Catalog status
CNRG.ai capacity projection for SINERGY.SU. Prices are not shown until the upstream catalog publishes customer-visible terms.
Updated at 2026-06-08T00:00:00.000Z
Reserved H100/A100 training cluster
Reserved GPU capacity for model training and fine-tuning workloads that need predictable availability, verified providers, and SLA visibility.
- Regions
- EU, MENA, APAC, North America
- Accelerators
- NVIDIA H100, NVIDIA A100, high-memory GPU nodes
- Workloads
- training, fine-tuning
- Offer status
- intake-ready
- Isolation
- dedicated, private-tenant
- SLA target
- up to 99.9% target for reserved windows
- Minimum reservation
- 12 hours
- Next step
- Open console intake or submit a search_capacity request with GPU class, region, duration, and isolation needs.
Inference and batch burst capacity
Elastic compute capacity for inference spikes, evaluation batches, RAG indexing, and temporary queues where teams need speed more than long-term reservation.
- Regions
- EU, MENA, APAC
- Accelerators
- NVIDIA L40S, NVIDIA A100, CPU/GPU mixed nodes
- Workloads
- inference, batch, rag-indexing
- Offer status
- intake-ready
- Isolation
- shared, dedicated
- SLA target
- best-effort spot or reserved SLA by provider
- Minimum reservation
- 1 hours
- Next step
- Collect latency region, expected tokens/jobs, peak window, and acceptable interruption policy.
SINERGY.SU CIS reserved capacity
Localized CIS capacity lane for teams that need regional provider relationships, Russian-language operator support, and CNRG.AI-backed accounting.
- Regions
- CIS, Kazakhstan, Armenia, Georgia, Turkey
- Accelerators
- NVIDIA H100, NVIDIA A100, NVIDIA L40S, provider-declared GPU nodes
- Workloads
- training, fine-tuning, inference, batch
- Offer status
- intake-ready
- Isolation
- dedicated, private-tenant
- SLA target
- localized SLA by provider and reservation window
- Minimum reservation
- 6 hours
- Next step
- Route through SINERGY.SU intake with language, region, compliance, payment, and provider-locality requirements.
Provider idle GPU monetization
Provider-side listing path for unused GPU clusters, custom availability windows, verification data, utilization controls, and settlement-ready usage records.
- Regions
- Global, CIS
- Accelerators
- provider-declared GPU nodes, CPU/GPU mixed clusters, storage/network add-ons
- Workloads
- provider-listing
- Offer status
- planned
- Isolation
- shared, dedicated, private-tenant
- SLA target
- provider-declared SLA after verification
- Minimum reservation
- Custom
- Next step
- Collect provider identity, hardware inventory, regions, windows, pricing rules, isolation model, and audit requirements.
Agentic capacity intake
Machine-readable intake lane for LLM clients and agentic planners that need to transform natural-language compute needs into search and reservation parameters.
- Regions
- Global, CIS
- Accelerators
- GPU class inferred from workload, memory inferred from model size, storage/network inferred from data path
- Workloads
- training, fine-tuning, inference, batch, rag-indexing
- Offer status
- intake-ready
- Isolation
- shared, dedicated, private-tenant
- SLA target
- derived from user constraints before compare_capacity
- Minimum reservation
- Custom
- Next step
- Extract workload class, accelerator target, region, budget, timing, SLA, isolation, and provider constraints.