GPU Rent Hub status and uptime. Nine components, three datacenters.
GPU Rent Hub publishes live status and 90-day uptime for compute, network, dashboard and API, crypto deposit detection and the Tor service across DFW-1 Dallas, IAD-1 Ashburn and PDX-1 Hillsboro; trailing twelve-month uptime is 99.97%. Updated automatically from the same probes that page us. Incidents are written by the engineer on call, not by marketing — which is why some of them are blunt.
Last check:
DFW-1 — compute
uptimeDFW-1 — network
uptimeIAD-1 — compute
uptimeIAD-1 — network
uptimePDX-1 — compute
uptimePDX-1 — network
uptimeDashboard & API
uptimeDeposit detection (all coins)
uptimeTor onion service
uptimeWhat incidents has GPU Rent Hub had since 2021?
Everything that affected more than one customer, back to the first month of operation. Maintenance windows are announced 7 days ahead in the dashboard and listed here only if they overran. Power, cooling and carrier detail for each site is on the US datacenters page.
IAD-1 edge router failover Degraded
Primary edge lost a line card. Failover worked; 25 minutes of elevated latency (+30 ms) to some East-coast destinations while the secondary re-learned routes.
Written by RK, on call.
Monero deposit detection delayed Degraded
Our monerod fell behind the chain after a restart. Deposits were credited late (up to 2h20) but all were credited at the rate of their actual first confirmation, not the late detection time.
Written by MT, on call.
DFW-1 upstream fibre cut Outage
A construction crew cut the Lumen path into DFW-1. Cogent and HE stayed up but our BGP re-convergence took 14 minutes longer than it should have because of a stale prefix-list on the edge. Compute was unaffected; customers saw packet loss of 30–60% for 2h29 total. Root cause on the prefix-list has been fixed and a third diverse path (Zayo) was ordered.
Written by RK, on call.
DFW-1 cage 2 PDU replacement overran Degraded
Planned 60-minute window ran 70 minutes. 96 A6000/A40 instances rebooted once instead of the announced zero-reboot swap; the A-feed transfer switch did not hold. Affected customers were credited 3 days.
Written by AV, on call.
API latency Degraded
Catalogue endpoint p99 went above 4 s after a stock-sync job locked the inventory table. Dashboards were slow; no instance was affected. Job moved to a read replica.
Written by MT, on call.
DFW-1 cooling loop alarm Degraded
Facility CRAH unit fault raised inlet temperatures in cage 4 to 31 °C. We throttled 8-way H100/H200 nodes to 80% power for 1h45 rather than risk thermal shutdown. No reboots. Customers on affected nodes were credited 2 days.
Written by RK, on call.
Dashboard unavailable Outage
A bad deploy of the dashboard broke login for everyone for 2.5 hours. The API and all instances were fine, which is cold comfort if you needed the console at 4 AM. We added a canary stage after this.
Written by AV, on call.
PDX-1 network flap Degraded
Upstream (HE) maintenance not communicated to us. 25 minutes of intermittent loss for PDX-1.
Written by MT, on call.
DFW-1 utility power event Outage
Utility feed dropped during a storm; generators started but the B-side UPS failed to carry cage 3 during transfer. 412 instances lost power for up to 4h45 (hard reboot). No data loss reported. Facility replaced the UPS string; we moved to 2N UPS on all cages by October 2024. Credited 7 days.
Written by RK, on call.
IAD-1 upstream degradation Degraded
Zayo path congested; we had one upstream at IAD-1 at the time. Second carrier (Lumen) was live by March 2023.
Written by AV, on call.
IAD-1 opening delayed Notice
First IAD-1 deployments were scheduled for 28 Nov and slipped two days waiting on a cross-connect. Pre-orders were credited a week.
Written by MT, on call.
Bitcoin deposit detection paused Degraded
Our first payment incident. bitcoind ran out of disk. Deposits were credited late, at the correct rate. Disk alarms were, in retrospect, a good idea.
Written by RK, on call.
DFW-1 switch failure Outage
The single top-of-rack switch in our first rack died. Forty RTX 3090s were offline for 1h45 while we swapped it. Every rack has had redundant ToR switches since May 2021.
Written by AV, on call.
A deposit incident delays crediting, never the rate you get: the mechanics are on paying for GPU rental with crypto, and the node policy behind the August 2026 entry on Monero GPU hosting. Downtime credits are one day of term per hour beyond 99.9%, as written in the terms of service.