As artificial intelligence drives new infrastructure demands, the conversation is expanding beyond large, centralized data centers. Adam Karon, COO and GM of Akamai's Cloud Technology Group, explains why the rise of interactive AI agents will require a more distributed approach to computing. While centralized infrastructure remains essential for large-scale training and batch processing, AI agents increasingly need to access data and systems across multiple locations while responding to users in real time. With approximately 4,000 points of presence globally, Akamai is positioning its distributed infrastructure to bring compute closer to users and applications, helping reduce latency and support emerging AI workloads.
Karon also discusses where businesses are beginning to see measurable value from AI, including responsive gaming characters, retail applications, real-time dubbing and media transcoding. However, he cautions that organizations must first prepare their data, systems and processes before expecting AI investments to deliver meaningful results. Beyond performance, Karon highlights the growing importance of data sovereignty and portability, encouraging businesses to minimize exposure, carefully select operating regions and avoid becoming locked into individual cloud providers. As AI adoption accelerates, the ability to run workloads in the right locations while maintaining flexibility and control could become a key differentiator.
Steve Saunders:
Hey, Adam, how are you?
Adam Karon:
I'm great, Steve. Great to be here.
Steve Saunders:
Well, I'm excited to have you on the show because Akamai is a company that I've been dealing with for a long time and has seen a lot of changes in its strategy. And I think right now is a particularly exciting time to be there because we're entering this extraordinary period of innovation and AI-enabled infrastructure, which is really driving this new digital economy. I do wonder, however, whether, particularly in the US, we've over-focused on building data centers and to the detriment of the network, which must surely connect them and connect the AI out into the rest of the world. Do you think that's a problem or what are your thoughts on that?
Adam Karon:
Well, I think that the large data centers are needed. They're needed for certain things that operate better in those big centralized locations, whether they're in the US or globally. Things like batch jobs or heavy training, like for a large frontier lab model. But I think you're right in that, as the agents themselves become the new users on the internet, and I think that's the transition that we're in, those interactive agents are going to be able to pull data from many different locations and they're going to action dozens of loops or round trips on every task that they're going to do. And for that, you're going to need to have your agents located much closer to the user that they're interacting with and to the data and systems that they interact with. So I don't know that it's a cloud versus edge more than just making sure that the right type of application runs in the right place at the right time. And sometimes that's central and sometimes that's distributed.
Steve Saunders:
Where's the money? AI is moving so quickly. The business case in some ways doesn't seem to have caught up with the technology. Where do you see people making money from AI? What are the actual business cases as opposed to pilot cases that you are seeing taking off?
Adam Karon:
Yeah, I mean we've seen it in a lot of places. Gaming studios has been a place we've seen it where people are generating simulated AI characters that can respond super fast, in milliseconds. Retailers are using AI at the point of sale or maybe with staff augmentation. Broadcaster is probably one of the first places, things like transcoding or realtime dubbing have been places we've seen the pilots themselves go to production, tremendous value gets created. I think one of the barriers that you see a lot is people try to layer AI on top of old systems and old data, and that doesn't work really well. You've got to make your systems and your data AI-ready for you to take advantage of it. So once the data's ready and once the processes are ready, then becomes the question of how do I run it the fastest way and the cheapest way possible, and measure the results of that per outcome.
Steve Saunders:
When you're talking to customers, what's the key benefit from an Akamai service versus something from a cloud hyperscaler? Where are you really differentiated?
Adam Karon:
We want to operate in the places, I think, where the hyperscalers aren't necessarily naturally. The hyperscalers aren't... While they have distributed capabilities, that's not their core of their business. And I think we operate in 4,000 pops all over the globe and we're able to distribute the compute close to end users. So for applications that fit that or for agents that really need to operate in many thousands of locations in every geography possible, Akamai really has a natural fit there as compared to other providers in the market.
Steve Saunders:
I've been hearing a lot about sovereignty this year from everybody, everybody I talk to. And actually sovereignty is more than one thing depending on who you're talking to. Where do you come down on this one? Is that an Akamai responsibility? And do you have advice for your customers such as factories or hospitals or something? How are they supposed to reconcile the proprietary nature of their business traveling over say a public cloud? And what can you offer them which is an improvement over that?
Adam Karon:
I think what customers really need to do is try to reduce the exposure that they have as much as possible. That would be my advice to them. And they need to choose the regions that they operate in on whatever provider they're using. They need to make sure that their data is portable and that their application's portable. So they're not locked in and their data's freely movable without being able to be blocked. I think it's the best thing that anybody can do at this point.
Steve Saunders:
It's such a pleasure to talk to you and congratulations on everything at Akamai. I'm really enjoying following the story.
Adam Karon:
No, it's a really exciting time and I appreciate you spending time with me to talk about it.