Six months after Iranian drone strikes broken three AWS information facilities within the UAE and Bahrain, Amazon nonetheless hasn’t restored full service to both area, and it gained’t even present a Bahrain replace till early 2027. That single, unresolved outage sits inside a dataset of 30,246 cloud and SaaS outages that monitoring agency IncidentHub tracked throughout the primary half of 2026 alone.
A Reliability Report With Uncomfortable Numbers
IncidentHub’s H1 2026 Cloud and SaaS Reliability Report counted these 30,246 outages throughout 1,082 monitored suppliers between January and June, with Might alone accounting for six,070 incidents. Solely 15.9% of the suppliers IncidentHub tracks recorded zero outages throughout the interval. Amongst main cloud suppliers, AWS and Microsoft Azure every logged 14 incidents, Google Cloud Platform logged two, together with an information heart hearth in India that took 21 days and 12 hours to completely mitigate, and Oracle Cloud logged two. Edge and CDN suppliers fared worse in uncooked rely: Cloudflare alone logged 487 outages in six months, in opposition to 161 for Fastly and 59 for Akamai.
Operational Impacts of Cloud Outages
Particular person incidents carried actual enterprise penalties past the rely. A Google Cloud automated system positioned the developer platform Railway’s manufacturing account into suspension on Might 19, chopping off each buyer workload working on it for roughly eight hours. A ransomware assault by the group ShinyHunters in opposition to learning-platform supplier Instructure knocked Canvas offline for an estimated 8,800 to 9,000 establishments proper as many colleges have been coming into last exams, with restoration at particular person campuses starting from a single day to greater than every week. GitHub logged 169 incidents over the identical six months, together with an April 27 incident by which a surge of scraping site visitors overwhelmed its search infrastructure and knocked out situation, pull-request, and Actions search company-wide for greater than six hours.
Why This Differs From the Multicloud Debate Corporations Already Know
Enterprise conversations about multicloud technique have principally centered on value and lock-in: whether or not spreading workloads throughout AWS, Azure, and Google Cloud lowers vendor pricing energy and eases the ache of switching suppliers later. IncidentHub’s numbers level towards a special argument. Workload focus on a single supplier, or on a single edge community like Cloudflare, isn’t a theoretical threat. It’s a documented, frequent operational occasion, one that may run from an eight-hour disruption to an outage that, six months on, AWS nonetheless hasn’t closed out.
Bodily and Geopolitical Threat: The Restoration That Isn’t Completed
The Center East strikes are the clearest illustration of why resilience planning has to account for bodily and geopolitical threat alongside routine technical failure. A regional outage brought on by drone strikes doesn’t match the failover assumptions constructed round software program bugs or {hardware} faults, and the restoration has borne that out. AWS’s UAE area, ME-CENTRAL-1, continues to be lacking one in every of its three availability zones, and the corporate says it would share a restoration replace within the coming months. Its Bahrain area, ME-SOUTH-1, hasn’t come again on-line in any respect, and AWS isn’t promising an replace earlier than early 2027. What the corporate described in March as a restoration measured in a number of months is now approaching a 12 months for one area, with no fastened finish date.
The Board Query This Knowledge Truly Raises
My take: the resilience case for multicloud structure is robust sufficient now that leaving it purely to whichever group owns cloud infrastructure undervalues the stakes concerned. A single-provider outage lasting the higher a part of a 12 months, or an edge-network incident disrupting 1000’s of downstream clients without delay, is a business-continuity occasion with income, authorized, and reputational penalties. That belongs in entrance of a board’s threat committee, not buried in an infrastructure group’s backlog.
Prioritizing Essential Workloads
None of this implies each firm wants full workload portability throughout three clouds. The operational overhead of sustaining that form of redundancy is actual, and for a lot of companies, it isn’t price the fee. The extra helpful commonplace popping out of IncidentHub’s information is narrower: establish which particular workloads would trigger essentially the most harm if their single supplier went down for days fairly than hours, and construct redundancy round these workloads particularly, fairly than treating multicloud as an all-or-nothing architectural stance.
IncidentHub’s second-half 2026 numbers will present whether or not the incident fee is climbing or leveling off, however the first half already provides boards a concrete baseline to weigh in opposition to their very own focus threat. Thirty thousand outages in six months isn’t an argument for panic. It’s an argument for realizing, upfront, which workloads sit contained in the outage rely’s blast radius, as a result of Bahrain’s clients already know what the worst case appears like: six months of silence, with the subsequent phrase not due till 2027.
Six months after Iranian drone strikes broken three AWS information facilities within the UAE and Bahrain, Amazon nonetheless hasn’t restored full service to both area, and it gained’t even present a Bahrain replace till early 2027. That single, unresolved outage sits inside a dataset of 30,246 cloud and SaaS outages that monitoring agency IncidentHub tracked throughout the primary half of 2026 alone.
A Reliability Report With Uncomfortable Numbers
IncidentHub’s H1 2026 Cloud and SaaS Reliability Report counted these 30,246 outages throughout 1,082 monitored suppliers between January and June, with Might alone accounting for six,070 incidents. Solely 15.9% of the suppliers IncidentHub tracks recorded zero outages throughout the interval. Amongst main cloud suppliers, AWS and Microsoft Azure every logged 14 incidents, Google Cloud Platform logged two, together with an information heart hearth in India that took 21 days and 12 hours to completely mitigate, and Oracle Cloud logged two. Edge and CDN suppliers fared worse in uncooked rely: Cloudflare alone logged 487 outages in six months, in opposition to 161 for Fastly and 59 for Akamai.
Operational Impacts of Cloud Outages
Particular person incidents carried actual enterprise penalties past the rely. A Google Cloud automated system positioned the developer platform Railway’s manufacturing account into suspension on Might 19, chopping off each buyer workload working on it for roughly eight hours. A ransomware assault by the group ShinyHunters in opposition to learning-platform supplier Instructure knocked Canvas offline for an estimated 8,800 to 9,000 establishments proper as many colleges have been coming into last exams, with restoration at particular person campuses starting from a single day to greater than every week. GitHub logged 169 incidents over the identical six months, together with an April 27 incident by which a surge of scraping site visitors overwhelmed its search infrastructure and knocked out situation, pull-request, and Actions search company-wide for greater than six hours.
Why This Differs From the Multicloud Debate Corporations Already Know
Enterprise conversations about multicloud technique have principally centered on value and lock-in: whether or not spreading workloads throughout AWS, Azure, and Google Cloud lowers vendor pricing energy and eases the ache of switching suppliers later. IncidentHub’s numbers level towards a special argument. Workload focus on a single supplier, or on a single edge community like Cloudflare, isn’t a theoretical threat. It’s a documented, frequent operational occasion, one that may run from an eight-hour disruption to an outage that, six months on, AWS nonetheless hasn’t closed out.
Bodily and Geopolitical Threat: The Restoration That Isn’t Completed
The Center East strikes are the clearest illustration of why resilience planning has to account for bodily and geopolitical threat alongside routine technical failure. A regional outage brought on by drone strikes doesn’t match the failover assumptions constructed round software program bugs or {hardware} faults, and the restoration has borne that out. AWS’s UAE area, ME-CENTRAL-1, continues to be lacking one in every of its three availability zones, and the corporate says it would share a restoration replace within the coming months. Its Bahrain area, ME-SOUTH-1, hasn’t come again on-line in any respect, and AWS isn’t promising an replace earlier than early 2027. What the corporate described in March as a restoration measured in a number of months is now approaching a 12 months for one area, with no fastened finish date.
The Board Query This Knowledge Truly Raises
My take: the resilience case for multicloud structure is robust sufficient now that leaving it purely to whichever group owns cloud infrastructure undervalues the stakes concerned. A single-provider outage lasting the higher a part of a 12 months, or an edge-network incident disrupting 1000’s of downstream clients without delay, is a business-continuity occasion with income, authorized, and reputational penalties. That belongs in entrance of a board’s threat committee, not buried in an infrastructure group’s backlog.
Prioritizing Essential Workloads
None of this implies each firm wants full workload portability throughout three clouds. The operational overhead of sustaining that form of redundancy is actual, and for a lot of companies, it isn’t price the fee. The extra helpful commonplace popping out of IncidentHub’s information is narrower: establish which particular workloads would trigger essentially the most harm if their single supplier went down for days fairly than hours, and construct redundancy round these workloads particularly, fairly than treating multicloud as an all-or-nothing architectural stance.
IncidentHub’s second-half 2026 numbers will present whether or not the incident fee is climbing or leveling off, however the first half already provides boards a concrete baseline to weigh in opposition to their very own focus threat. Thirty thousand outages in six months isn’t an argument for panic. It’s an argument for realizing, upfront, which workloads sit contained in the outage rely’s blast radius, as a result of Bahrain’s clients already know what the worst case appears like: six months of silence, with the subsequent phrase not due till 2027.
