large models efficiently without too much quality loss.

MiMo V2.5 Pro

by Xiaomi  |  Mixture of Experts (MoE)  |  Released April 2026  |  MIT license

Xiaomi's flagship open model. Despite having 1.02 trillion total parameters, only 42 billion activate per inference pass — the MoE architecture routes each question to a small subset of 'expert' sub-networks, making it efficient to run while retaining the knowledge of a much larger model.

Parameter Counts

Total parameters 1,020 billion
Active per inference pass 42 billion (only these fire per question, but all must be loaded)
Context window 1 million tokens

Memory Calculation

Total parameters 1,020 billion
× 1 GB per billion parameters = 1,020 GB
+ 20% safety buffer + 204 GB
Minimum GPU memory needed 1,224 GB

Minimum Hardware to Run This Model

Option A — Minimum cost (H100 PCIe cluster)
Cards needed16 × H100 PCIe 80GB
Total GPU memory16 × 80 GB = 1280 GB
GPU cost only$440,000

Server nodes, networking, and installation add ~$50–100k on top of GPU cost.

Option B — Clean infrastructure (DGX H100 cluster)
Servers needed2 × DGX H100
Total GPU memory2 × 640 GB = 1280 GB
Total cost$700,000

Complete, tested servers — no extra chassis or networking to source separately.

Request a Quote

GLM 5.1

by Zhipu AI  |  Mixture of Experts (MoE)  |  Released 2026  |  MIT license

Zhipu AI's reasoning-focused open model, built on a 754B MoE architecture. Designed for complex reasoning and software engineering, it achieves 94.6% of Claude Opus 4.6's coding benchmark — a remarkable result for an open-source model trained without NVIDIA hardware.

Parameter Counts

Total parameters 754 billion
Active per inference pass 40 billion (only these fire per question, but all must be loaded)
Context window 200,000 tokens

Memory Calculation

Total parameters 754 billion
× 1 GB per billion parameters = 754 GB
+ 20% safety buffer + 151 GB
Minimum GPU memory needed 905 GB

Minimum Hardware to Run This Model

Option A — Minimum cost (H100 PCIe cluster)
Cards needed12 × H100 PCIe 80GB
Total GPU memory12 × 80 GB = 960 GB
GPU cost only$330,000

Server nodes, networking, and installation add ~$50–100k on top of GPU cost.

Option B — Clean infrastructure (DGX H100 cluster)
Servers needed2 × DGX H100
Total GPU memory2 × 640 GB = 1280 GB
Total cost$700,000

Complete, tested servers — no extra chassis or networking to source separately.

Request a Quote

DeepSeek V4 Pro

by DeepSeek  |  Mixture of Experts (MoE)  |  Released April 2026  |  MIT license

DeepSeek's most powerful open model. With 1.6 trillion total parameters and a 1-million-token context window, it rivals the best closed-source frontier models on most benchmarks — and is free to use commercially under the MIT license.

Parameter Counts

Total parameters 1,600 billion
Active per inference pass 49 billion (only these fire per question, but all must be loaded)
Context window 1 million tokens

Memory Calculation

Total parameters 1,600 billion
× 1 GB per billion parameters = 1,600 GB
+ 20% safety buffer + 320 GB
Minimum GPU memory needed 1,920 GB

Minimum Hardware to Run This Model

Option A — Minimum cost (H100 PCIe cluster)
Cards needed24 × H100 PCIe 80GB
Total GPU memory24 × 80 GB = 1920 GB
GPU cost only$660,000

Server nodes, networking, and installation add ~$50–100k on top of GPU cost.

Option B — Clean infrastructure (DGX H100 cluster)
Servers needed3 × DGX H100
Total GPU memory3 × 640 GB = 1920 GB
Total cost$1,050,000

Complete, tested servers — no extra chassis or networking to source separately.

Request a Quote