Learn from past cloud-first mistakes for better AI
The cloud computing revolution brought many innovations, but it also taught the pitfalls of rapidly adopting new technologies without a well-thought-out strategy. Today, IT leaders are at a similar crossroads with the rise of generative AI. As companies build their AI factories in this new era, IT leaders have an opportunity to learn from their cloud-first sins of the past and build strategically in a way that prioritizes security, governance, and cost-effectiveness for the long term, avoiding mistakes that will need to be corrected later.
The Cloud Retrospective: A Learning Curve
The rise of cloud computing a decade ago ushered in shadow IT, where teams and even individual employees swiped a credit card to gain instant access to vast compute and storage resources. Over time, many organizations struggled with cost, security, and governance issues that led them to rethink the underlying model. Today, organizations are at risk of falling into a similar scenario known as shadow AI, where teams turn to public clouds or API service providers in their rush to develop or adopt AI solutions. If these initiatives are not properly monitored, the cost of running GenAI services in the public cloud could quickly spiral out of control, causing a number of problems in the long run.
Applying lessons from shadow IT to generative AI
As organizations build their AI strategies, the lessons learned from the cloud era can be particularly valuable. Here are some insights to consider:
Build cost-efficiently now
Long before today’s acceleration of AI, a shift was already underway as companies began to rethink the cloud operating model. A 2021 article by Andreessen Horowitz estimated a $100 billion market value gap among the 50 largest publicly traded software companies investing in the cloud. They attributed the effects of Public cloud spending.
GenAI offers great potential – McKinsey estimates that generative AI could create annual added value of $2.6 to $4.4 trillion(1)– but that also comes at a price. Instead, companies can focus on efficiency from the ground up, unlike in the early days of cloud adoption, when long-term costs were often overlooked.
A recent study by Enterprise Strategy Group found that running an open source Large Language Model (LLM) on-premises with Retrieval-Augmented Generation (RAG) was 38 to 75 percent less costly than in the public cloud, and 88 percent less costly with the API-based approach.(2)Armed with this knowledge, leaders can build from the ground up for long-term success, rather than achieving short-term gains that require course corrections.
Prioritize an on-premise first strategy that brings AI to your data
Cost is just one consideration in an increasingly AI-driven world. AI is only as valuable as the data it is linked to. But where is that data? Gartner predicts that by 2025, 75% of the data generated by enterprises will be created and processed outside of a traditional central data center or cloud.(3)This means that companies must give the security and sovereignty of their data the highest priority, secure access and management, and increase transparency.
Given the lessons learned from previous cloud computing experiences, organizations should consider whether an on-premises first strategy or integrating AI with data makes the most sense in an AI world, especially for applications where control over data and compliance are critical. This approach allows organizations to build on a strong foundation of existing infrastructure and strategically deploy AI services in the public cloud where it makes sense. In addition, organizations can sidestep issues such as data gravity, avoiding the need to rethink strategies and the challenges associated with re-engineering in a few years.
Continuously evaluate, learn and adapt
GenAI, like all areas of technology, will evolve. What works today may not be your ideal strategy tomorrow. But the goal should be to build a foundation that gives you the most flexibility and reusability while giving you complete control over your data, infrastructure and management. This may also mean working with the right partners who can advise you on technological advances and help you adapt strategies on the fly.
Sunny prospects in the age of AI
The move to GenAI is an opportunity to apply the hard-earned lessons learned from the cloud computing era. However, by making cost efficiency a priority from the start, considering an on-premise first approach, and continuously adapting and learning throughout the process, IT organizations can use GenAI in a way that is both innovative and sustainable. This balanced approach will help avoid past mistakes and lay a solid foundation for future innovation.
Learn more about the Dell AI Factory in this Webinars.
(1) The Economic Potential of Generative AI: The Next Productivity Frontier, McKinsey
(2) Maximizing AI ROI: On-premises inferencing with Dell technologies can be 75% more cost-effective than public cloud, Enterprise Strategy Group, April 2024
(3) What Edge Computing Means for Infrastructure and Operations Managers, Gartner
