He Turned Ohio’s Biggest Paper Mill Into an AI Data Center | Andrew Keys

Andrew Keys (0:00)

We are expending in the United States ten times more in 36 months than it took America to build the United States Highway systems in 35 years.

Amanda Cassatt (0:11)

And that's inflation adjusted?

Andrew Keys (0:13)

We've actually converted one of the largest paper mills in America to become a data center.

Amanda Cassatt (0:19)

I'm Amanda Cassatt, and you're watching Endgame. The data center business makes total sense to me. What does this have to do with Ethereum?

Andrew Keys (0:26)

These machines can't have bank accounts. They're not LLCs. They're not human beings. AI is the production half that will continuously produce intelligence output, and Ethereum is the coordination half. Settlement, identity, payments, proofs. These are two halves of the same coin.

Amanda Cassatt (0:48)

My guest, Andrew Keys, is known for being early to Ethereum. Now he's on to a new prediction with a fresh thesis, how Ethereum powers the machine economy. It's a great episode on where crypto and AI are actually converging, and the surprising places he believes value will accrue. I'm your host, Amanda Cassatt, and you're watching Endgame. Don't forget to like, subscribe and comment. Where I want to start is, you were one of the earliest and most successful investors in ETH. How did you find ETH so early?

What was that story, and what were the characteristics that attracted you to it when it was 30 cents?

Andrew Keys (1:27)

I came to Ethereum via another business where simply put, I dealt with payment systems for medical records.

Very quickly, if you went to the doctor and gave your insurance card to the person at the front desk, all of your health records would be sent to Chennai, India, where a dealer would fight with Blue Cross Blue Shield in India, and eventually Dr. Jones got paid. And our software business would essentially process that revenue cycle management. So I learned pretty well the problems of legacy payment systems at the same time Bitcoin happened. But I always thought, great, Alice could send Bob value without the bank in the middle. But there had to be more sophisticated processes. You know, if something happens, then send the payment else failure. Those if-then-else statements were poorly phrased as smart contracts. And basically, that was kind of the first iteration that I heard of this kind of peer-to-peer payment technology that you can embed logic. And I literally went to the first ever Ethereum meetup and met Joe Lubin before Ethereum started. And essentially, he was talking about creating a crypto alphabet for an industry that hadn't yet developed on a software that hadn't yet launched, that had no applications, no developer tools. But it was kind of a grandiose idea. And there was enough pattern recognition where you could see that Bitcoin grew in tremendous value. I think at that time, it was $100. And I said, well, it could basically have this thing that was $100.

But you could basically programmatically meter every type of transaction on a peer-to-peer basis. That's probably worth more than the version of this, like the 1 where it's just Alice sending Bob value. And so I took what little I had and tried to make a little bit more of it.

Amanda Cassatt (3:31)

Okay, so let's zoom to the present.

You have written this piece, super interesting. I want to go into the thesis. Does the machine economy actually need Ethereum? You make this strong case that machines need payment and verification. But what proves that those rails need to be Ethereum?

Andrew Keys (3:47)

So I think first and foremost, AI cannot produce identity, reputation, proofs. But AI is great at producing, let's just call it output. We see these infinite images, memes, videos that are being created, content, ad nauseum. But we're not able to necessarily reconcile the provenance of any of those assets.

And I think that Ethereum is best poised to give a cryptographic attestation, a payment rail, identity, or reputation, mainly because it is the most, it's in the most hegemonic position now. You know, in terms of liquidity for payments, I mean, 50% of all stable coins happen on Ethereum layer one, you know, 25% happen on layer twos that sync to Ethereum, and then everybody else is the other 25%. So I think that the agents are naturally going to go where there's already liquidity.

Amanda Cassatt (4:48)

That makes sense. Are you seeing that happening today? How are you seeing the...

Yeah, are they making that choice now? Are you expecting them to do something different than what they're doing?

Andrew Keys (4:58)

I think it's early days.

You know, I think, you know, I outlined like a concept of standards, and for any of the OG Ethereum people, they could remember the EEA that created, you know, the standard for a permission flavor of Ethereum. If you want to rewind further, when Java became J2EE, Java 2 Enterprise Edition, it's when it had clean web APIs and clean database APIs. And so if you were an engineer in Mumbai or Paris or San Francisco, you had the same APIs. And that's what kind of permeated Java as the most used software language. And with this, I think that you can see early standards. One, which is X402, that was created by Coinbase, and now is primarily driven by CloudFlare.