Early on May 7, smoke rose over an industrial park in Almere, Netherlands. Firefighters from Amsterdam, Schiphol and beyond raced to the scene. Inside a facility run by NorthC Datacenters, flames had taken hold in a rear compartment packed with technical gear and the emergency power supply. That single blaze brought down IBM Cloud operations for hours.
Customers noticed first. Services hosted in the AMS03 facility simply stopped responding around 0715 UTC. Tickets marked severity one sat unanswered. Support channels went quiet. One enterprise user reached an account manager directly before learning the truth. The entire datacenter had lost power. Four hours. Minimum.
Yet IBM’s official status page stayed green. No incidents listed. No alerts pushed. Monitoring service StatusGator picked up reports from at least ten users marking the platform as down. Downdetector captured a spike in complaints that stretched past noon. The mismatch between reality and the public dashboard raised eyebrows fast.
IBM confirmed the cause later that day. A spokesperson told The Register: “IBM is aware of a fire at a datacenter in Amsterdam which serves IBM, in addition to others. The facility has been evacuated and there are no reported injuries. We are working closely with emergency services, addressing the effect on our operations, and coordinating directly with affected clients to address any impacts.”
The site sits just miles from Schiphol airport. NorthC’s Almere facility spans 26,000 square meters and delivers 11 MW of capacity. The fire broke out around 08:45 local time in a space housing the backup power systems. Dutch public broadcaster NOS reported massive smoke plumes visible from the A6 highway. Fire crews declared the affected compartment lost and focused on containment. No dangerous substances were detected in the smoke, yet authorities issued an NL-Alert urging nearby residents to close windows and shut off ventilation.
NorthC itself stated that everyone inside the building had been evacuated in time. University of Utrecht saw applications and its website falter. Medical communication provider Infomedics lost phone and chat access. General practitioner systems across the Netherlands went unreachable. The collateral damage spread beyond cloud tenants.
But for IBM Cloud customers the outage hit harder than expected. Many run production workloads in the region. Some maintain hybrid setups that suddenly lost a leg. Sev 1 tickets lingered without updates until personal outreach produced answers. The episode exposed gaps in how outages get communicated when the provider itself depends on a third-party facility.
This incident arrives against a backdrop of repeated trouble. Network World documented multiple authentication-related disruptions throughout 2025 that left users locked out of resources. One lasted over 14 hours. IBM had already adjusted its basic support tier last September, removing the ability for those customers to open or escalate cases through the portal. Self-service via the console became the new normal. The policy change drew criticism at the time. Now it looks even more consequential.
Power infrastructure failures rarely announce themselves politely. A diesel tank on site required specialized crash tenders from airports to keep it cool during firefighting. Emergency power should have bridged the gap. Instead the compartment burned. Investigators have not yet released an official cause. Until they do, speculation runs toward electrical faults or equipment overheating, common triggers in such facilities.
Enterprise technology leaders have watched cloud providers promise redundancy for years. Multi-zone architectures. Automatic failover. Yet when the underlying physical plant suffers a direct hit, those abstractions sometimes collapse. IBM maintains multiple sites across Europe. Clients with resources only in AMS03 discovered the limits of that strategy the hard way.
By late afternoon on May 7, IBM’s status page finally acknowledged the situation under “Data Center Power Infrastructure” for Amsterdam 03. The listing noted an ongoing investigation tied to a facility provider fire alarm or hazard requiring evacuation. Services began recovering in stages. Full restoration timelines varied by workload.
Recent coverage adds context to the pattern. Data Center Dynamics reported the fire on May 7 and linked it directly to IBM’s reported issues at its Amsterdam 03 facility. NorthC posted updates confirming the blaze had been downgraded and brought under control overnight. The company emphasized quick evacuation and cooperation with authorities.
Industry observers point to broader questions about colocation risk. When cloud giants lease space from specialized operators, they trade some control for scale. Fire suppression systems, redundant power feeds and rigorous maintenance become the operator’s responsibility. A single missed detail can cascade. In this case the emergency power area itself became the ignition point or at least the hardest hit zone.
Customers have grown accustomed to hyperscaler outages that resolve within minutes. This one lasted hours. Communication lagged. The status page disconnect frustrated teams already scrambling to reroute traffic or activate backups. One sysadmin posted on X that the episode served as a reminder that resilience differs from true backup. The comment gained traction among infrastructure professionals.
IBM has not released a formal root cause analysis yet. Past incidents prompted detailed post-mortems published on the company’s status site. Expect the same here once the fire investigation concludes and technical teams complete their review. In the meantime, affected clients are receiving direct outreach. Some have been offered credits or assistance migrating workloads.
The episode underscores an uncomfortable truth for technology buyers. No provider stands immune to physical events. Fires, floods, grid failures. They happen. The difference lies in preparation, transparency and speed of recovery. IBM Cloud’s track record over the past year shows several high-severity events. Each adds pressure to demonstrate stronger operational maturity.
Enterprise architects now face fresh calculations. Should they spread workloads across more providers? Invest in stronger disaster recovery automation? Demand better visibility into facility-level risks from their cloud partners? The Almere fire supplies fresh data for those discussions.
Smoke has cleared over the NorthC site. Fire crews have stood down. Yet the memory of those silent hours will linger for teams that watched dashboards go dark without warning. For IBM, the task ahead involves not only fixing the immediate technical issues but rebuilding confidence that such surprises can be minimized. The cloud runs on trust as much as it runs on silicon and power. Both must hold firm.
WebProNews is an iEntry Publication