Local AI on NVIDIA, built and managed by Redeco IT Group.
We advise, supply, implement and manage NVIDIA AI infrastructure. From a DGX Spark in your office to DGX B300 and H200 NVL servers in your server room or our data centre racks in the Netherlands. Your data stays yours.
No data sent to OpenAI, Anthropic or Google
We set up AI infrastructure on-premise or run it in our own data centre racks in the Netherlands. Your business data stays within your own walls, under your control and managed by us.
Your business data
- Documents, email and administration
- Customer and production data
- Prompts, knowledge sources and models
Runs on your infrastructure
- Your own server room or data centre
- Our data centre racks in the Netherlands
- Hardware, models and management by Redeco IT Group
No request leaves your environment or the Netherlands.
Independent advice
We choose hardware and models based on your use case, not a single supplier.
Data stays in the Netherlands
Prompts, documents and models never leave your own environment.
One partner
Advice, hardware, software and management from one partner, with one dedicated point of contact.
From advice to rack installation
- 1
Advice and sizing
We map your use case, users and data to determine which NVIDIA hardware and models fit.
- 2
NVIDIA hardware supply
As a NVIDIA Certified AI Advisor, we supply the full portfolio, from DGX Spark to DGX B300, networking and licences.
- 3
Installation and implementation on-site or in our data centre racks in the Netherlands
We install the hardware and set up the AI stack. In the data centre, our engineers provide data centre remote hands: racking, cabling and networking.
- 4
Management and monitoring
We keep hardware, models and software up to date, secure and monitored. One point of contact for everything.
Local AI from advice to everyday use
One partner for advice, custom software and complete AI infrastructure on-premise or in our own data centre racks in the Netherlands.
AI assessment & advice
Together, we identify where AI can save your business time, money or frustration.
Custom software
From prototype to production. We build AI applications that fit your processes precisely.
Local AI infrastructure
We set up complete AI stacks on-site or in our own data centre racks in the Netherlands. No dependency on major providers.
Managed by us
We manage the entire stack: hardware, models, integrations and updates. One partner for everything.
Training & adoption
Your team learns to use AI safely and effectively, with practical workshops and clear guidelines.
Secure & GDPR-compliant
Data, prompts and models meet your security and compliance requirements. European, on-premise.
NVIDIA DGX B300: for training and AI reasoning at scale
The DGX B300 is built for large-scale inference, reasoning models and training. We supply, install and manage the system ourselves, keeping your data and models under your control.
- Sizing for your models and users
- Power and cooling capacity checked in advance
- Installation in your server room or our racks
- Management and monitoring by Redeco IT Group

8x Blackwell Ultra GPUs
DGX B300 configuration
2.1 TB HBM3e
Total GPU memory
up to 144 PFLOPS FP4
sparse
Training & inference
Reasoning models at scale
NVIDIA H200 NVL and RTX PRO 6000: proven power for production inference

NVIDIA H200 NVL
141 GB HBM3e per GPU. For inference in standard rack servers, scalable card by card. Powerful production AI with large language models and many concurrent users.

NVIDIA RTX PRO 6000 Blackwell Server Edition
96 GB GDDR7 per GPU. A versatile server GPU for AI inference, fine-tuning and visual workloads.
H200 NVL or DGX B300?
Not sure? We provide independent advice based on your use case. Book an intro call
Networking and software
A GPU system is more than just the GPU. We supply and configure the complete networking and software stack for your AI environment.
Networking
NVIDIA ConnectX SuperNICs
High throughput per GPU node for fast data feeds to your GPUs.
NVIDIA InfiniBand
Low-latency clustering of GPU systems, the standard for training clusters.
NVIDIA Spectrum-X Ethernet
AI-optimised Ethernet fabric for inference and modern network design.
NVIDIA BlueField DPUs
Offloading and network security directly on the node, independent of your GPU capacity.
Software
NVIDIA AI Enterprise
A supported and secure AI software stack, with updates and support.
NIM microservices and model catalogue
Production-ready inference using optimised NVIDIA models and open-source alternatives.
NVIDIA vGPU licences
Share GPU capacity across workstations, VMs and teams, scaled per user.
Operating system and drivers
Operating system selection, drivers and container runtime configured and updated by us.
Together, we determine which combination fits your cluster or server. Book an intro call
AI in your office, under your control.
The NVIDIA DGX Spark brings local AI to your team's desk. No data is sent to OpenAI or Google; your business data stays within your own walls. We supply, install and manage it.
- Hardware selected and supplied by Redeco IT Group
- Local models installed and optimised
- Monitoring, updates and security by our team
Ideal for a single team or a proof of concept, before scaling up to H200 NVL or DGX B300.

GB10 Superchip
Blackwell GPU · Grace CPU
1 PFLOP
FP4 AI compute
128 GB
Unified Memory
4 TB SSD
Local storage
AI as part of your IT
Combine local AI with IT management, security and Microsoft 365 for one partner across your entire digital environment.
Frequently asked questions about local AI
What is the NVIDIA DGX Spark?
The NVIDIA DGX Spark is a compact desktop AI supercomputer. Built around the NVIDIA GB10 Superchip with Blackwell GPU and Grace CPU: 1 PFLOP FP4 AI compute, 128 GB unified memory and a 4 TB SSD.
When should I choose a DGX Spark rather than an H200 NVL or DGX B300 server?
A DGX Spark suits a single team or proof of concept with models of up to around 200 billion parameters. Choose servers with H200 NVL or DGX B300 for many concurrent users, larger models or training.
What is the difference between NVIDIA H200 NVL and DGX B300?
H200 NVL has 141 GB HBM3e per GPU and excels at production inference. DGX B300 (Blackwell Ultra) has 288 GB HBM3e per GPU and FP4 support, designed for large-scale training and reasoning.
What is needed to install a DGX B300 or H200 NVL system?
These systems require substantial power and cooling. We assess your server room in advance or install the system in our data centre racks in the Netherlands.
Do you also supply NVIDIA DGX B300 and H200 NVL?
Yes. As a NVIDIA Certified AI Advisor, we supply the full NVIDIA portfolio, including networking and NVIDIA AI Enterprise licences, with implementation and management.
Can you install the hardware in our data centre?
Yes. We install and implement on-site, in your data centre or in our data centre racks in the Netherlands.
Why choose local AI instead of ChatGPT or Copilot?
Because your data stays yours. With local AI, nothing is sent to OpenAI, Anthropic or Google. Models and business data run on-premise or in our own data centre racks in the Netherlands, under your control and managed by us.
Which models can you run locally?
Open-source models such as Llama and Mistral, plus specialist models where appropriate. Together, we choose what works best for your use case.
What does management include?
We manage the complete stack: hardware, models, software, monitoring, updates and security. One dedicated point of contact for everything.
What types of businesses is this suitable for?
From small and medium-sized businesses starting with one team or a proof of concept to larger organisations and AI platforms with their own server rooms or data centre environments.
Do you also build custom software around AI?
Yes. Alongside infrastructure, we build AI applications that fit your processes precisely: internal tools, RAG on your documents and integrations with existing systems.
Contact
Talk to us about local AI
From your first AI idea to a complete NVIDIA environment: we help you decide what fits your organisation, your data and your team.
Your use case comes first
We discuss what you want to achieve with AI, who will use it and which data needs to stay within your own environment.
The right hardware and software
Together, we assess whether DGX Spark, H200 NVL or DGX B300 fits, including networking, models and security.
A practical next step
You get a clear picture of implementation and management, on-site or in our data centre racks in the Netherlands.
A dedicated contact. A response within one working day.
Tell us about your plans or questions. You do not need to have selected any hardware yet.
