Managed means we own the pager for the infrastructure we built. Not "we'll look at it during business hours" — actual on-call, actual drills, actual written reports with measured numbers. If an SLA is going to be missed, you hear it from us first.
The difference between tiers is how fast we answer, how often we rehearse failure, and whether the response target is contractual. Nothing in a lower tier is a broken promise — it is simply a narrower one.
99.9% uptime target
from $1,900 /cluster/month
99.95% uptime target
from $3,600 /cluster/month
99.99% uptime target
from $7,900 /cluster/month
from $220 /GPU node/month
Priced per node, not per GPU card. Includes driver/fabric health, scheduler tuning, and utilisation reporting.
from $4,000 /month
A prepaid hour pool you can drain anytime. For teams who build themselves but want a senior engineer one message away.
SLAs are usually written in language that means less than it appears to. These are our definitions, stated before you sign rather than after something goes wrong.
Managed work is not reactive only. Here is the recurring schedule by tier — the drills are the part that separates a real operations practice from a monitoring subscription.
Measured, not promised Median first response and quarterly drill outcomes are published on our status page. If we miss, you see the miss — that is the point of publishing.
If you already have a cluster that needs looking after, start with an audit. We'll tell you what condition it is in and what it would take to run it properly.