Anthropic IPO and AI stock bubble

Investment Analysis Clubs Macro Economic Trends and Risks
1

Patrick Boyle, one of my favorite analysts, discusses the Anthropic IPO and the AI stock bubble in his usual deadpan sarcastic (but data-supported) way.

It’s worth investing a half hour in this video. Boyle’s takeaways are many but the bottom line is that investing in a stock IPO for $2 trillion of a frontier AI company with relatively minimal profits from actual end-users at a time that open source models and technology to route end-user AI queries to the cheapest, most distilled AI apps and price competition may not be the best use of investors’ hard earned cash.

I especially enjoyed Boyle’s reference to Scott McNeely’s famous 2002 Post-Dot-Com quote which highlights the stark mathematical reality of current valuations.
Here is the full excerpt from Scott McNealy’s April 2002 interview with BusinessWeek, reflecting on Sun Microsystems trading at 10 times revenue at its dot-com peak:

“At 10 times revenues, to give you a 10-year payback, I have to pay you 100% of revenues for 10 straight years in dividends. That assumes I can get that by my shareholders. That assumes I have zero cost of goods sold, which is very hard for a computer company. That assumes zero expenses, which is really hard with 39,000 employees. That assumes I pay no taxes, which is very hard. And that assumes you pay no taxes on your dividends, which is kind of illegal. And that assumes with zero R&D for the next 10 years, I can maintain the current revenue run rate. Now, having done that, would any of you like to buy my stock at $64? Do you realize how ridiculous those basic assumptions are?”

I had several chuckles along the way. When Boyle says something that doesn’t seem to make sense on its fact, understand that he’s just being sarcastic in his usual deadpan style.

Investors need to base valuations on profits, not on revenues. Estimating the value of a company on the TAM, which stands for Total Addressable Market - is pure wishful thinking that doesn’t take competition and pricing cycles into account. In the video, he discusses how tech companies and investment bankers use TAM figures—which describe the total potential revenue available if a company were to capture 100% of its target market—to justify massive, multi-trillion-dollar valuations. That’s pure speculation and was the inflator of the dot-com bubble.

I wouldn’t touch the Anthropic IPO with a 10 foot pole. Nor would I buy any bonds offered by any AI or AI-adjacent company.

40% to 45% of the S&P 500 is comprised of companies explicitly tied to the AI ecosystem (enabling hardware, hyperscale cloud infrastructure, power/utilities, and core software platforms). Roughly 60% to 75% of total index returns since the start of the current AI expansion cycle (late 2022 through 2026) have been directly generated by AI-linked companies.

The rise of high-capability open-source models, model distillation, and rapid hardware efficiency optimizations acts as a major deflationary force on model providers’ pricing power. The current trajectory of capital expenditure versus real end-user revenue is structurally unsustainable.

According to Gemini: Absent a sudden macroeconomic shock, a margin-driven valuation correction is most vulnerable to occurring over a 12 to 24-month horizon. The tipping point will arrive when hyperscaler management teams signal a deceleration or flatlining in capex growth, triggering an immediate multi-multiple contraction in high-P/E hardware and semiconductor suppliers.

It’s typical of end-stage bubbles to attract investors who feel so smart because their brokerage statements are growing rapidly. (This was true even before there was such as thing as a brokerage statement - cf. “Manias, Panics and Crashes.”)

This bubble won’t pop suddenly because some of the hyperscalers have massive revenues to support their capex. But they are already borrowing huge amounts. Interest rates are rising. When the boardrooms start to pull back because the revenues don’t support the capex, the whole ecosystem will implode. Not overnight. Maybe starting with a plateau of the SPX as we are seeing now.

One way to protect against extreme valuation concentration is to move assets from a cap-weighted to an equal-weighted SPX.

The divergence between the Equal-Weight S&P 500 (RSP / S&P 500 Equal Weight Index) and the Cap-Weighted S&P 500 (SPY / S&P 500 Index) across the three major downturns highlights a critical rule: Equal weighting excels when a downturn is driven by valuation multiple compression in mega-cap tech, but offers limited protection during broad, credit- or demand-driven economic recessions.

The conservative investor will select stocks of growing companies with low P/E ratios in industries that aren’t heavily cyclical, coupled with a ladder of short-to - moderate duration TIPS (or A rated and better bonds). Index investors will shift to equal-weight funds since GDP growth is currently strong and there’s no sign of a recession.

Wendy

7 Likes
2

I haven’t found too many funds such as these available - mainly a couple Invesco vehicles with a 0.20% expense ratio. What have you found?

Pete

3

Core S&P 500 Equal Weight Funds

  • Invesco S&P 500® Equal Weight ETF (Ticker: RSP)

    • Type: U.S.-listed ETF

    • Expense Ratio: 0.20%

    • Description: The benchmark standard and largest equal-weight S&P 500 ETF. Rebalances quarterly so that each of the 500 companies target an equal ~0.20% allocation.

  • Invesco Equally-Weighted S&P 500 Fund (Ticker: VADAX / VEIYX / VADGX)

    • Type: Mutual Fund

    • Description: A traditional open-end mutual fund vehicle offering the same strategy for retirement accounts or institutional platforms that favor mutual funds over ETFs.

  • Xtrackers S&P 500 Equal Weight UCITS ETF (Ticker: XDEW / EWSP)

    • Type: European/UCITS ETF

    • Expense Ratio: 0.15%

    • Description: Designed for European and international investors needing UCITS-compliant access to equal-weighted U.S. large-cap equity.

  • iShares S&P 500 Equal Weight UCITS ETF (Ticker: EWSR)

    • Type: European/UCITS ETF

    • Expense Ratio: 0.15%

    • Description: BlackRock’s UCITS vehicle tracking the S&P 500 Equal Weight Index.

2 Likes
4

Cash flow is harder to spoof than profits. During the 2008 meltdown many companies booked credit sales as profit by having their “finance” arm manage the credit.

GoogleAI:

That is a highly accurate observation. Cash flow is widely considered a cleaner metric than net profitbecause revenue recognition and accrual accounting leave significant room for interpretation, whereas actual cash entering or leaving a bank account is much harder to manipulate.

During the lead-up to the 2008 financial crisis, many corporations—most notoriously within the automotive and consumer goods sectors—heavily relied on their captive finance arms (like GMAC for General Motors) to artificially prop up their core business metrics through aggressive credit schemes.

During the dot.com bubble well respected economists were saying that cash is king but they included cash from angel investors and VPs as cash flow instead of just cash from sales.

GoogleAI:

During the dot-com bubble of the late 1990s and early 2000s, the traditional phrase “cash is king” was fundamentally distorted by Wall Street analysts, venture capitalists (VCs), and some economists. In standard corporate finance, “cash flow” strictly refers to Cash Flow from Operations (CFO)—the money a business generates from selling its actual products or services. However, during the dot-com mania, the definition was dangerously expanded to include Cash Flow from Financing (CFF), which includes cash injections from angel investors, VCs, and Initial Public Offerings (IPOs).

This conflation allowed structurally unprofitable companies to mask their massive operational losses under the guise of having “strong cash positions.”

Maybe a better guide than either profit or cash flow is understanding the business which was Peter Lynch’s main guide, “Buy what you know.”

GoogleAI:

Peter Lynch’s famous philosophy—“Buy what you know”—is one of the most powerful and accessible rules in investing history.

While financial metrics like profit margins and free cash flow tell you how well a business is doing, understanding the actual product, service, and customer experience tells you why it is doing well and whether it can keep doing it.

AI gave more details which you can find by repeating my queries.

The Captain

1 Like
5

By that he meant “Buy what you understand.” About a third of his classic book “One Up on Wall Street” is a discussion of how to understand and what to look for on corporate balance sheets. He was looking for Growth at a Reasonable Price (GARP). He popularized (but I don’t believe invented) the popular valuation metric Price/Earnings Growth (PEG) ratio, and used EPS growth as tool, for example. He is very much a profit and cash flow guy.

Since we’re talking about the AI stock bubble, I had Gemini create a Lynch-like table of PEG ratios and EPS growth for big publicly traded AI companies. I’ve posted a similar tables before, but I asked Gemini to do a Lynch-style analysis as well. You can see the results below.

A couple notes: All of these companies flunk a Lynch test of already attracting large institutional investors. He likes stuff that has’t been discovered. Also, Amazon, Microsoft, and especially Alphabet’s earnings are a little wonky because they all include large unrealized capital gains from AI and SpaceX investments. So their PEG ratios should probably be a little higher.

But for the most part they look pretty good from a Lynchian perspective. Nvidia especially. Bonkers EPS growth and tiny PEG (smaller is better). The outlier is Tesla which Lynch would run away from screaming with his hair on fire.

Company Ticker Trailing P/E Ratio 3-Year EPS Growth Rate Current PEG Ratio Peter Lynch Rule of Thumb Context
Amazon AMZN ~20.1 ~88.3% (Trailing/Recent surge) 0.23 Heavily discounted relative to recent explosive earnings growth (PEG < 0.5).
Alphabet GOOGL ~17.2 ~112.4% (Trailing acceleration) 0.15 Trades at a very low PEG due to sharp margin expansion and robust profit growth.
Nvidia NVDA ~28.4 ~142.6% (3-Yr annualized average) 0.23 Exceptionally low PEG ratio; astronomical earnings growth has temporarily outpaced its multiple expansion, making it look statistically cheap under a strict Lynch screen.
Tesla TSLA ~320.3 ~17.7% (5-Yr EBITDA / 3-Yr baseline) 22.32 Extremely high PEG; P/E significantly outpaces current earnings growth rate.
Meta META ~28.6 (ex-NRI) ~24.0% (5-Yr baseline metric) 1.19 Fairly valued to slightly above Peter Lynch’s ideal 1.0 threshold.
Microsoft MSFT ~28.5 ~21.0% (3-Yr average baseline) 0.88 – 1.52 Hovering near fair value, aligning closely with consistent enterprise cloud and AI growth.
8 Likes
6

There are two sides to Lynch investing, “What you know” and “Valuation.” They are not mutually exclusive.

The Captain

2 Likes
7

This made me curious about the dot-com bubble long term outcome so I asked AI. Interesting that despite all the companies going bust, VGT still beat SPY. Glad I kept my AAPL. It offset my CMGI.

Company / Ticker / Fund 2000 Bubble Peak Status Price (9/25/2026) Est. Total Return (With Dividends) Multi-Decade Fate & Modern Status
:crown: AAPL (Apple) Consumer niche tech $228.45 +23,500%+ Shifted to mobile ecosystems; became a dominant mega-cap cash engine.
:package: AMZN (Amazon) Early online book retailer $194.20 +3,780%+ Survived a 93% crash; built the multi-billion-dollar AWS cloud backbone.
:laptop: MSFT (Microsoft) Dominant PC operating system $516.17 +1,115% Stagnated for 15 years, then exploded via Azure cloud and enterprise AI.
:shield: VGT (Vanguard Tech) Diversified tech fund $126.17 (Split-adj) +840% Broke even and outpaced the broad market by rotating into new mega-caps automatically.
:balance_scale: SPY (S&P 500 ETF) General market benchmark $771.35 +580% The gold standard benchmark for broad, diversified equity compounding.
:globe_with_meridians: CSCO (Cisco) Networking hardware king $106.70 +112% Stock price barely cleared its 2000 nominal high; dividends provided the only real gain.
:floppy_disk: INTC (Intel) Semiconductor manufacturer $123.00 +105% Severely lagged the index for over two decades before recent manufacturing turns.
:hot_beverage: JAVA (Sun Microsystems) Server infrastructure anchor N/A (Delisted) -96% Eradicated by commoditized server shifts; acquired by Oracle for cheap in 2009.
:office_building: CMGI (Steel Connect) High-flying tech incubator N/A (Delisted) -99.9% Dropped 99% during the crash; corporate remnants pivot to supply-chain management.
:magnifying_glass_tilted_left: Inktomi Search backend infrastructure N/A (Delisted) -99% Peak $25B valuation vanished as clients cut spend; bought by Yahoo in 2003 for $241M.
:world_map: Excite Core internet web portal N/A (Delisted) -100% Merged disastrously, went bankrupt in 2001. The legacy portal domain was fully shut down in August 2026.
3 Likes
8

People keep mentioning equal weight, but what’s the case for it?

SPY trounced RSP in recent history and by allocating more to smaller companies you are buying companies with lower quality financials.

Since January 1, 2020, SPY had total return of 163.58% and RSP returned 104.94%

So SPY gives you 60% more dollars over just the last nearly 6 years.

SPY has better risk-adjusted returns: Sharpe Ratio 0.58 vs 0.35.

SPY has better annual earnings growth: 15.5% vs 9.2%.

9

Diversification ==> lower risk and steadier returns.

DB2

1 Like
10

The S&P Equal Weight index was created in 2003, since that time RSP has returned 10.94% vs. 11.06% for the cap weighted index, but most of that underperformance was in the last five years. Backtests back to 1970 show equal weight outperforms by about a percent.

Equal weight is sort of a mid-cap strategy (which tends to outperform over time), combined with an anti-momentum (pull the flowers, water the weeds) component, which is sort of a contrarian play. That works most of the time, but it doesn’t work all of the time.

Since 2022, the Mag 7 have increased by nearly 200% and now make up 35% of the S&P 500. The rest of the index was up like 40% or something. So equal weight would have missed out of most of that run up.

But at 35% of the index, a regular index investor is actually pretty heavily concentrated in the AI/Tech sector. The Mag7 only make 0.14% of RSP. Should the bubble burst, the same forces that caused RSP to lag the S&P would stanch the worst of the bleeding.

There is also a Nasdaq equal weight ETF, QQQE. It has really lagged the cap weighted index.

https://www.spglobal.com/spdji/en/research/article/more-equal-than-others-20-years-of-the-sp-500-equal-weight-index/

3 Likes
11

I would not have guessed those numbers to be so close.

I wonder if smaller caps will ever recover their former glory?

Same for value?

Same for small cap value?

12

Everything has its day in the sun. And mostly they all get their time in the barrel too; the trick is to know when to move from one to the other. At the moment the entire SPY is being dragged upwards (unreasonably, in my view) by a few super-out-performers. That, as they say, cannot go on forever, and when it stops those weights will cause the same kind of selling volume that the buying has been providing in the opposite direction.

I know, I know, you want to know “when.”

May 3, 2027. Happy?

3 Likes
13

Yep. Your table shows the power of Long-Term Buy & Hold.

intercst

14

Switching from the standard S&P500 to “equal weight” gives you larger allocation to the poorer performers. It’s a proven way for “smart people” to under perform.

Prof. Terrance Odean studied this phenomenon in his landmark 2000 paper.

When people sell their “winners” to diversify, the new stock they bought invariably under performs the one they sold.

I learned this lesson myself in 1996 when I sold 1/3 of my DELL stock to “diversify”. It quickly became a $1MM mistake. It’s been LTB&H for me ever since.

intercst

3 Likes
15

I sold all of my AOL and Cisco in 1999. Was that a mistake?

I sold everything, and I mean everything (longstanding purchase of BRK in a taxable account excepted) in 2008. Winners, losers, all of them. I’m pretty sure that on e worked out, too.

LTBH is for lucky people, or lazy, or both. :wink:

4 Likes
16

It was not. interest is suffering survivor bias when he talks about his DELL bet. The same bet in AOL or Cisco, at the same time, would have been a disaster. Ditto for INTC.

1 Like
17

I would guess:

AI Overview

Modern Portfolio Theory (MPT) is a mathematical framework developed by Harry Markowitz in 1952 that helps investors maximize return for a designated level of risk through strategic diversification. [1, 2]

Which I dislike.

The Captain

18

Not likely. Most high quality small caps now wait until they are midcap before going public. The result is that a greater percentage of the small caps are lower quality these days.

3 Likes
19

Samesies. The leaked prospectus is chock-full of risk. Apart from it being an unprofitable company…

“A quarter of its revenues came from just two customers, the company said, “and as part of its risk factors, warned that many of its largest clients were not locked into long-term contracts and could cut or stop spending,” according to Reuters.”

True dat! Here’s a simple analysis that should give AI investors pause.

image
image960×561 62.8 KB

“In other words,” he continued, “the tech silo is betting on a future in which demand for AI and tech services explodes, while the silos covering the companies that would pay for those services see a much more modest outlook. Both cannot be right at the same time.”

4 Likes
20

That’s mungofitches theory. There is also a theory that private equity snatches up good performing small companies and they never go public.

Markets and the business environment change over time. Small cap value historically has been the best performing sector, but maybe that won’t be true going forward.

1 Like