NVIDIA H200 Compute from £1.25/hr — Early Access

Carbon3.ai is launching the CarbonCore Platform for Creators in January 2026. This Early Access Programme lets a limited number of AI Creators access £1.25/hr NVIDIA H200 compute ahead of launch

Apply for Early Access

View Details

Performance you can build on

Frontier-scale power

Fine-tune and serve large models on a single 8xH200 powerhouse with low latency and high throughput.

High-speed I/O

Large-memory capacity and ultra-fast NVMe keep datasets and checkpoints close to the GPUs for rapid iteration.

Supercomputer fabric

Scale seamlessly to multi-node clusters over 3,200Gbps fabric for distributed, high-performance fine-tuning.

What early access includes

A controlled pre-launch onboarding programme for AI creators of all sizes.

£1.25/hr NVIDIA H200 compute – monthly reserved

3-month Early Access pricing window (from activation)

Early feature access to CarbonCore ahead of January launch

Eligibility review to align capacity with workloads

No payment required during application

How early access works

Apply

Submit your project and expected GPU needs.

Review

We assess workload suitability and capacity availability.

Activate

Approved creators begin using £1.25/hr NVIDIA H200 compute at platform launch.

Core platform capabilities

High-performance compute and essential AI services.

GPU compute & scaling

Provision NVIDIA H200 capacity for fine-tuning (previously “training”) and inference

Scale compute up/down based on model demand

Reserved monthly GPUs for predictable performance

Model fine-tuning pipelines

Distributed fine-tuning (previously “training”) on LLMs, agents and multimodal models

Preconfigured PyTorch/JAX/TensorFlow environments

Runtime metrics for cost, utilisation and performance

Model deployment & serving

Deploy models to scalable, production-ready endpoints

Autoscale inference workloads based on traffic

Support for RAG, agentic systems and generative apps

Workflow orchestration

Automate data → fine-tune → validate → serve cycles

Trigger multi-step AI workflows programmatically

Integrate with CI/CD or MLOps pipelines

Observability & governance

Real-time GPU monitoring, logs and metrics

Role-based access control (RBAC)

Audit trails and policy-driven controls

Programme terms (summary)

Early Access provides discounted £1.25/hr NVIDIA H200 compute on monthly reserved nodes (8 GPU minimum). The 3-month Early Access pricing period runs from the official programme launch date — not the date you sign up or are approved. Access is capacity-limited and subject to eligibility review. GPU hours may be capped during pre-launch to ensure fair usage. Full terms issued upon approval.