Artificial intelligence is ushering in a new world, with AI-powered intelligent systems driving business operations. Digital infrastructure must evolve to support this new world, so much so that fundamental IT concepts, including the cloud, will have to be rethought in light of emerging technologies, use cases, and business requirements.
Today it is common to have several cloud instances and multiple management consoles whose use depends on where a workload is running. And cloud use is nearly ubiquitous, with most organizations comfortably keeping even sensitive data in their cloud instances.
Now, the prospect of AI-driven operations is pressing organizations to re-evaluate IT environments, and several aspects of traditional cloud computing will evolve. Three major changes are these:
- Abandoning the use of unique, disparate cloud instances and adopting a single cloud operating model
- Moving beyond traditional cloud security models to a framework that also mitigates risk and supports compliance and governance issues
- Achieving new levels of IT operational efficiency
The single cloud operating model: This may be the biggest change and will require a rigorous commitment. AI applications require a cloud operating model that offers consistency and interoperability between key elements and management tools. The goals are to have a cohesive view from the edge to cloud to data center in an effort to reduce unicorn cloud instances or other outliers that get set up separately and require special attention. A single cloud operating model will leverage a broad set of capabilities and functional elements. One benefit of this enhanced infrastructure will be less complexity in the cloud control plane. It is important for the cloud operating model to apply to both public cloud and on-premises data centers so they can be operated without barriers, creating a true hybrid environment. This unified infrastructure will become the status quo as more computing returns to the data center due to privacy, compliance, and legal demands. Organizations will also find that meeting the performance and capacity demands of the AI cloud will require leveraging the latest and most capable technologies from AI leaders such as NVIDIA.
Data protection and security: With the AI cloud model, data protection and security must evolve from simply protecting against unauthorized access and theft to boosting protections so that all compliance, risk, and legal mandates are fully satisfied. While AI carries the promise of deriving hidden insights from an organization’s data, it also has the potential for exposing that data in new and alarming ways. Without adding effective guardrails to keep the organization within statutory requirements, AI will not be secure enough to be deployed for many new use cases that require the integration of especially sensitive data. Solving this problem will require cooperation between experts from both IT and the business side of the organization, who in effect will be charting new territory where off-the-shelf standards or frameworks will prove inadequate to the task. One solution that seems likely to be adopted by many organizations is to run the most sensitive AI apps on premises, in a private cloud environment, to ensure there is no data leakage to public generative AI instances or large language models (LLMs).
IT operational efficiency: To function effectively, this new operational model that organizations adopt must place AIOps as the driving principle. Human-centered cloud operating models naturally tend to focus on individual workloads, not the entire infrastructure. Having a single cloud operating model is the first step in consolidating workloads, making it possible to leverage AI to dramatically improve efficiency and responsiveness of IT operations and staff. AIOps will allow IT teams to offload many of their mundane and repetitive tasks, something that is of great value to overloaded teams. And AIOps will make more efficient use of the resources in the IT estate, always choosing the right infrastructure for the right workloads and never forgetting that performance and resource demands must be balanced with compliance, regulatory, security, and other considerations.
Hybrid Cloud Economics Demand a New Pricing Model
AI will also create new cloud economics. Innovative options for financing will make it possible to apply a usage-based pricing model to all infrastructure, both on premises and in hyper-scalers. This will support a true hybrid cloud approach where cost is less of a determinant of where resources should be utilized. Data mobility and the need to stream data across infrastructure must also be key components of AI cloud planning, including both the technical issues of capacity, performance, and throughput, in addition to the cost of data egress and ingress. A new, flexible pricing model will be essential because new apps or LLMs may require moving data in ways that cannot be foreseen.
AI will bring significant changes to the cloud we know today. This new world is still emerging, and it is not possible to provide all the details that need to be considered in an AI cloud strategy, but the key trends discussed above demand a different kind of scalability and a totally new approach to workload services. And, because so much is unknown, it is important to maintain flexibility to support new AI use cases as they emerge. It is time to re-evaluate the legacy cloud and start making plans to better position IT and the business for the new world of AI.
