On July 29, 2025, enterprises relying on Microsoft Azure’s East US region experienced an unexpected disruption that reverberated across numerous organizations. Attempted allocations for virtual machines failed. The root cause wasn’t a network breach, misconfiguration, or other complex technical mishap. It was something shockingly basic: a lack of capacity. A sudden surge in demand outstripped the available computing resources, rendering Azure unable to fulfill virtual machine requests for several users. Although Microsoft flagged the issue as resolved by August 5, numerous administrators reported lingering challenges, highlighting a deeper problem with the “elastic” cloud solutions we’ve come to trust.

This Azure incident wasn’t an anomaly. In the past few years, such capacity-related outages have become more prevalent, impacting various public cloud providers. Enterprises were assured that cloud environments were engineered to scale automatically and effortlessly, capable of handling unpredictable spikes in demand without issue. But recent reality checks, including this Azure crisis, suggest otherwise.

Organizations are increasingly finding themselves grappling with the hard truth: The public cloud isn’t some magic infrastructure immune to the challenges of physical systems; the cloud is simply someone else’s computer, with all the usual limitations, pitfalls, and infrastructure constraints. Enterprises must recognize this and account for the possibility—the inevitability—of failures in scalability.

Promises of endless scalability

Public cloud providers built their reputations on a simple yet powerful claim of instant scalability up or down in response to user demand. Need more computing power to handle a traffic spike? No problem! Just allocate more virtual machines. The seamless flexibility offered by Microsoft Azure, Amazon Web Services (AWS), Google Cloud, and others was a major selling point for enterprises transitioning away from on-premises data centers. Companies reasoned that offloading infrastructure management to hyperscalers would grant them access to virtually limitless computing power while eliminating headaches related to hardware provisioning.

The Azure East US region incident underscores a serious flaw in this narrative. “Limitless capacity” falls apart when demand exceeds supply. Cloud providers, while expansive in scale, still rely on physical data centers with finite infrastructure. When a region’s compute resources are exhausted, virtual machine allocation simply fails, leaving enterprises scrambling for alternative solutions. This time, the issue stemmed from surging demand for certain types of compute instances, likely compounded by enterprisewide Kubernetes upgrades coinciding with the end-of-life timeline for Kubernetes 1.30. These overlapping pressures likely overwhelmed the system.

Scalability in the cloud isn’t inherently limitless. When we talk about elasticity, what we really mean is the capacity to scale within the confines of available infrastructure—an infrastructure that, at the end of the day, is still constrained by physical hardware and resource limitations.

Holding providers accountable

As cloud capacity challenges continue to emerge, enterprises must reassess how they engage with public cloud providers. The first step is a renewed focus on service-level agreements (SLAs). For years, SLAs have served as a measure of trust between cloud providers and their customers. These agreements outline performance metrics such as uptime, latency, and response times. However, metrics like “available capacity” or “scalability thresholds” are rarely addressed explicitly in standard contracts, leaving enterprises without a clear recourse when capacity issues arise.

Enterprises should revisit their SLAs and consider stricter requirements. A well-drafted SLA should include clauses that address failure to allocate resources and enforceable commitments related to scalability, geographic availability, and redundancy. Compensation for falling short of these guarantees must also be outlined, whether in the form of monetary payments, service credits, or some combination.

Enterprises should also insist on telemetry visibility: consistent, clear insights into cloud resource usage and availability. Monitoring tools alone aren’t enough if the cloud provider does not transparently communicate overall capacity trends and projected constraints. Customers using the Azure East US region, for instance, would have greatly benefited from earlier warnings that demand in certain instance classes was exceeding availability. Microsoft suggested using alternative instance types or migrating workloads to East US 2, but many enterprises learned of these options far too late, after their operations had already been disrupted.

Responding to failure in the cloud age

Capacity issues will almost certainly arise again. The question is how enterprises will respond and adapt. Enterprises must treat the cloud not as a perfect, endlessly scalable solution, but as normal infrastructure subject to failure and constraints. They should pursue practical steps, including the enforcement of robust SLAs, diversification of workloads across multiple regions or providers, and internal contingency plans for capacity-related failures.

Organizations should also consider hybrid or multicloud strategies to mitigate risk. By spreading workloads across multiple providers or maintaining a baseline capacity in private data centers, enterprises can ensure that critical operations remain insulated from capacity constraints on any one platform. This hybrid model, though more complex, recognizes that no single provider can always meet 100% of a company’s compute needs.

The cloud computing industry must address this growing trust gap around scalability. Providers need to increase transparency about capacity constraints and communicate more proactively during demand surges. Customers should feel confident that even when workloads demand rapid scaling, they won’t encounter unexpected resource allocation errors.

Ultimately, the Azure East US region incident serves as a wake-up call to enterprises and providers alike. Although the cloud offers unprecedented flexibility, scalability is not an abstract, automatic guarantee; it’s a shared responsibility that must be actively negotiated, prepared for, and enforced through clear accountability. If the promises of elastic computing are to remain credible, the industry must embrace greater transparency, accountability, and collaboration with its customers.