Hardware
NVIDIA A100 80GB SXM on the HGX platform. Bare-metal, direct customer relationships.
Synterran Compute
Capacity in Synterran’s Atlanta AI factory.
NVIDIA A100 capacity available on the current cluster, and NVIDIA B300 capacity opening for reservation ahead of Q3 2026 delivery. Direct partnership, bare-metal, configured to your workload — in Synterran’s Atlanta AI factory, next door to our engineering lab developing the Synterran Cycle.
A100 available now · B300 reservations open · Atlanta
Available now
A100 capacity.
Synterran operates NVIDIA A100 capacity on bare-metal infrastructure, with 99.9%+ cluster availability in continuous production since July 2025.
Configuration
16 vCPUs and 64–256 GB RAM per GPU, SXM NVLink fabric at 600 GB/s bidirectional. 8-GPU domains scoped to your workload.
Terms
Committed reservations:
- $1.19/hr — 12+ months
- $1.39/hr — 6 months
- $1.69/hr — monthly
Full-node minimum (8 GPUs per reservation).
Q3 2026
B300 capacity, reservations open.
NVIDIA B300 capacity is opening for reservation ahead of Q3 2026 delivery in Synterran’s Atlanta AI factory.
Hardware
NVIDIA B300, 288 GB HBM per GPU. Liquid-cooled, on the HGX platform.
Topology
Reserve in 8-GPU NVLink domains. Inter-domain fabric — InfiniBand or Ethernet — scoped to your workload.
Terms
Committed reservations:
- $3.99/hr — 3 years
- $4.49/hr — 2 years
- $5.49/hr — 1 year
Full-node minimum (8 GPUs per reservation).
Beyond the rack
Why Synterran?
Founder-direct
Direct line to the person accountable for your infrastructure.
A100 today, B300 next
Priority access to new generations.
Synterran Cycle
In active development.
Redundancy from the rack up
Reliability & certification.
- Power. OCP BBU direct on the busbar — rack-level battery.
- Network. Primary fiber with 5G failover on independent ground infrastructure.
- Cooling. Liquid cooling with redundant pumps and cooling towers.
- Tenant isolation. Bare-metal hardware dedicated per reservation. Per-tenant VLANs.
- Monitoring. Continuous infrastructure monitoring with on-site technical escalation for incidents.
- Certification. SOC 2 and ISO 27001 pursued during Deployment 1 buildout; certified capacity available after completion.
Fit
Is this a fit?
Direct-partnership capacity isn’t for every workload.
- You want bare-metal capacity on committed terms.
- You benefit from workload-scoped topology — NVLink domain sizing, inter-domain fabric selection, storage integration tuned to your job.
- You don’t require certifications your delivery window predates — or you’re aligned with the SOC 2 / ISO 27001 timeline above.
Reserve
Start the conversation.
-
1
Scope
Topology, density, term.
-
2
Written proposal
Configuration, pricing, delivery.
-
3
Signed term sheet
Reservation confirmed.
First response within a business day. Initial B300 allocation is finite; reservations fill in the order terms are signed.