The artificial intelligence boom is real, but credit investors say there's more to AI trading than just chips.
“This is a super-sized microcycle that lasts for many investors' careers,” said Diameter Capital Partners co-founder and managing partner Scott Goodwin, who quoted his partner Jonathan Lewinson as saying.
AI represents what Diameter Capital sees as a long-term disruptive cycle, but buying the most obvious winners is not the only way to do it, he said on the “Goldman Sachs Exchanges” podcast published Friday.
Diameter Capital, which manages about $25 billion in assets, has focused on where AI demand can cause less obvious bottlenecks, and where those bottlenecks appear in credit markets.
AI opportunities beyond chips
With this in mind, Diameter will purchase unsecured bonds from mid-sized telecommunications companies in 2023.
Goodwin said the bet is rooted in the idea that as companies move from training AI models to actually using them, demand will shift from individual chips to networks that carry data.
“It had to come out of the data center. How would it get out? It would go out on commercial fiber optics and pipes,” he said.
Goodwin said the carrier has signed more than $10 billion in contracts with hyperscale cloud providers, and its debt has been restored to par.
Diameter Capital also made a “big bet” on satellite companies related to radio spectrum. The bet later paid off when the company sold its spectrum assets and the debt returned to par.
Goodwin's comments come amid growing debate about whether AI's inflated valuations are sustainable and whether investors are missing out on other opportunities related to the technology.
risk and reward
Mr Goodwin warned that some of the AI credit boom may carry risks that are difficult to price, particularly in chip finance.
He said some investors are taking on “residual risk,” or the riskiest part of a chip financing deal, betting on how much the hardware will be worth in a few years. Cutting-edge companies frequently update their technology, which can quickly make chips obsolete for some customers.
“We're going to call some really smart people in Silicon Valley, we're going to call some really smart people in Big Tech companies, and we're going to ask them what the residual value of these chips is going to be in three, four, five, six, seven years.” “None of them have a clue.”
Goodwin said the next step is not just spending on infrastructure, but destroying competition rather than capital spending.
“Who will be the companies that adopt AI and take a step forward against their competitors? And who will be the losers?” he asked.
“This is really interesting because it’s actually a longer cycle than the capex cycle,” he said.
