fungible thoughts

Artificial Intelligence

TLDR: room to run; applications vs models; changing work force & education

The changes we have yet to see in AI far outweigh the changes we have seen since the “dawn” of AI. The pace of innovation has been eyewatering, with a new groundbreaking model being shipped to market seemingly every month. The pace of innovation is accelerating but culture is slower to change than science. We’ll begin to see the larger effects of AI models once the adoption of AI shows lower barriers to entry.

Leading AI companies are throwing money at developing new models and rushing to create a market for application of AI models, all while bleeding capital. NVIDIA runs at a 75% gross margin and are still on pace to grow net income by 10% this year. Companies are willing to pay just about any price to get their hands on hardware to train and use models.

The cost of saying please and thank you is apparently in the tens of millions of dollars, as per Sam Altman.

Here’s the bottom line business case: massive amounts of capital being thrown at widespread adoption of AI, accelerating pace of innovation, minimal use case viability and implementation. We’re at the very beginning, the first batter of the first inning, of AI implementation into society.

What’s different from the race for search dominance, which saw AltaVista, AskJeeves, and Yahoo overthrown is the pocketbooks of these companies is absolutely cavernous. As of Q1 2025, Apple has over $100B in cash on its books – IN CASH!!

In the battle for search, we saw start-ups become behemoths. The current landscape means incumbents are investing to play defense as well as offense, as demonstrated by the massive amounts of capex being thrown at AI. I’m skeptical of the case that some unknown will be come a beneficiary of the “winner takes all” model, which we saw in the tech booms of the 00’s. The incumbents seem primed to win the AI race given the capital requirements around creating and maintaining AGI models and databasing/warehousing.

Barring some cataclysmic event, say deleting the entirety of the Social Security Administration’s databases, widespread adoption of AI use cases seems inevitable. AI code applications being able to self replicate and optimize will happen well before the regulation banning such activities will be enacted. If you think the hype is overblown, you’re in a luddite camp. It’s coming whether you like it or not.

What’s not set in stone is how culture will be impacted. We’re adaptable, but the changes required to build a functioning society in parallel with AI are massive. We’re looking at changes to how people work and the areas in which humans can complement AI. Education will change massively. In universities, AI writes the test questions for professors and AI answers the questions for students. We’ll need a whole new system for assessing competence and ability and my chips are on private industry rather than traditional educational pathways.

Overall Market valuations: high but fair?

Let’s start with defining markets as the S&P 500, which represents the 500 largest companies listed in the US. Valuing “markets” on a price-to-earnings-growth ratio rather than a price-to-earnings (trailing) ratio paints an entirely different picture of where we are in terms of how expensive stocks are. PE ratios point to a frothy, overvalued market. When factoring in growth expectations, markets look fairly valued, just slightly expensive. What’s important to highlight is the optimism surrounding AI and the dominance of five to ten companies within the S&P 500 index. If the AI case still has legs, and we’re operating on the assumption that the incumbents will win the AI arms race, we’re in a great position to buy equites.

Interest rates. Hang with me, I know it’s boring. Interest rates are significantly higher than 2022. Expectations of where interest rates will be in the future have also changed significantly. In 2022 and 2023, markets operated on the expectations that rates were set to fall. The entire market was convinced that the US was going into a recession.

“100% confidence of a recession in 2024 into 2025” was the hottest narrative on the market, coming from analyst desks at the largest, most respected names on Wall Street. Then we got “US exceptionalism”, the idea that the rest of the world would go into recession but the pro-business Trump administration would lead us into a US dominated world again. That narrative was replaced by sell US, buy Europe. Liberation day spurred Germany to pass a massive bill increasing defense spending across Europe. The leading German index DAX is up +20% this year and the US is largely flat YTD.

The media and analyst have been chasing a narrative by driving while looking through the rearview mirror. With massive policy uncertainty, not just in the US but globally, we’re in a market where the future is extremely hard to predict. A world model of markets which is both easily understandable and academically proven is comparing the expected return versus the return available in bond markets. With higher interest rates, the long term average equity market return looks less popular. Big institutional money (think: pensions, endowments) can meet their return requirements by moving down the risk ladder into government and corporate debt rather than investing heavily in equity markets.

With higher rates of return being offered in bond markets, we should see less buying pressure on equties, leading to lower equity market returns than we saw under a low/zero interest rate environment. Circling back to why the incumbents should win the AI arms race, a lower interest rate environment also makes it harder to finance a cash flow negative company. Combine this with the massive amount of high valuation companies sitting on private equity books that can’t go public due to uncertainty in monetary and fiscal policy which has led to a boom in private credit markets.

A rise in popularity in private credit markets was fueled by higher interest rates. Private equity companies can recapitalize, returning capital to limited partners ad themselves, by converting their shares of private companies into debt securities. Private credit is a less democratized system than the traditional method of IPO’s. It’ll most likely lead to more income inequality and more sub-prime lending and increased financial complexity built in to the instruments being sold to institutional clients. (read: Klarna. The article doesn’t directly support my above statements, but it does rhyme.)

What’s the TLDR? Invest in public markets, but be aware that interest rates are higher now and you perhaps shouldn’t expect your 401k to go up 20% per annum in perpetuity. Financial complexity is on the rise, which is great for creating liquidity for non-liquid holdings of private equity firms, but perhaps not great for the long term health of global capital markets. Stocks seem fairly valued on a PEG basis, but keep an eye on medium term growth expectations and disruptions in the AI space.

Do your own research. Ideas presented here are not to be treated as investment recommendations. My opinions are my own and do not reflect on my employer.

Thanks for reading,

/Tommander-in-chief

BONUS: Crypto

TLDR: room to run; congress bill; systematic trading

“The GENIUS Act would establish a legal framework for issuing stablecoins in the United States.”

The GENIUS Act recently passed a key vote through congress; this bill is designed to regulate stable coins, i.e. who can issue them and defining the legal terms surrounding a “stable coin”. Congress taking action around stable coin regulation paves the way for broader adoption. In the writers opinion, the crypto space, as a whole, has plenty of room to run in the long run. Buy-and-hold broad basket or systematically trading a broad basket of coins are both viable strategies to approach crypto.

There exist academically proven factors, such a momentum or mean reversion, whose edge has been whittled away as strategies designed to profit from these edges have been more widely adopted. Given the barriers to entry in crypto markets for larger, more established institutions (read: big banks and trading desks), these edges still exist in crypto and are accessible to retail investors.