Howard Marks has been investing for more than 50 years and is one of the most widely read investors alive. He co-founded Oaktree Capital. Warren Buffett has said that when he sees a Marks memo in his mail, it's the first thing he opens.
This week, Marks published a memo on artificial intelligence. He opened it by calling himself a layman.
Then he did something you might not expect from an investor with his track record. He asked an AI to help him think about AI.
He asked Anthropic's Claude to score how predictable each part of the AI business is, from 1 (unknowable) to 10 (fully predictable). The AI gave the AI companies themselves, the labs building these models, a 2. It compared them to early-stage venture capital. Claude called its scores judgment calls, not hard data
Marks pointed out the irony. Claude is made by Anthropic, one of those labs. The machine rated its own maker as nearly unknowable.
That's the whole memo in one scene. One of the most careful investors alive is saying it plainly. Nobody knows yet who wins.
Netscape was right about the internet
Marks isn't an AI skeptic. He believes it will change companies, the economy and daily life in ways we can't picture yet.
His point is narrower and more useful to you. Being right about a technology and making money from it are two different bets.
He reminds us of the internet. It changed everything, as promised. Yet Netscape, Yahoo and Napster were household names in 2000, but the lasting winners turned out to be others, like Amazon and Google
The AI that Marks consulted summed it up well. The direction of AI is fairly predictable. What investors actually get paid for is not: how fast it happens, how large the profits are, and who keeps them.
Eleven questions, and no easy answers
Marks built his memo around eleven questions. Four of them matter most for your money.
Who will profit? If several AI models become equally smart, they may end up competing on price. Then AI users win, not the sellers.
Will today's leaders still lead in ten years? Two of the top labs were unknown to most people a few years ago. What stops someone newer from doing the same to them?
Can anyone forecast the earnings? Marks offers a simple test. Guess a company's earnings five years out. Would you bet the real number lands within 50% of your guess, either way? For most AI companies, that's a hard bet to take.
Who pays for all of this? Marks expects AI to need about a trillion dollars of new capital a year. Goldman Sachs estimates close to $600 billion in AI-related debt has been issued globally so far this year.
He also does some simple math. Suppose you need to get four things right to win a bet, and you have a 60% chance on each. Your odds of getting all four right are about 13%.
Owning shares in an AI company can mean being right about even more: the technology, the demand, the competition, the financing and the timing
Some may own more AI than you think
This part of the memo matters most for investors. You may never have bought a single AI stock. Depending on where the rest of your money is invested, you may still own a lot of AI.
Most U.S. index funds track the S&P 500 or something close to it. According to J.P. Morgan, a group of 42 AI-related companies produced roughly two-thirds of that index's price gains since January 2024.
It isn't only stocks. Marks notes that AI-related bonds now make up about 12% of a major U.S. investment-grade corporate bond index. If you own U.S. corporate bond funds, the safest part of your portfolio may carry some of the same bet.
Marks calls this a fault line running through portfolios. Bad news on chips, spending or demand wouldn't be isolated to the tech sector. It could move the whole market.
Canadians lived through this once already. In 2000, Nortel Networks made up about a third of the TSE 300, today's TSX. Investors who "just owned the index" found out how concentrated they really were when it collapsed.
AI isn't Nortel in every way. Most of today's biggest AI spenders also earn large profits from search, advertising, cloud computing and online retail. Their businesses don't live or die on AI alone, though chipmakers selling into the boom have more in common with Nortel. The point is that an index can unknowingly become a bet on one idea.
You might be wondering where we stand. We don't overweight AI in the portfolios we manage. We spread your money across low-cost index ETFs in different regions and asset classes, so no single theme drives your results. AI is one part of your portfolio, not a bet layered on top. We do see risk where we don't manage the accounts. A group RRSP or defined contribution pension at work, or a DIY trading account. If you have money in any of those, it's worth a second look, and we're happy to review it with you.
What this means for you
Marks doesn't tell anyone to avoid AI. He says uncertainty is no excuse to sit out something this important. He also says price matters more than the story.
He ends with a few questions worth asking about any investment. Here they are in plain terms.
How many things have to go right for me to earn a decent return?
If I'm wrong, how much could I lose?
How much good news is already baked into the price?
Does this add a new kind of risk to my portfolio, or more of the risk I already carry?
That last one is where we focus most of our time when creating portfolios for our clients. We can't predict which AI company wins. Neither can Marks, and neither can the AI. So we start somewhere else. Before we invest anything, we put in writing what your money needs to do, when you'll need it, and how big a drop you can sit through. We check every holding against those answers, including how much of your portfolio leans on one theme like AI. Each year you see what you own, the risk and the performance. When markets get excited about something new, we go back to your plan before we change a thing.
Marks closes with a line Warren Buffett wrote in Forbes in 1979: "You pay a very high price in the stock market for a cheery consensus."
So here's my question for you. Nobody can tell you who wins the AI race. Is your plan built so it doesn't need to know?
That's how we build every plan. Your goals, your timeline and your comfort with risk come first. We then design the portfolio with a range of outcomes in mind, without leaning on any one forecast. If you'd like to see how that works in your own portfolio, give us a call.
This document is provided as a general source of information and should not be considered personal, legal, accounting, tax or investment advice, or construed as an endorsement or recommendation of any entity or security discussed. All investments involve risk, including the potential loss of principal. Leveraged ETFs and other complex investment vehicles may not be suitable for all investors and should only be used with a full understanding of their risks. Asset class performance varies over time, and diversification does not ensure a profit or protect against a loss. Every effort has been made to ensure that the material contained in this document is accurate at the time of publication. Market conditions may change, which may impact the information contained in this document. All charts and illustrations in this document are for illustrative purposes only. They are not intended to predict or project investment results. Individuals should seek professional advice, as appropriate, regarding any particular investment. Investors should consult their professional advisors prior to implementing any changes to their investment strategies. The opinions expressed in the communication are solely those of the author(s) and are not to be used or construed as investment advice or as an endorsement or recommendation of any entity or security discussed. Mutual funds and other securities are offered through De Thomas Wealth Management, a mutual fund dealer registered in each province in which it conducts business and a member of the Canadian Investment Regulatory Organization (CIRO).
