Hosting
Managed cloud hosting, in fixed monthly bands
We provision, deploy, monitor, back up, and update your infrastructure on a major public cloud, or on your own. Your quote names a flat monthly band for your size, not a usage multiple — a traffic spike becomes a stated overage, never a disputed invoice.
Starter
Up to about 5,000 monthly active users · one region · light media
A launch-stage footprint with monitoring, backups, and incident response included.
Flat monthly band, quoted
Growth
About 5,000 to 50,000 monthly active users · one region · moderate media
Growing user bases and traffic, still on a predictable flat band.
Flat monthly band, quoted
Scale
50,000+ monthly active users, or media-heavy · multi-zone
High-traffic and media-heavy platforms with room for spikes.
Flat monthly band, quoted
Enterprise
Multi-region · custom SLA · compliance boundary
Sized individually around your residency, uptime, and audit requirements.
Flat monthly band, quoted
Your band and its overage rate are named in the same written quote as the build: one document, before you sign.
How overage and the floor work
Overage is a stated rate, not the band multiple
Going over your band's limits is pure infrastructure cost, not additional management work, so your quote states one overage rate up front and overage is billed at the rate your agreement states, never at the band multiple.
Every band has a floor
Incident response has a fixed cost regardless of footprint size, so every band starts at a stated monthly floor — a client on a tiny footprint never pays less than what a single support call actually costs to staff.
What every band includes
Provisioning, deployment, monitoring with alerting, encrypted backups, security updates, and a named person who answers when something is wrong. The band is the whole bill for running the system; there is no separate management fee.
Why bands instead of a usage multiple
Media-heavy platforms — video, audio, and image-heavy catalogs — have egress that can rise tenfold in a single month. Usage-multiple billing turns that into a disputed invoice; fixed bands with a stated overage rate keep the bill predictable instead.
GPU hosting
Five ways to run GPU workloads, depending on the shape of the load
GPU rental is the dominant cost in these workloads, so it is never folded into the bands above. GPU capacity is quoted individually, never estimated on this page — hardware pricing moves too fast for a published range to stay honest.
Shared inference pool
Per-request or per-minute on our pooled capacity
Best for
Low and spiky volume, the cheapest way in
Dedicated GPU
Reserved monthly instance
Best for
Steady production load with a predictable bill
Burst capacity
On-demand, hourly
Best for
Batch jobs and periodic heavy runs
Client-owned hardware
You buy the GPUs, we operate them — management fee only
Best for
Cost-sensitive workloads at scale
On-premise air-gapped
Your hardware, your site, licensed software plus a support contract
Best for
Regulated, sovereign, and classified work