Kenso Trabing - The AI Compute Crisis & Web3 Opportunity - TOKEN2049 Dubai 2025

Well, it's no secret that the AI industry is booming, but what a lot of people don't talk about basically is the energy opportunity. So AI is using a ton of water, but it's also using a ton of electricity. What's really interesting here is that 65 to 90% of the opex of an AI workload is actually directly related to the price of electricity. So what's interesting about what we're doing is basically related to the cost side of a lot of our customers. What I mean by that is this.

According to a recent survey by Intel, 56% of AI practitioners say that cost is the number one challenge related to implementing AI solutions. What we do is focus on a vertically integrated approach from the substation to the application layer. That's possible because there's three layers of commodification occurring in the industry right now. What I mean by that is this. There's commodification occurring at the software layer.

There's commodification occurring at the hardware layer and electricity is already a commodity. So drilling down on the software layer at the top of the stack. Basically, what you have going on is open-source models converging with closed source models when it comes to performance. Further down the stack, when it comes to the hardware layer, what's really interesting is that anyone can basically go out and buy an Nvidia GPU. And what's interesting about that is obviously the firmware side.

They've got a huge edge with CUDA and the name of the game in the hardware space is Nvidia right now. And then at the bottom of the stack, obviously electricity is already a commodity. That's part of the reason we've positioned ourselves in Paraguay, a small landlock country in between Argentina and Brazil because the electricity is very cheap there. And not only is it affordable, it's 100% renewable. So something we've also done in Paraguay is gone ahead and partnered with the government to go ahead and basically implement a breast cancer detection model partnering with their Ministry of Health and their Ministry of Technology.

That means that our servers are actually going to be in a government data center going ahead and solving real world problems that have a really high net benefit for society. What's cool about what we're doing too is a lot of companies that understand the value of not just affordable compute but renewably generated compute um is basically Nike the Nike the enterprise the uh exclusive representative of Nike in Parlay is now going to start working with us when it comes to um going ahead and implementing renewably generated AI agents to basically conduct predictive analytics on um their suite of products. Another company that's going ahead and basically using us uh is Honda. And what's cool about that also is again that it's cheap, it's renewable, and um these are really high quality AI agents that are basically being integrated with uh their enterprise resource management systems. Lastly, the way we plan to grow is basically using linear returns to bootstrap our geometric returns.

What I mean by that is this. If you have a wholesale electricity contract, it's really hard to balance your resources. So, you're using all the electricity, say for a particular purpose. So, what we do when we're not using all of our electricity, believe it or not, is we mine Bitcoin. So, what I mean by that is this.

The way we plan to grow this side of the business is basically by selling Bitcoin futures. And what that enables us to do is get cash upfront to go ahead and grow that side of the business while bootstrapping the AI side of the business. I'm going to keep this presentation really short because it's really hot out here. So um this is the end of the talk but looking forward to going ahead hearing all of your feedback on social media, Telegram and beyond and um looking forward to biddling the future together. Thank you very much.