Recent outages at Microsoft and AWS prove that even the world's largest cloud providers can experience significant downtime. Business resilience can't be outsourced entirely to a single provider. Organizations need architectural flexibility, including choices like on-premises, private cloud, or hybrid deployment, to keep critical data and analytics available when disruption strikes.
Recently, businesses and governments around the world confronted an uncomfortable reality: even the technology infrastructure we rely on every day can become unavailable.
On November 12, 2026, a major Microsoft outage disrupted Outlook, Teams, and other Microsoft 365 services for more than 12 hours globally. Organizations experienced interruptions to essential business functions including email, calendars, and virtual meetings. The effects extended into critical segments like government agencies, financial services, healthcare, airlines, and other businesses.
This wasn't an isolated reminder. Just over a year earlier, on October 20, 2025, a major AWS service disruption originating in its US-EAST-1 region affected applications and services around the world. The disruption impacted businesses and popular digital services, preventing some users from performing routine activities and leaving organizations dependent on affected infrastructure waiting for services to recover.
Two of the world's largest technology providers. Two different incidents. One important lesson: no infrastructure is immune to disruption.
For business, data and technology leaders, that raises a more consequential question. If the infrastructure supporting your most critical data and analytics became unavailable tomorrow, could your business keep operating?
Resilience can't simply be outsourced. This isn't an argument against the cloud.
Cloud computing has transformed enterprise IT, delivering tremendous advantages in agility, scalability, and access to innovation. Microsoft and AWS operate extraordinarily sophisticated infrastructure. But moving critical workloads to a highly resilient provider isn't the same as designing a resilient business.
As organizations consolidate more applications, data, and operational processes into a small number of technology ecosystems, they can also introduce dependencies that only become apparent when something goes wrong.
The November 2026 Microsoft outage illustrates the potential consequences. When essential services became unavailable, the impact wasn't confined to IT departments. Business communications, meetings, and other everyday operations were disrupted.
The 2025 AWS outage demonstrated a similar interconnectedness. A disruption originating in one AWS region cascaded across applications and services used around the world.
The lesson isn't "don't use the cloud." It's "don't let resilience depend entirely on somebody else's cloud."
Not every workload carries the same business risk. But some data powers decisions and processes organizations simply can't afford to interrupt: fraud detection, financial transactions, telecommunications networks, supply chains, manufacturing operations, healthcare systems, and increasingly, AI-driven applications.
For these workloads, organizations should ask:
These aren't simply infrastructure questions. They're business resilience questions, and the answers should guide how your organization designs for the next disruption, not just the last one.
Resilience doesn't necessarily mean moving everything back on-premises. It doesn't mean distributing every workload across multiple clouds, either. It means making deliberate architectural choices based on the importance of the data and workload involved.
Some organizations may determine that their most critical analytical workloads should remain on infrastructure they control. Others may use private cloud environments. Still others will build hybrid architectures combining on-premises infrastructure and public cloud resources.
Choose an architecture that keeps your most sensitive workloads under direct control if sovereignty and compliance outweigh the convenience of full cloud reliance. Choose a hybrid model if your organization needs the agility of the cloud alongside the assurance of on-premises control for specific, high-stakes workloads.
What matters most is retaining the choice to put critical data and analytics where the business needs them, rather than being locked into one deployment model.
Organizations are already becoming more deliberate about where their workloads and data should run. Our 2025 survey conducted in partnership with Foundry found that over 80% of the organizations planning to repatriate are prioritizing the move of their AI and analytics workloads from the public cloud to on-premise, private cloud or hybrid environments by 2027.
That doesn't signal a wholesale retreat from the cloud. Instead, it points to a more selective approach to workload placement as organizations weigh considerations such as cost, security, performance, control, sovereignty, and resilience. For critical data and analytics, the question is increasingly less about choosing “cloud or on premises” and more about determining which environment best meets the requirements of each workload.
For organizations evaluating repatriation, resilience should therefore be part of the decision alongside economics and performance. Bringing selected workloads back to infrastructure an organization controls, or adopting a hybrid architecture, can provide another option for workloads where greater control over data, infrastructure, and operational dependencies is important.
That flexibility becomes increasingly important as AI moves into production. AI systems depend on continuous access to trusted enterprise data. If applications, agents, and models can't reach the data or analytics that provide their business context, the availability of the AI model itself doesn't solve the problem.
AI readiness, therefore, can't only be about models and compute. It also needs to encompass the resilience, availability, governance, and sovereignty of the data foundation underneath them.
For organizations building AI into mission-critical decisions and processes, this changes the architecture discussion. Leaders need to consider not only how quickly they can build AI applications, but also how those applications continue operating when an infrastructure provider, region, or service becomes unavailable.
Rocket® Vertica™, recognized by IDC as a worldwide leader in analytical databases, is a sovereign, AI-native data lakehouse built for enterprise-scale, mission-critical analytics and AI. It gives organizations the flexibility to run high-performance analytics where their data and business requirements dictate: on premises, private cloud, Kubernetes or across hybrid architectures. That flexibility allows organizations to make deployment decisions based on the characteristics of each workload, including performance, economics, governance, sovereignty, and resilience.
Vertica combines a massively parallel processing (MPP) architecture with Eon Mode, which separates compute from shared storage and enables organizations to scale resources independently as analytical requirements change. Its open architecture and support for Apache Iceberg give organizations greater flexibility in how they build their enterprise data environments without tying critical analytics to a single infrastructure approach.
Vertica also brings advanced analytics, machine learning, and AI closer to trusted enterprise data. By analyzing data where it lives and minimizing unnecessary data movement, organizations can maintain greater control over sensitive and mission-critical data while continuing to derive real-time insights from it.
Together, these capabilities enable organizations to make architecture and deployment decisions based on the requirements of each workload, including performance, economics, governance, sovereignty, and resilience.
For one organization, that could mean keeping mission-critical analytics close to operational systems. For another, it could mean maintaining sensitive data within infrastructure it directly controls. For another, a hybrid approach may provide the right balance of cloud agility and infrastructure control.
There's no single architecture for resilience. But organizations should have the freedom to put critical data and analytics where they make the most sense for the business. Vertica provides that choice, helping organizations balance resilience with high-performance analytics, sovereignty, control, and the flexibility to adapt as their requirements change.
The November 2026 Microsoft outage and the October 2025 AWS disruption are reminders that even the world's largest technology platforms can experience failures.
The question isn't whether organizations should choose cloud or on-premises. The better question is this: how much choice and control do you have when the environment you've chosen isn't available?
For the data and analytics your organization simply can't operate without, resilience needs to be designed into the architecture from the beginning. Because when critical data powers your business, where it lives, where it's processed, and whether you can continue using it during a disruption aren't just IT decisions. They're business decisions.
Ready to build an architecture that keeps your critical data available, no matter what happens upstream? Explore Rocket Vertica to see how your organization can gain the flexibility and control resilience requires.
What's New in Vertica 26.3: Accelerating the Journey from Analytics to AI-Powered Decision Making
Wie KI die Strategien zur Cloud-Rückführung verändert
Why Security Is Driving Cloud Repatriation of Analytical Databases in Large Enterprises
KI hat entdeckt, was den Menschen entgangen ist. Ist Ihr Großrechner als Nächstes dran?
Warum Token-Effizienz bei der Bewertung von On-Premises- und Frontier-Modellen wichtig ist
Wie modernisiert man eine CICS-Anwendung, ohne den Produktionsbetrieb zu unterbrechen?
