
Oct 21, 2026 · 1h 14m
Listen from 9:08
Listen at 9:08
Arnott challenges the hidden costs of passive investing
Rob Arnott on the Hidden Costs of Passive Investing and the Real Winners From AI
The episode connects AI’s investment boom to index mechanics, infrastructure economics, accounting distortions, and the risk of mistaking recent performance for durable value.
- 1AI may transform productivity, but infrastructure suppliers could capture less value than businesses using the technology.
- 2Cap-weighted indexes embed active decisions and impose trading costs that expense ratios do not reveal.
- 3Rapid hardware obsolescence, intangible investment, and overfit backtests complicate how investors measure durable returns.
Don't miss
Rob Arnott explains why cap-weighted indexing is not entirely passive and how index rebalancing can create measurable performance drag.
The brief
Rob Arnott tells Kai Wu that AI is a genuine productivity revolution, but investors may be overexposed to the companies building its infrastructure rather than those capturing its benefits.
The conversation reframes passive investing: index additions, deletions, and rebalancing create forced trades that hedge funds can anticipate, leaving costs beyond the published expense ratio.
Arnott questions whether AI infrastructure can earn adequate returns as chips become obsolete, power and cooling demands rise, and vendors help finance customer purchases.
The episode broadens the accounting problem: R&D can create durable intangible value while technical advantages decay quickly, making leadership and strategy harder but potentially more useful measures of durability.
Arnott closes by warning that investors often overfit backtests and chase recent winners, while neglected value strategies may benefit when valuations mean-revert.
What was said on this episode
40 statements · 24 positive · 14 negative · 2 neutral
Nearly half of the S&P 500 consists of AI-revolution leaders.
“Nearly half of the S&P 500 is companies that are leaders in the AI revolution.”
Listen at 2:42
AI will probably become the strongest lifetime productivity enhancer for the global economy.
“the AI revolution is incredibly powerful. It is very real. It is probably going to be the most powerful productivity enhancer for the global macroeconomy in our lifetimes.”
Listen at 2:51
AI will probably become the strongest lifetime productivity enhancer for the global economy.
“It is probably going to be the most powerful productivity enhancer for the global macroeconomy in our lifetimes.”
Listen at 2:56
Technological revolutions create more jobs than they destroy.
“Technological revolutions always create more jobs than they destroy.”
Listen at 4:02
Every technological revolution is accompanied by an investment bubble.
“But every technological revolution is accompanied by a bubble.”
Listen at 5:29
DeepSeek disrupted OpenAI.
“OpenAI was disrupted by DeepSeek.”
Listen at 6:33
OpenAI disrupted Google.
“Google was disrupted by OpenAI.”
Listen at 6:36
Most top-ten technology stocks from 2000 underperformed the S&P afterward.
“The rest underperformed the S&P.”
Listen at 7:09
AI will likely produce a similar bubble outcome.
“AI is likely to be no exception.”
Listen at 7:31
Only some current AI leaders will remain dominant in ten years, and perhaps one or two in 25 years.
“The dominant players today Some will still be dominant in a decade. Maybe one or two will still be dominant in a quarter century”
Listen at 7:34
Customers capture more value than innovators in technological revolutions.
“Every technological revolution requires that the customers win more than the innovators.”
Listen at 7:51
Ralph Lauren is a smart AI user for selling products.
“Is it a smart user of AI using it as another angle to sell product? Yes.”
Listen at 9:08
AI’s biggest winners will be the broader economy using the technology.
“the big winners are going to be the rest of the economy that's being left in the dust.”
Listen at 9:37
The next 500 companies’ cash flow grew three percentage points faster than the top 500.
“the underlying fundamentals, the cash flow of the next 500, has been growing 3% per year faster. Than the top 500.”
Listen at 11:31
“highly likely to outperform the large-cap growth that so utterly dominates our attention these days by perhaps 600, 800 basis points per annum. For the coming decade”
Listen at 12:22
Tesla rose roughly 50% between S&P announcement and index inclusion.
“Between the November announcement and the addition to the index, it rose roughly 50%.”
Listen at 18:26
Tesla rose 900% from March lows to S&P 500 inclusion.
“From the March lows to the addition to the S&P 500, it rose 900%.”
Listen at 18:55
S&P indexers would need to buy roughly 25% of SpaceX’s float upon inclusion.
“SpaceX gets added. The indexers have to buy 25% of the entire market capitalization, or at least the entire float of SpaceX.”
Listen at 19:31
S&P indexers will buy 25% of SpaceX’s float from reluctant sellers.
“between the IPO and the addition to the S&P, S&P indexers will have to buy 25% of the float from people who don't want to sell.”
Listen at 22:31
Index deletions push stocks to abnormally low valuations, creating buying opportunities.
“the stocks that get dropped are pushed down to abnormally low valuations on their way out. And that actually creates a buying opportunity.”
Listen at 26:58
Trading at index announcements would have improved returns by about 25 basis points annually since 1989.
“your performance, um, over the last— since 1989 would have been about 25 basis points per annum better.”
Listen at 28:59
RACWI beat the S&P 500 by 70 basis points annually over 35 years.
“if you do that with 500 stocks, It beats the S&P 500 by 70 basis points per annum for 35 years.”
Listen at 31:40
RACWI outperformed MSCI ACWI by over 1% annually with slightly over 1% tracking error.
“it's outperformed MSCI ACWI by over 1% per year with just over 1% per year tracking error.”
Listen at 35:31
RACWI added approximately 1% annually versus the S&P with 0.6% tracking error.
“against the S&P, it's added almost exactly 1% a year with 0.6% tracking error.”
Listen at 36:11
RACWI modestly overweights Nvidia relative to the S&P 500.
“If Nvidia is 7% of the S&P, it'll be 7.2% of our index.”
Listen at 37:12
Many AI providers, including Nvidia, lend customers money to purchase their products.
“many of the providers— Nvidia is a beautiful example— are lending money to their customers to buy their product.”
Listen at 40:00
AI customers must derive more value than cost for chip vendors to profit.
“the AI customers have to value the AI product by more than it costs in order for the AI vendors to value the chips for more than they cost.”
Listen at 42:31
Markets price AI infrastructure as if end customers will not benefit.
“the market is pricing things as if the end customer is irrelevant.”
Listen at 43:07
If AI chips become uneconomic after three years, half of annual spending could be written off.
“if that cycle is 3 years and you're doing depreciation of the purchase over 6 years, then that means that let's say there's $1 trillion of spend in a given year on AI tech. That means half a trillion of it will be a write-off in 3 years.”
Listen at 47:38
Small-cap value and emerging-market value trade at CAPE ratios around 10.
“Small-cap value, emerging markets value are trading dirt cheap, CAPE ratios of 10.”
Listen at 48:21
Capitalizing R&D and one-third of G&A doubles book-value value-factor efficacy.
“the efficacy of a book value-based value factor doubles.”
Listen at 52:17
R&D-adjusted value investing produced eight times growth-investor wealth over 60 years.
“a simple price-to-book value factor would have given you 4 times the wealth if you were a value investor as a growth investor over the last 60 years. Throwing in R&D and a little bit of G&A doubles that. You wind up 8 times as rich instead of 4.”
Listen at 53:12
Technological competitive moats are generally not durable.
“moats are, um, not durable.”
Listen at 55:27
Competitive moats are fragile at best.
“So, so I think moats are fragile at best.”
Listen at 56:46
Durable competitive advantage depends more on responsive leadership than conventional moats.
“a durable moat really is more a function of leadership that has its finger on the pulse of their clientele”
Listen at 59:05
AI has probably made the firm’s programmers three to four times more productive.
“Our programmers are probably 3 or 4 times as productive as they used to be.”
Listen at 1:04:49
AI produces more thorough paper reviews than human referees in Arnott’s experience.
“I'll get a 10-page review of the paper that is better than any, any of the hundreds of referee papers that I saw when I was editor.”
Listen at 1:06:35
AI has not independently generated new ideas in Arnott’s experience.
“I haven't seen AI come up with any independent new ideas in the times I've worked with it.”
Listen at 1:09:39
Optimizing for attractive backtests does not establish a robust investment strategy.
“Looking for a great backtest assures you of a strategy that has a great backtest.”
Listen at 1:11:02
Mean reversion could make value investing exceptionally strong in the coming decade.
“any mean reversion could mean— would mean that value could be absolutely stupendous in the coming decade.”
Listen at 1:12:28
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Books & mentions
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.
