IBM Power: Doing Hybrid by Choice

This month, IBM had a briefing on its Power platform for AI, which currently is in use by over 40K users. This may not seem like a lot, but with AI failures outnumbering AI successes, it is important to look at firms that have delivered successful AI implementations, and IBM has been aggressively working and demonstrating AI for several decades.

Currently, most AI is run in the cloud, with a tiny but increasing amount running on the edge and a larger portion moving to on-premises. For most of these implementations, the issue is that they are inefficiently hybrid because this model was adopted by plan, not by necessity. In other words, the implementations’ hosting preferences drove the decisions, not the needs of the enterprise.

IBM wants to change that by providing an increasing number of solutions that optimize AI so that it runs in the most efficient, reliable, and productive hybrid environment dictated by the needs of the business, not the limitations of any individual AI provider.

The AI Hybrid Problem

The issue we have at the moment is that most AI is run in the cloud because that is where you can get the easiest access to the highest variety of AI models. The cloud isn’t the most economical, reliable, or highest-performance place to run AI because it is a remote shared service, not something that IT has full control over. As we increasingly depend on AI to run our businesses, there is also an increasing need to control the infrastructure these AI instances run on so that we know it’s secure, that our confidential and proprietary information isn’t leaking, and that we can better assure critical service levels.

This isn’t to say the cloud is bad. It does have advantages like a shared cost model, cloud data centers tend to be able to better handle most weather events, and in the future, most may be nuclear powered. But today, they’re still plagued by enough unknowns in terms of security, exposure, staffing, power, and latency that they aren’t ideal for efforts that are tied to real-time, automated decision-making.

For instance, if an AI is monitoring and running operations and the cloud goes down, it takes out any chance of automated mitigation because the AI will fail right along with the service. If it were on-premise or fully hybrid (with failover), it would be able to respond to the outage and limit the impact on end users.

Why Choose IBM Power for AI

The advantage that Power has over x86 for AI is that it tends to be less used, so the number of potential attackers who understand the platform is reduced. It is a very different architecture, which means much of the existing malware may not run on it well, if at all. It comes from IBM, which has been far more security and ethics-focused than firms that came up after it.

This is why banks and many healthcare providers

prefer IBM. If security, reliability, and availability are extremely high priorities, then IBM is typically at the top of the bidder list. It is through IBM’s comprehensive focus on infrastructure and excellence that it has been able to roll out some of the most secure and efficient platforms currently in the market, and it remains one of the few companies still able to provide mainframe solutions that continue to excel in those three critical areas.

So, it just makes sense to look at IBM for AI, given its advantages and experience in this segment and the ability to help design solutions that are less ad hoc and more tailored to best utilize the hybrid requirements that surround them.

Wrapping Up

IBM will never be the cheapest, but when considering AI, focusing on the cheapest providers, many of which are having a very hard time understanding, let alone deploying AI, it is best to focus, particularly in these early years, on companies that are subject matter experts. With Power and watsonx, IBM is one of the most experienced AI providers currently in the market, and it partners heavily with NVIDIA, making IBM one of the better choices when looking to deploy hybrid AI.