With over 30 years of experience in enterprise technology, David Linthicum is a globally recognized thought leader, innovator and influencer in cloud computing, AI and cybersecurity. He has authored over 17 best-selling books, 7,000 articles and over 50 courses on LinkedIn Learning. He’s a frequent keynote speaker, podcast host and media contributor on digital transformation, cloud architecture, AI and cloud security.
I spoke with Linthicum about how enterprises are, in some cases, struggling to find the ideal platform for their AI software, and the challenges companies are facing as they use traditional cloud providers for this role. We also spoke about agentic AI, language models, and cloud-based generative AI.
Key Points: AI and the Enterprise Cloud
Selected Quotes: “I don’t know why we need 36 LLMs out there”
Linthicum’s comments about top cloud providers, AI, and the evolution of LLMs are informed by his long experience as a cloud and AI consultant.
So how do you see the big three cloud players (Microsoft, AWS and Google) falling short in terms of AI workloads?
“So you would think they would have kind of a lock on that industry people who are looking for AI systems, but enterprises are finding a couple of things they need to maintain control of their information for compliance issues they’re dealing with. They’re not quite ready to trust the big public cloud providers to make that happen.
“But the big thing is cost. Some of these AI systems, it’s massive amounts of GPUs that you need for this stuff. It’s going to be $10,000 a day that enterprises are going to have to pay to run this and they’re running the numbers—and I’m working with a few of them as well as a consultant—and the numbers just aren’t working out. It’s too expensive.
“And also they didn’t view the scalability. And also if you look at the core features and functions of some of the big cloud platforms, big cloud providers, they’re not really optimized for AI workloads.
“They’re optimized for their own stuff, they’re optimized for typical applications, but they haven’t optimized for the way that AI uses storage and processors and things like that. Everything’s very tightly coupled.”
I remember in the early days of cloud, or even past the very first days, there was sort of a race to the bottom in terms of cost because each cloud provider wanted to grab market share in these really important days. So the fact that they’re high priced now for AI workloads suggests they haven’t started to play that game yet in the AI era
“You got it right, and there was a race to the bottom of the early days of cloud. You got to remember that was 2008 to 2013 when that whole airline price war kind of took off. And I think what happened now is they got 15 years into it, they have many billions of dollars of infrastructure that they made and they can’t afford to take a loss on the infrastructure like some of the startups can.
“So they’re putting a price point out that they view as fair and probably as something in the market market will accept, and it’s just way more expensive than some of the other alternatives that you have out there. And the reason is they provide a really good advanced service. Public cloud providers should be given their due.
“However, the cost of those services is out of reach for many of the enterprises out there and they also don’t need 10,000 services. They need a couple of services to support their AI workloads. They find it’s easy to move data around these various platforms fairly easy than it was 10 years ago where that wasn’t the case 10 years ago. So now you’re facing alternatives.”
A lot of companies aren’t actually building generative AI models. They’re accessing generative AI using an API. So they don’t have model training costs, they just have usage costs. Given that, generative AI can be fairly cost effective given how productive it is. Agree, disagree?
I agree. And it is basically almost free. You can use generative AI systems or the APIs for free or a couple of pennies a month, whatever you want to pay for it. And most of the models out there, open source stuff, you can run them on your computer. So we’ve kind of reached the saturation point where everybody knows how to build it.
“The issue with generative AI is going to be a couple of things. Number one, we’ve run out of past of historic data to train the thing with, we’re generating data all the time on the open internet, but we’re not going back 40, 50 years in training the AI systems on everything that’s out there. So it’s going to have a limited amount of value and a diminishing amount of value moving forward.
“I don’t know why we need 36 LLMs out there and I get a news release every week and these things cost $150 million to build and run the power of a small city, but they’re going to have a diminishing value. And I think that people need to understand how it’s going to be used in the context of a business problem.
“One of the things that enterprises are doing right now, they’re running around in a circle: ‘Well, we have generative AI, what do we do?’ We know it’s going to be a game changer for the business, but what business problems do we apply it on? It hasn’t proven itself in that context yet. And so they’re falling back and, I think for good reason, onto small language models and agentic-AI based deployments, so then they can take or leave generative AI. That’s part of it.
“So generative AI as a model is a huge step forward in terms of where things are going to go. However, our ability to find business value out of that, I think, is going to be a struggle for at least the next several years. So it hasn’t proven itself yet.”
Do you think that the rise of agentic AI is going to be the beginning of major job losses for knowledge workers?
“I’m always a little skeptical about when people say that there’s going to be displacement. I think there’s going to be change and I think people are going to adapt and adopt to different tools and technology. We may see a shift in people moving from one career path to another career path, but I think on the whole it’s going to be a positive trend for information workers because they’re going to be able to do things in a much faster, much better way.
“You’ve got to remember, it’s not the fact that we’re running out of work. We have plenty of work to do it, we just don’t have enough people to do it. And so we’re able to automate them and use ag agentic AI or AI as a force multiplier. I think that’s going to bring value to them. So I’m not as scared of it as other people are. There is going to be some changes, very much like the Industrial Revolution in the mid 1800’s when we moved from a farm-based economy where everything was manual into something that was automated.
“We adapt to it, we work it, we make it better. We leverage that technology to take us to the next level. So I think I’m a little bit more positive in terms of its impact on the market and employers.”
