Two Truths and a Lie: AI Planning Advice for Infrastructure Strategists

Generative AI has exploded the imagination of CEOs and employees alike, representing a technological leap unlike any we've seen in recent years. While technology has certainly seen its share of fads-think Surround Sound, 3D TV, or virtual reality--these trends quickly faded from the corporate planning budgets when reality didn't meet hype and users lost interest.

Generative AI is different. The rise of tools like ChatGPT showed that AI benefits are now accessible to everyone, not just data scientists. The rapid proliferation of AI promises a healthy ROI for internal productivity gains and for stickier customer relationships.

So now we have infrastructure owners scrambling to plan for the onslaught of AI applications. In fact, a recent study revealed that 99% of surveyed customers plan to upgrade their AI applications or infrastructure.¹

Based on our interactions with hundreds of enterprises, here is some initial advice on how to plan responsibly for your AI-powered future.

An AI Truth - AI will stretch your infrastructure…and your infra teams.

AI requires new hardware and software that are beyond what most enterprises are using today.

For Hardware: Innovations in GPUs and CPUs, along with supply constraints, means access to these chip sets is now on the critical path to AI innovation… and can drive organizations to look beyond their traditional domains for this - whether on-premises, at the edge, or in the cloud.

Conclusion: AI planners should develop AI apps on a platform that can access GPUs on-premises and in the public cloud - ideally with a single operating model.

For Software: Most AI models are containerized while most datacenter teams are familiar with virtual machines. Running AI in a private datacenter helps protect data and gives a transparent pricing model - but container skills will be essential.

Conclusion: Infrastructure owners should look at how to converge management for containers and virtual machines so that a single team can manage both at scale.

An AI Truth - AI apps require the same enterprise value you deliver today.

Enterprise infrastructure owners cut their teeth delivering "enterprise value," namely resiliency, Day 2 operations, and security for business-critical apps. AI apps are no different and this offers a career-path accelerator for infrastructure teams.

Conclusion: AI apps will quickly become mission-critical once internal productivity gains and external customer value gets into production. Pick a platform that delivers resiliency, seamless hardware and software updates, and fast remediation for security risks-consistent with your current offerings for business-critical apps today.

An AI Lie - Infrastructure will make your AI successful.

While infrastructure plays a key role in AI success, infrastructure alone is not enough. In fact, we can quickly determine who will be successful with AI - successful companies identify a business challenge, quantify the opportunity, and drive to an ROI-driven outcome.

Conclusion: Don't treat AI as a science project. Instead, find a real use case, define the success metrics, pick the right AI tools, e.g. large language models (LLMs), chatbots etc., validate the use case quickly, and plan for production of this workload.

Make sure your teams consider security, data privacy, and AI governance from the get-go. Determine how your existing team can extend their skills into GPUs, containers, and LLMs. It's good for your business and for your team's career development.

At Nutanix, we're excited to bring our enterprise infrastructure value to AI apps and help teams take on new hardware and software challenges. Bring your AI vision to us and we can help you deliver an AI truth.