Reflections and Predictions:
What 2025 Taught Us and What’s Coming in 2026

A Look Back at 2025: Reflections and Surprises

If I had to summarize 2025 in a single phrase, it would be that it’s the year AI got real.

AI stopped being something organizations were exploring and became something they were dependent on. That shift from exploration to necessity was the defining narrative for enterprise technology.

But adopting AI at scale isn’t the same thing as adopting a new analytics tool or adding another cloud service. AI touches the entire infrastructure stack because it changes how applications are architected. As a result, it changes how data moves. It changes the way we look at compute. It changes how security, privacy, and sovereignty concerns lead to new views on where workloads should optimally reside—whether on-premises, in the cloud, at the edge—or (increasingly) all three.

Looking back, 2025 was full of surprises. Some shifts happened faster than expected. Others highlighted weak spots in the way most organizations currently work. And a few significant developments reshaped how enterprise IT teams think about AI and the infrastructure behind it. These include:

• Inference shifted back to on-prem and to the edge. Organizations realized that while training may belong in the public cloud, inferencing benefits from the predictable pricing, low latency, and tight data control offered on-prem or at the edge.
• CPUs made an unexpected comeback for GenAI workloads. This was especially true in retrieval and vector database pipelines, where the CPU–GPU pairing proved more effective than many expected. This is because CPUs often perform better on tasks like search and filtering, while GPUs handle parallel computations.
• Power became the new limiting factor. Hyperscalers hit real-world power ceilings, forcing everyone to rethink what “scaling AI” actually means.
• Neoclouds emerged seemingly overnight. Specialized cloud providers delivering high-performance GPU infrastructure for AI and machine-learning workloads have sprung up, signaling a major shift in the service provider landscape.

Together, these trends point to a future where AI is not a standalone innovation, but a distributed capability woven into every part of the enterprise, including applications, data, operations, and even physical locations.

With all that in mind, I want to share three predictions for 2026 and the years ahead. Taken together, these trends will shape how enterprises build, run, and secure their systems for the next decade.

Predictions for 2026

#1: Enterprises will shift from AI-first to AI-smart

After a year of chaotic AI trials, enterprises will lock into ROI-driven AI use cases in 2026 now that they recognize that enthusiasm isn’t enough. Many organizations spent 2025 racing to launch pilots and plug AI into every workflow. This next year will bring about a more thoughtful application of AI in what I call an “AI-smart” approach. It means treating AI as a mission-critical service that solves a clear business challenge and is reliable, governable, and designed for the long haul.

This maturity hinges on three capabilities.

• Resiliency: AI will quickly become something users rely on every day, which means any slowdown or outage will be unacceptable. Meanwhile, the hardware and frameworks behind AI, such as GPUs, CPU accelerators, ARM systems, and fast-changing model architectures will continue to evolve constantly. Platforms need to absorb this churn without disrupting operations.
• Day-2 operations: Deploying an AI service can be quick. But keeping it healthy over time is a real challenge. Updates, patches, model refreshes, vector databases, compliance requirements, and distributed deployments all demand tight coordination between developers and IT.
• Integrated security: AI introduces new data sensitivities and must run consistently across cloud, on-prem, and edge environments. Security policies must extend over distributed environments with sovereign control unified across traditional business critical applications, not stitched together after the fact.

Enterprises that master the resiliency, operations, and security needs of AI will be the ones that turn AI into lasting competitive advantage.

#2: Evolution of the Sovereign Edge

The edge isn’t an afterthought anymore, it’s the new frontline. Enterprises are realizing that AI will lead to more distributed infrastructure as AI logic moves to the source of the data, which is frequently generated at remote, edge locations where sensors reside. As AI spreads outward and data volumes explode, the edge can’t remain a collection of disconnected remote sites. It will increasingly be considered a sovereign element of distributed infrastructure with the same consistency, governance, and security needs as any centralized environment. With everything from industrial sensor information to medical imaging and video analytics sourced at the edge, it’s no longer practical or compliant to ship everything to a central cloud. Processing data where it originates is simply more efficient.

This shift will give rise to the sovereign edge—a globally managed but locally autonomous layer of compute and AI that keeps data under the organization’s control while enabling modern, distributed workloads.

To make this work, three capabilities are essential.

• Global management provides centralized visibility and consistent operations across potentially thousands of sites.
• Integrated security ensures that policies can be implemented locally and propagated globally to meet growing sovereignty and compliance requirements.
• Edge resiliency allows sites to withstand outages, preserve state, and continue operating even with limited connectivity or without on-site expertise.

As these foundations take hold, the sovereign edge will redefine enterprise architecture and become the natural home for real-time and regulated AI workloads.

#3: The Rise of Platform Wars

Organizations looking to scale their AI ambitions are discovering that stitching together point solutions and juggling disconnected tools isn’t a viable long-term strategy. The next decade won’t be about cloud-versus-cloud or which Kubernetes solution to adopt. It will be defined by a broader competition: full-stack platforms that can run AI reliably anywhere—alongside traditional applications, across clouds, datacenters, and an increasingly important sovereign edge.

What enterprises need now is a unified foundation that delivers flexibility, consistency, and operational strength across all environments. This is why platform architecture is quickly becoming one of the most consequential decisions IT leaders will make.

Winning platforms will stand out in three areas.

• Resiliency: The ability to operate predictably across mixed hardware, evolving accelerators, and distributed locations.
• Modernization: The ability to support new GPUs and application models without forcing architectural overhauls or duplicating operations for VMs and containers.
• Choice: Teams want the freedom to pick their preferred container stacks, model frameworks, processors, and deployment locations, not be locked into rigid ecosystems.

In the end, the platform wars come down to one test: Which platform can run AI anywhere with confidence? The leaders will be those that unite resiliency, modernization, and choice into one cohesive operating model.

The AI-Smart Road Ahead

If 2025 was the year AI became unavoidable, 2026 will be the year it becomes intentional. The scramble to experiment is giving way to the clearer reality that AI is only as valuable as the infrastructure, governance, and operational discipline supporting it. The organizations that win won’t be the ones chasing every new model. They’ll be the ones aligning AI to business outcomes and building on platforms that can sustain it.

That’s the shift toward becoming AI-smart. It means treating AI as a secure, resilient service. It means securing and governing data consistently across datacenters, clouds, and the sovereign edge. And it means simplifying operations so teams can run AI and traditional applications together without fragmentation or complexity.

At the same time, data gravity is pushing outward, accelerating the rise of the sovereign edge as a first-class element of enterprise architecture. Platform choices are becoming strategic decisions with long-term consequences. And the push for resiliency, consistency, and choice is reshaping how IT teams think about modernization.

2025 put AI in the spotlight. 2026 will decide who owns the stage, and who gets left behind.

©2025 Nutanix, Inc. All rights reserved. Nutanix, the Nutanix logo and all Nutanix product and service names mentioned are registered trademarks or trademarks of Nutanix, Inc. in the United States and other countries. Kubernetes and the Kubernetes logo are registered trademark of The Linux Foundation in the United States and other countries. All other brand names mentioned are for identification purposes only and may be the trademarks of their respective holder(s).