AI agents are not new. AI itself was old, but it is new again, and AI agents and agent AI are the new chapter. Known for its autonomous decision-making and complex goal management, agent AI will fundamentally transform how enterprises operate. A key change is that public cloud providers are often not the preferred platform choice. Large and small enterprises are turning to smaller, decentralized platforms such as on-premise hardware servers and smaller devices. Let’s explore what is driving this transition and its future impact on enterprise AI.
Agents are on the rise
Agent AI refers to artificial intelligence systems that have autonomous decision-making capabilities and can act independently to achieve specific goals. We’ve seen many examples of this idea over the years, most recently in personal digital assistants on our phones and devices, and the increasing automation of everything from home HVAC systems to automobiles.
These systems have advanced reasoning, learning, and adaptation capabilities and can process complex information, make informed choices, and perform tasks without constant human supervision. Using sophisticated algorithms and vast datasets, Agent AI can analyze environments, predict outcomes, and initiate real-time actions. This form of AI aims to increase efficiency and effectiveness by providing intelligent, goal-oriented solutions across a variety of domains, including healthcare, finance, and transportation.
Until now, public cloud services such as AWS, Microsoft Azure, Google Cloud, etc. have dominated the cloud environment. However, the unique demands of agent AI are causing enterprises to rethink and ultimately migrate away from public cloud solutions for several reasons:
Data sovereignty and security are crucial, especially in regulated industries like finance, healthcare, and government, where a private cloud or on-premise servers provide greater control over how data is processed and stored, ensuring compliance and reducing the risks associated with data breaches.
Agent AI applications often require a high degree of customization and optimization. In a non-public cloud environment, the infrastructure can be fine-tuned to meet specific requirements and operate more efficiently. Enterprises enjoy better performance and resource management compared to standardized services in the public cloud.
The cost structure of public cloud services can be unpredictable and prohibitive. Many businesses are still struggling with post-pandemic cloud bills that have doubled or tripled what they expected. Things have gotten so bad that an entirely new field called finops has been born. By moving to their own hardware servers and small, purpose-built devices, businesses can gain more predictability and control over their costs, avoiding the ongoing subscription fees and variable costs inherent in the public cloud.
Agent-based AI means smaller, non-cloud systems
Edge computing (running AI at the edge of the network on small devices) is gaining traction as part of this shift. Edge computing processes data locally, reducing latency and improving security; sensitive data is kept closer to the source. For example, automakers are deploying edge AI to process real-time vehicle data to improve performance and safety without relying solely on cloud connectivity.
Decoupled distributed systems running AI agents require hundreds of low-power processors that must run independently. Cloud computing is not typically suited for this, but it is possible to be nodes in these distributed AI agents running in heterogeneous complex deployments outside of public cloud solutions.
The continued maturation of agent AI will drive further migration away from the public cloud. Enterprises will increasingly invest in purpose-built hardware, from intelligent IoT devices to advanced on-premise servers, tailored for specific AI tasks. This transition will require a robust integration framework to ensure seamless interaction between diverse systems and optimize AI operations across the board.
Yes, this adds complexity, heterogeneity, and operational costs, but it also leads to more practical AI deployments aligned with business needs. Let’s face it, companies will never build large language models for their own use – it’s too expensive, even on public cloud providers. Agents and small language models are a more viable architectural option for AI.
In this case, doing what’s best for the enterprise is not good for the public cloud providers. But there’s no need to cry for them. The cloud is not the most optimized or cost-effective option, but it is still the “easy button” for AI. That makes it attractive for many AI builds and deployments, and many enterprises will choose that path.
The integration of agent AI marks a key shift in enterprise strategy, moving companies away from public cloud solutions. By adopting non-public cloud technologies and investing in adaptable, secure, and cost-effective infrastructure, enterprises can realize the full potential of agent AI. This strategic shift will drive operational efficiencies and more closely align AI adoption with the specific needs and goals of the business.
