AXION Cloud
14 managed services, all billed per second
Same container images, same telemetry schema and same CLI as the hardware you own. Rent one accelerator for two minutes or 4,096 of them for three years.
Catalog
Every service
Cloud Instances
Accelerators by the second, no minimum term.
- SKUs
- M400, M400X, M200, V90, V40, V20
- Billing
- Per second, 2-minute minimum
Reserved Clusters
Contiguous capacity with a fabric you can benchmark.
- Size
- 32 – 4,096 accelerators
- Terms
- 1 month – 3 years
Interruptible Capacity
Cheap accelerators with a 90-second warning.
- Discount
- 62–78% off on-demand
- Warning
- 90 s SIGTERM + hook
Serverless Endpoints
Deploy a model, get a URL, pay per token.
- Models
- Catalog + your own weights
- Cold start
- 8.6 s median (warm cache)
Batch Jobs
Queue a training run and go home.
- Scheduling
- Gang, topology-aware, fair-share
- Resilience
- Auto checkpoint + node replacement
Fine-tuning Service
Upload examples, get a private adapter.
- Methods
- LoRA, QLoRA, full SFT, DPO
- Data
- Private, not used for base training
Model Gallery
Try any model in the catalog before you deploy it.
- Models
- 410 first-party and partner
- Playground
- Free, rate limited
Vector Store
Retrieval that keeps up with the model.
- Scale
- 10 B vectors per namespace
- Latency
- 6 ms p50 at 1 B vectors
Runbook Cluster Manager
Operate your own cluster the way we operate ours.
- Scope
- On-prem, colo, and AXION Cloud
- Health
- Predictive node drain on degradation
Outpost Fleet
Manage 10,000 edge devices you will never visit.
- Scale
- 100,000 devices per tenant
- Updates
- Signed, A/B partitions, auto-rollback
Loom Cloud
Streamed digital twins, no workstation required.
- Streaming
- WebRTC, 4K 60, 40 ms
- Concurrency
- 500 viewers per stage
Play Stream
Your game library, rendered in a datacenter.
- Tiers
- Free (ads, queue) / Plus / Ultra
- Max quality
- 4K 120 HDR or 1080p 240
Cloud Insight
Know what every token and every watt cost you.
- Metrics
- MFU, memory pressure, fabric health
- Attribution
- Per team, project, endpoint, model
Private AI Factory
Our operations team, your building.
- Model
- Customer-owned or leased hardware
- Operations
- 24/7 AXION NOC, on-site SLA
Instance types
On-demand rates
Two-minute minimum, per-second thereafter. Interruptible capacity gives 90 seconds of warning before reclamation.
| Type | Accelerators | VRAM | vCPU | Host RAM | Fabric | On-demand | Interruptible |
|---|---|---|---|---|---|---|---|
| v20.1x | 1 × V20 | 24 GB | 8 | 64 GB | 25 GbE | $0.62/hr | $0.19/hr |
| v40.1x | 1 × V40 | 48 GB | 16 | 128 GB | 100 GbE | $1.48/hr | $0.44/hr |
| v90.1x | 1 × V90 | 96 GB | 24 | 256 GB | 200 GbE | $2.94/hr | $0.91/hr |
| v90.4x | 4 × V90 | 384 GB | 96 | 1 TB | 4 × 400 G RDMA | $11.20/hr | $3.58/hr |
| m200.8x | 8 × M200 | 1.1 TB | 192 | 2 TB | 8 × 400 G RDMA | $22.40/hr | $6.94/hr |
| m400.1x | 1 × M400 | 288 GB | 36 | 512 GB | 800 G RDMA | $4.72/hr | $1.51/hr |
| m400.8x | 8 × M400 | 2.3 TB | 288 | 4 TB | 8 × 800 G RDMA | $34.80/hr | $9.62/hr |
| m400x.8x | 8 × M400X | 3.4 TB | 288 | 6 TB | 8 × 800 G RDMA | $41.60/hr | $11.80/hr |
Inference
Token pricing
| Axiom-L 8B | TTFT 210 ms | $0.04 / M |
| Axiom-L 70B | TTFT 340 ms | $0.24 / M |
| Axiom-L 500B | TTFT 620 ms | $0.9 / M |
| Axiom-Reason 120B | TTFT 880 ms | $0.6 / M |
| Axiom-Code 34B | TTFT 260 ms | $0.11 / M |
| Axiom-VL 40B | TTFT 410 ms | $0.35 / M |
| Axiom-Mini 2B | TTFT 90 ms | $0.01 / M |
| Axiom-Embed | TTFT 40 ms | $0.012 / M |
| Cadence Speech | TTFT 240 ms | $0.05 / M |
Footprint
11 regions
- us-west-2Oregon
- us-east-1Virginia
- us-central-1Texas
- eu-west-1Ireland
- eu-north-1Sweden
- eu-central-2Frankfurt
- ap-northeast-1Tokyo
- ap-southeast-2Sydney
- ap-south-1Mumbai
- sa-east-1São Paulo
- me-central-1Dammam