【The Plain Bagel】美股估值逆天？标普500到底是贵还是便宜？

浑水摸鱼清源

AI-assisted summary of 【The Plain Bagel】美股估值逆天？标普500到底是贵还是便宜？ — original source. Check the original for full context.

Source published: 2026-10-12

Key ideas

The S&P 500's Shiller CAPE ratio has climbed above 40x, near historical highs from the dot-com era, but the metric is heavily distorted by smoothing 10 years of historical data during a period of ~30% year-over-year earnings growth. (4:51)
Factoring in Wall Street's aggressive 27.3% 5-year annualized earnings growth projection, the S&P 500 forward PEG ratio drops to 0.69, creating an optical signal of multi-decade undervaluation despite an elevated forward P/E of ~19x. (7:37)
While S&P 400 mid-caps and S&P 600 small-caps trade at cheaper forward P/E multiples of ~15x and ~14x, international equities in the MSCI ACWI ex USA offer the widest valuation margin at just 12.5x forward earnings alongside surging growth forecasts. (10:00)
The US Equity Risk Premium has compressed to approximately 4%, near multi-year lows, as elevated 10-year Treasury yields diminish the relative incentive to hold volatile equities and threaten capital-dependent AI infrastructure projects. (15:00)

Detailed notes

## Background This video was published on Bilibili by 浑水摸鱼清源 on October 12, 2026, featuring content from financial analyst and creator The Plain Bagel (Richard Coffin). According to the page description, the video investigates whether soaring US equities are trapped in an AI bubble by dissecting the S&P 500 across multiple valuation benchmarks, including the Shiller CAPE, forward P/E, PEG ratios, and the equity risk premium.

## TL;DR Host Richard Coffin evaluates whether the S&P 500 is historically overvalued by contrasting contradictory valuation indicators across the market. While backward-looking metrics like the Shiller CAPE ratio sit above 40x near dot-com bubble extremes, aggressive analyst forecasts project 27.3% annualized earnings growth, producing an optically cheap forward PEG ratio of just 0.69. However, this apparent bargain relies on severe analyst optimism bias, heavy market concentration in the 'Magnificent 7', unsustainable margin expansion from corporate cost-cutting, and capitalized AI infrastructure spending that obscures actual cash burn. Ultimately, with rising 10-year Treasury yields compressing the Equity Risk Premium to around 4%, the market prices in near-flawless execution with very little margin for error.

## Atmosphere & style The presentation delivers a calm, methodically grounded antidote to financial sensationalism, blending technical equity analysis with Richard Coffin's trademark dry, self-deprecating wit. Rather than hyping doom or pushing euphoria, Coffin acts as an empathetic guide who openly acknowledges how disorienting conflicting valuation metrics can be for ordinary investors. The emotional arc moves from the dizzying noise of headline AI narratives into a rigorous, sobering deconstruction of Wall Street assumptions, concluding on an earnest reminder to remain humble and diversified.

## Bulletin - [04:51](https://www.bilibili.com/video/BV1mBpg6iEbx?t=291s) The S&P 500's Shiller CAPE ratio has climbed above 40x, near historical highs from the dot-com era, but the metric is heavily distorted by smoothing 10 years of historical data during a period of ~30% year-over-year earnings growth. - [07:37](https://www.bilibili.com/video/BV1mBpg6iEbx?t=457s) Factoring in Wall Street's aggressive 27.3% 5-year annualized earnings growth projection, the S&P 500 forward PEG ratio drops to 0.69, creating an optical signal of multi-decade undervaluation despite an elevated forward P/E of ~19x. - [10:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=600s) While S&P 400 mid-caps and S&P 600 small-caps trade at cheaper forward P/E multiples of ~15x and ~14x, international equities in the MSCI ACWI ex USA offer the widest valuation margin at just 12.5x forward earnings alongside surging growth forecasts. - [15:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=900s) The US Equity Risk Premium has compressed to approximately 4%, near multi-year lows, as elevated 10-year Treasury yields diminish the relative incentive to hold volatile equities and threaten capital-dependent AI infrastructure projects.

## Key takeaways - [00:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=0s) Diagnosing an overarching market bubble is far more complex than identifying speculative individual outliers like SpaceX, which reached the top 10 US market cap despite ranking outside the top 200 in revenue. - [02:54](https://www.bilibili.com/video/BV1mBpg6iEbx?t=174s) Discounted cash flow (DCF) modeling cannot be practically applied to an entire index because compounding assumptions across 500 companies create an unusable margin of error, making market-cap-weighted relative P/E ratios the preferred standard. - [04:51](https://www.bilibili.com/video/BV1mBpg6iEbx?t=291s) The Shiller CAPE ratio exceeds 40x—its highest level since the dot-com crash—primarily because its 10-year backward-looking smoothing method understates recent explosive corporate earnings growth. - [04:51](https://www.bilibili.com/video/BV1mBpg6iEbx?t=291s) Unlike the dot-com bubble where companies lacked fundamental profitability, today's S&P 500 trailing P/E remains well below historical peak extremes and is supported by real corporate profits. - [07:37](https://www.bilibili.com/video/BV1mBpg6iEbx?t=457s) The S&P 500 forward P/E sits at roughly 19x, elevated compared to its 16x historical baseline but down from dot-com peak levels of 25x as earnings have kept pace with equity price increases. - [07:37](https://www.bilibili.com/video/BV1mBpg6iEbx?t=457s) Consensus analyst expectations project a 27.3% annualized 5-year corporate earnings growth rate, driving the forward PEG ratio down to an unusually low 0.69. - [10:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=600s) S&P 400 mid-caps (15x forward P/E) and S&P 600 small-caps (14x forward P/E) appear cheaper on pure earnings multiples, yet carry higher PEG ratios (0.80 and 0.96) than the mega-cap S&P 500 due to lower projected growth. - [10:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=600s) International equities represented by MSCI ACWI ex USA trade at an attractive forward multiple of 12.5x, offering strong earnings revisions and a larger margin of safety than US benchmarks. - [12:12](https://www.bilibili.com/video/BV1mBpg6iEbx?t=732s) Forward valuation ratios are distorted by systematic Wall Street optimism, where rolling 25-month consensus estimates reliably trend downward over time as economic realities set in. - [12:12](https://www.bilibili.com/video/BV1mBpg6iEbx?t=732s) Index-level valuations are heavily skewed by the 'Magnificent 7,' which account for roughly one-third of the total index capitalization and trade at substantial valuation premiums over the remaining 493 components. - [12:12](https://www.bilibili.com/video/BV1mBpg6iEbx?t=732s) A wide gap exists between projected revenue growth (15.6%) and bottom-line earnings growth (>27%), signaling that expected profits rely heavily on temporary AI cost-cutting rather than sustainable top-line expansion. - [15:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=900s) Massive corporate investments in data centers and AI chips are capitalized on balance sheets rather than expensed upfront, risking major future write-downs if hardware depreciates faster than accounting assumptions predict. - [15:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=900s) At roughly 4%, the Equity Risk Premium offers minimal extra compensation for equity risk over risk-free debt, leaving the market highly vulnerable to multiple compression if Treasury yields remain elevated.

## Detailed walkthrough ### [00:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=0s) The AI Bubble Debate & The Problem with Predicting the Future

The single biggest question floating around stock investing right now is whether AI is going to kill us—followed closely by whether AI is currently in a massive bubble. Headlines and financial commentators have been sounding the alarm relentlessly, flashing warnings that the bubble is bursting in South Korea, popping in tech, or collapsing in video AI ("Wait a minute, that's me," the host jokes, recognizing his own past thumbnail). The anxiety is understandable: tech giants have completely taken over major market indices, pouring hundreds of billions of dollars into massive infrastructure for artificial intelligence ventures that remain, for the time being, largely unprofitable. At the same time, the broader economy faces stubborn headwinds from rising oil prices, climbing Treasury yields, and elevated inflation. Yet despite this wall of worry, the stock market continues to notch fresh all-time highs, leading many investors to assume valuations must be dangerously stretched.

There are certainly plenty of individual warning signs and questionable valuations across the tech landscape. SpaceX is a prime illustration of this disconnect: it recently broke into the top 10 most valuable US companies by market capitalization, even though its actual revenues do not even crack the top 200. But diagnosing whether the market *as a whole* is in a bubble is far trickier than looking at individual outliers. Bulls argue that while AI stocks look expensive today, their immense future earnings potential justifies the premium—the businesses spending heavily now stand to reap unprecedented rewards tomorrow. The future will inevitably land somewhere between two extreme poles: a utopia where everyone retires wealthy with a Tesla Optimus robot folding their laundry, or an Armageddon where AI takes over and destroys the economy because "tech bros flew too close to the sun." Because there are simply too many unpredictable variables to forecast that outcome, the more grounded exercise is to evaluate whether the market—specifically the S&P 500—is historically expensive based on hard valuation data.

- **Macro headwinds vs. Market peaks**: The S&P 500 continues printing all-time highs despite high inflation, rising bond yields, and higher energy prices. - **Extreme valuation disconnect**: SpaceX cracked the top 10 market cap list in the US while ranking outside the top 200 in corporate revenue. - **The Core Paradox**: One set of valuation charts makes the market look like an unmistakable, overinflated bubble, while another set makes it look historically attractive.

---

[02:54](https://www.bilibili.com/video/BV1mBpg6iEbx?t=174s) Valuing an Index: Why DCF Fails and Relative P/E Takes Over

To evaluate the broader market, the benchmark of choice is the S&P 500. While commonly thought of simply as the 500 largest US companies, it is technically 500 of the largest *eligible* US stocks by market cap; companies must meet strict profitability and listing history criteria, which is why private or unseasoned names like SpaceX are excluded. Looking purely at the price level of the index tells investors very little: yes, the S&P 500 has doubled over the past four years, but that raw price gain is meaningless unless measured against fundamental business performance.

When evaluating a single company, an analyst would typically build a Discounted Cash Flow (DCF) model, forecasting all future cash flows, discounting them back to the present value based on timing and risk, and adjusting for net debt. Applying a DCF model to an entire index, however, is practically impossible. Building 500 individual cash-flow models would require stacking so many speculative assumptions on top of one another that the compounding margin of error would render the final valuation figure useless. Instead, index valuation relies on relative valuation: aggregating a single fundamental metric across all 500 constituents into a market-cap-weighted average. The most universal baseline is corporate earnings, expressed through the classic Price-to-Earnings (P/E) ratio (share price divided by earnings per share).

```mermaid flowchart TD accTitle: Index Valuation Methodologies accDescr: Flowchart contrasting Discounted Cash Flow against Relative Market-Cap Weighted P/E for index valuation.

subgraph DCF["Individual Stock Approach: DCF"] A["Estimate 500 Sets of Future Cash Flows"] --> B["Apply Discount Rates & Adjustments"] B --> C["Compounding Margin of Error (Unusable)"] end

subgraph Relative["Index Approach: Relative P/E"] D["Company Stock Price / EPS"] --> E["Calculate for all 500 Constituents"] E --> F["Market-Cap Weighted Aggregate P/E"] end ```

---

[04:51](https://www.bilibili.com/video/BV1mBpg6iEbx?t=291s) Shiller CAPE vs. Trailing P/E: The Distortion of Rapid Earnings Growth

When investors look at the famous Shiller P/E—also known as the Cyclically Adjusted Price-to-Earnings (CAPE) ratio—the market looks terrifying. The CAPE ratio currently sits above 40x, a dizzying height not seen since the peak of the dot-com bubble and fast approaching an all-time record. This single chart is often brandished by bears as proof of an impending "dot-com crash 2.0." However, the CAPE ratio's methodology explains why it looks so extreme: it divides the current price by the average inflation-adjusted earnings of the past 10 years to smooth out business cycle volatility. While smoothing prevents cyclical swings from distorting long-term value, it penalizes companies during periods of explosive, structural earnings acceleration by heavily downweighting recent profits.

That lag is critical right now. In Q3, the S&P 500 delivered year-over-year earnings growth of nearly 30%—marking the third consecutive quarter with growth topping 25%. By comparison, the 25-year historical average earnings growth rate for the S&P 500 is only about 6%. While critics rightly point out that building 3 to 4 data centers a week across the US cannot continue forever, genuine AI-driven productivity gains mean the CAPE ratio's 10-year backward look severely overstates today's overvaluation. In contrast, looking at a standard trailing 12-month (TTM) P/E ratio paints a much less apocalyptic picture. While trailing P/E is far more volatile—skyrocketing during the 2008 Great Financial Crisis when corporate earnings temporarily vanished—today's trailing P/E, though elevated, sits well below the speculative stratosphere reached during the dot-com peak. Today's market leaders have real, massive net profits backing up their market capitalization.

---

[07:37](https://www.bilibili.com/video/BV1mBpg6iEbx?t=457s) Forward P/E and the Shockingly Low Forward PEG Ratio

Because equity investors price stocks based on where companies are heading rather than where they have been, forward-looking valuation metrics offer a very different perspective. Forward P/E takes the current price and divides it by consensus estimated earnings over the next 12 months (NTM), aggregating analyst forecasts across Wall Street. Drawing from consensus data compiled by Ed Yardeni of Yardeni Research, the forward P/E for the S&P 500 currently hovers around 19x. While that is above the long-term historical average of 16x, it is far below the dot-com peak of 25x and has actually been gently drifting lower over the past year as earnings caught up to stock prices.

The most provocative signal comes from the forward PEG ratio (Forward P/E divided by the expected 5-year compound annual earnings growth rate). In standard financial analysis, a PEG ratio above 1.0 indicates that a stock is expensive relative to its growth, while a PEG below 1.0 suggests it is cheap. Because analysts have dramatically upgraded corporate earnings forecasts alongside the AI infrastructure boom, the consensus 5-year earnings growth estimate has surged over the past two years to an eye-popping 27.3%. Dividing an approximate 19x forward P/E by this growth rate yields an S&P 500 forward PEG ratio well below 1.0 (around 0.69)—making the index look like the cheapest it has been in decades. This sets up an extraordinary contrast: trailing metrics make the market look historically overpriced, while forward growth expectations make it look like a rare bargain. Yet, as the host warns, accepting that the market is "cheap" at face value based on forward estimates ignores critical underlying flaws: "We're not done beating this dead horse till I say so."

- **Yardeni Research Data**: S&P 500 consensus forward P/E is ~19x vs. a long-term average of 16x and a dot-com peak of 25x. - **Surging Growth Expectations**: 5-year consensus annualized earnings growth rate has exploded to 27.3%. - **Historically Depressed PEG**: Forward PEG sits at a multi-decade low (~0.69), giving the surface-level impression of deep undervaluation if analyst forecasts prove accurate.

[10:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=600s) Valuation Beyond the S&P 500: Small Caps, Mid Caps, and International Markets

Up to this point, the assessment of whether stocks are cheap or expensive has relied solely on the historical performance of the benchmark index itself. Broadening the perspective requires evaluating how the S&P 500 compares relative to its peers within the broader equity universe. A logical place to start is domestic small- and mid-cap equities, represented respectively by the S&P SmallCap 600 and the S&P MidCap 400. In terms of forward price-to-earnings (P/E) multiples, these smaller companies look distinctly more affordable: the mid-cap S&P 400 trades at roughly 15 times forward earnings, and the small-cap S&P 600 sits around 14 times, compared to the S&P 500 trading at nearly 20 times forward earnings.

```mermaid flowchart TD accTitle: Cross-Market Valuation Comparison accDescr: Comparison of Forward PE and PEG ratios across S&P 500, mid-caps, small-caps, and international equities.

subgraph US["U.S. Market Spectrum"] SP500["S&P 500 Forward P/E: ~20x PEG: 0.69"] SP400["S&P 400 Mid-Cap Forward P/E: ~15x PEG: 0.80"] SP600["S&P 600 Small-Cap Forward P/E: ~14x PEG: 0.96"] end

subgraph Global["Global Equity"] MSCI["MSCI ACWI ex-U.S. Forward P/E: 12.5x Strong projected earnings"] end

Examining growth-adjusted metrics scrambles this traditional dynamic. When factoring in the price/earnings-to-growth (PEG) ratio, the S&P 500 actually emerges as the cheapest of the three. While the comparison chart is somewhat crowded, the data is unambiguous: the S&P 400 carries a PEG of 0.8, the S&P 600 registers at 0.96, and the flagship S&P 500 sits at the very bottom with a PEG of just 0.69. The mega-cap index maintains an optically higher price multiple entirely because analysts project far superior earnings growth rates for its dominant constituents.

Widening the lens globally reveals an even starker valuation discrepancy. Using the MSCI All Country World Index ex USA (MSCI ACWI ex USA)—which tracks equities across 22 developed and 24 emerging economies and captures approximately 85% of global market capitalization outside the United States—shows that international stocks have enjoyed a quiet, robust run through 2025 and into the current period. Even after that solid performance, the ex-U.S. index trades at an undeniably cheap forward P/E of just 12.5 times by historical standards. While five-year compound earnings growth data is unavailable to calculate an identical PEG figure, international earnings estimates are surging and are forecast to expand at an annualized pace comparable to the S&P 500, despite international equities changing hands at a steep, persistent discount.

- **S&P 400 (Mid-Cap)**: Forward P/E of ~15x, forward PEG ratio of 0.80. - **S&P 600 (Small-Cap)**: Forward P/E of ~14x, forward PEG ratio of 0.96. - **S&P 500 (Large-Cap)**: Forward P/E of ~20x, forward PEG ratio of 0.69. - **MSCI ACWI ex USA**: 22 developed and 24 emerging markets (~85% of international equities); trades at ~12.5x forward earnings with rapid upward revisions.

---

[12:12](https://www.bilibili.com/video/BV1mBpg6iEbx?t=732s) The Four Big Caveats: Forecast Bias, Market Concentration, and Cost Cuts

At this point, viewers might feel pulled in every direction. Host Richard Coffin acknowledges the exasperation: "Now I know what you're thinking, by God, Richard, we've been all over the place now. Just give it to me straight. Is the S&P 500 expensive or not?" The answer cannot be reduced to a binary yes or no without dissecting several structural caveats, blind spots, and accounting quirks that prop up these attractive forward multiples.

The first major complication is the unreliability of Wall Street forecasts. Any forward-looking valuation is only as good as the consensus estimates feeding into it. To illustrate analyst fallibility, Yardeni Research tracks consensus expectations on a visualization aptly titled "squiggles." The chart plots actual trailing operating earnings per share for the S&P 500 alongside 25-month estimate paths mapped for every single quarter. Looking at these squiggles reveals a systematic pattern of analyst behavior: whenever the broader economy and corporate profits trend upward, analysts project absurdly optimistic trajectories. The estimate lines for individual quarters consistently slope downward over time as reality sets in, meaning the apparent "cheapness" of forward multiples often reflects ungrounded Wall Street euphoria rather than realized earnings power.

```mermaid flowchart TD accTitle: Structural Caveats in Equity Valuation accDescr: Flow showing the four primary caveats challenging current S&P 500 valuation models.

subgraph ForecastRisks["Forecast & Structure"] A["1. Systematic Analyst Optimism ('Squiggles' revisions trend down)"] B["2. Mega-Cap Concentration (Mag 7 drives 1/3 of the index)"] end

subgraph AccountingRisks["Earnings Composition"] C["3. Margin Expansion vs. Sales (Earnings outpace 15.6% rev growth)"] D["4. Capitalized AI Spending (Capex on balance sheet, not P&L)"] end

The second caveat is the unprecedented concentration of the S&P 500. While ostensibly representing 500 diverse businesses, index returns and aggregate earnings are disproportionately dictated by the "Magnificent 7" (Alphabet/Google, Amazon, Apple, Meta, Microsoft, Tesla, and Nvidia). These seven giants account for nearly a third of the index's total market capitalization and command far higher valuation multiples than the remaining 493 components. Furthermore, private AI flagships like OpenAI and Anthropic are not yet public; insider reports indicate they operate with massive net cash burn. Once such companies eventually tap the public markets, their unprofitable structures could drag on aggregate index earnings and distort multiples even further.

The third issue lies in how corporate America plans to hit these aggressive profit goals. Forward projections show revenue growing by only 15.6%—respectable, but nowhere near the 27%-plus earnings growth penciled in for the bottom line. This implies that the bulk of expected earnings growth stems from severe cost reductions, primarily achieved through AI adoption. This assumption has a clear expiration date: cost-cutting can only go so far, and the underlying AI providers delivering these enterprise efficiencies cannot subsidize corporate America forever. Once loss-making AI startups adjust their pricing to sustainable unit economics, enterprise software fees will climb, narrowing the profit margins corporations currently claim to have permanently unlocked.

- **Analyst revision bias**: Yardeni Research’s "squiggles" demonstrate that analysts consistently set over-optimistic targets across the 25-month rolling forecast window, later revising them down as quarterly results approach. - **Top-heavy index**: The Magnificent 7 represents roughly one-third of the S&P 500 by weight and trades at a noticeable premium over the broader constituents. - **Top-line mismatch**: Revenue is projected to grow by just 15.6%, meaning bottom-line earnings targets rely almost entirely on aggressive, unsustainable cost reductions. - **AI vendor dynamics**: Free or subsidized enterprise AI tools will eventually transition to profitable commercial models, eroding corporate margin savings.

---

[15:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=900s) Hidden Accounting Traps: Capex Depreciation and The Gravity of Bond Yields

The fourth caveat concerns the accounting treatment of artificial intelligence infrastructure. Most corporate AI outlays are classified not as operating expenses on the income statement, but as capital expenditures (Capex) parked on the balance sheet. Massive cash sums spent procuring advanced graphics chips and constructing colossal data centers are treated as multi-year balance-sheet investments to be amortized over time rather than expensed upfront. This accounting convention keeps current income statements looking pristine, even while companies pour billions into hardware. If advanced AI chips suffer physical obsolescence or commercial depreciation far faster than the lengthy schedules assumed by corporate accountants, companies will eventually face painful write-downs, revealing that current earnings figures are overstated.

```mermaid flowchart LR accTitle: Equity Risk Premium and Treasury Yield Dynamics accDescr: Relationship showing how higher bond yields compress the equity risk premium.

subgraph ERP["Equity Risk Premium Calculation"] EY["Earnings Yield (1 / P/E Ratio)"] TY["10-Year Treasury Yield (Risk-Free Rate)"] ERPVal["Equity Risk Premium (~4%)"] EY ---|"Minus"| TY TY -->|"Yield spike"| ERPVal end

subgraph Impact["Market Mechanism"] ERPVal -->|"Lowest in years"| CapitalShift["Less incentive to hold equities"] CapitalShift -->|"Acts as gravity"| Valuations["Multiple Compression for AI Tech"] end ```

The final lens through which to judge the market is the Equity Risk Premium (ERP). Calculated by taking the earnings yield (the inverse of the P/E ratio, representing the earnings return achieved when buying an index at prevailing prices) and subtracting the 10-year Treasury yield, the ERP gauges the marginal reward an investor receives for bearing equity volatility over risk-free government debt. Today, the equity risk premium hovers around 4%. While not quite as depressed as during the peak madness of the dot-com bubble, it sits at its lowest point in years. Investors are accepting near-record-low additional compensation to own stocks over safe Treasury bonds.

The primary culprit behind this compressed ERP is the sharp run-up in 10-year Treasury yields. Elevated yields act as a persistent gravitational pull on equity valuations, drawing institutional capital out of risk assets and into guaranteed paper. This dynamic presents real danger for capital-hungry AI ventures that depend on boundless investor liquidity to fund ongoing infrastructure expansion.

Host Richard Coffin concludes on an intentionally grounded note: none of this is an argument for trying to time the market, nor does it rule out the possibility that AI productivity delivers unprecedented corporate wealth that "delivers us to Valhalla." Instead, it highlights the extreme fragility of the consensus narrative. Analysts have baked near-flawless execution into equity prices, leaving investors with very little margin for error. As Coffin wraps up with trademark self-deprecating humor:

> "And sure, maybe I didn't actually clear anything up for anyone, but hopefully it gives you a more nuanced understanding of where we are today... Then, hey... We had fun together, didn't we? What a fun journey that was together, huh? Be safe out there."

- **Capex vs. OpEx distortion**: AI hardware and data centers are capitalized on balance sheets instead of expensed immediately, masking corporate cash burn. - **Depreciation mismatch**: Faster-than-anticipated obsolescence of cutting-edge chips poses a future risk of massive balance-sheet write-offs. - **Compressed ERP**: The Equity Risk Premium (Earnings Yield − 10-Year Treasury Yield) sits near 4%, offering minimal excess yield for bearing stock volatility. - **Bond yield gravity**: High risk-free yields create an attractive alternative to equities, specifically threatening debt- or equity-dependent AI buildouts.

## Notable quotes > [00:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=0s) Wait a minute, that's me — Richard Coffin

> [00:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=0s) tech bros flew too close to the sun — Richard Coffin

> [04:51](https://www.bilibili.com/video/BV1mBpg6iEbx?t=291s) dot-com crash 2.0 — Richard Coffin

> [07:37](https://www.bilibili.com/video/BV1mBpg6iEbx?t=457s) We're not done beating this dead horse till I say so. — Richard Coffin

> [12:12](https://www.bilibili.com/video/BV1mBpg6iEbx?t=732s) Now I know what you're thinking, by God, Richard, we've been all over the place now. Just give it to me straight. Is the S&P 500 expensive or not? — Richard Coffin

> [12:12](https://www.bilibili.com/video/BV1mBpg6iEbx?t=732s) squiggles — Richard Coffin

> [15:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=900s) delivers us to Valhalla — Richard Coffin

> [15:00](https://www.bilibili.com/video/BV1mBpg6iEbx?t=900s) And sure, maybe I didn't actually clear anything up for anyone, but hopefully it gives you a more nuanced understanding of where we are today... Then, hey... We had fun together, didn't we? What a fun journey that was together, huh? Be safe out there. — Richard Coffin

## Mentioned The Plain Bagel, Richard Coffin, S&P 500, SpaceX, Tesla, Tesla Optimus, Ed Yardeni, Yardeni Research, Robert Shiller, S&P MidCap 400, S&P SmallCap 600, MSCI ACWI ex USA, Alphabet, Google, Amazon, Apple, Meta, Microsoft, Nvidia, OpenAI, Anthropic, U.S. Department of the Treasury

## Video description 【The Plain Bagel】美股估值逆天？标普500到底是贵还是便宜？ How BAD Are Stock Valuations Really？

美股最近涨得这么疯狂，难道真的身处巨大的泡沫之中吗？作为财经博主，今天我们不聊虚的，直接用全网最硬核的多维估值指标，把标普500指数的底裤扒个干净。无论是席勒市盈率、动态市盈率，还是PEG和股权风险溢价，我们用真实的数据和严谨的逻辑来看一看，现在的市场估值究竟处于什么历史水准。

市盈率指标的巨大分歧当我们用传统的相对估值法来审视标普500指数时，会发现不同的市盈率指标给出了完全截然相反的信号。如果看大名鼎鼎的席勒市盈率（CAPE，即周期调整市盈率），目前的数值已经飙升到了40倍以上，创下了自2000年互联网泡沫破裂以来的历史次高水平。这是因为CAPE通过平滑过去10年的通胀调整后收益，过滤掉了短期经济周期的波动。但这也带来了弊端：在经历爆发式增长的时期，它会严重低估近期的盈利增长，从而高估整体估值。相比之下，如果采用过去12个月的简单滚动市盈率，虽然目前同样处于较高位置，但距离当年互联网泡沫巅峰时期的疯狂仍有明显差距。这意味着，现在的龙头企业背后至少有真实的净利润在支撑，并非全靠纯粹的空气支撑。

前瞻估值与意想不到的PEG 如果把目光投向未来，用Ed Yardeni Research提供的未来12个月一致预期市盈率来看，目前的数值大约在19倍左右，虽然高于16倍的长期历史平均水平，但也远低于互联网泡沫时期25倍的峰值，并且在过去一年里一直在缓慢回落。更让人跌眼镜的是前瞻PEG指标。根据分析师一致预期，目前标普500成分股未来5年的复合盈利增长率飙升到了27.3%，这使得标普500的向前PEG仅为0.69，处于几十年来罕见的低位。换句话说，如果分析师的预测成真，那么结合惊人的成长性来看，目前的标普500指数甚至可以说是被低估的。

大小盘及全球市场的估值对比为了更全面地评估，我们把标普500与其它市场做个横向对比。如果你去看代表中小型企业的标普400中盘股和标普600小盘股，会发现它们的向前PE分别只有15倍和14倍左右，明显比接近20倍的标普500便宜。然而，从PEG角度来看，标普400和600的PEG分别为0.8和0.96，依然没有标普500的0.69来得低。不仅如此，如果我们把目光转向美国以外的全球成熟及新兴市场，其向前PE仅为12.5倍，同样预期未来几年会有强劲的盈利增长。这意味着虽然美股独领风骚，但全球其他地区正提供着更高的安全边际与估值折扣。

不可忽视的四大结构性隐患尽管前瞻指标看起来很美，但我们在实际操作中必须清醒看到四大风险。首先是预测的盲目乐观。分析师在经济上行期往往带有系统性的过度乐观倾向。其次是严重的市场集中度。以谷歌、亚马逊、苹果、微软、Meta、特斯拉和英伟达为首的“七巨头”不仅占据了指数近三分之一的权重，而且它们的估值明显高于其余493家公司。第三，目前的盈利大增主要靠企业降本增效而非疯狂的收入增长，这种靠人工智能大幅压缩成本的红利未来可能随着AI使用成本的上升而边际递减。最后，海量的AI基建支出目前大量计入资产负债表的资本支出（Capex），并没有完全体现在利润表的费用端。如果芯片等设备的实际折旧寿命比会计假设更短，那么真实的利润水分将会被大大挤压。

股权风险溢价与操作指引审视当前市场的终极标尺是股权风险溢价（ERP）。通过计算当前市盈率倒数的收益率，再减去10年期美债收益率，我们发现目前的动态股权风险溢价仅为4%左右。这不仅是近些年来的最低点，也意味着投资者承担股票市场的风险所获得的超额回报被严重压缩，美债的高收益率正对股市估值构成明显的“引力”压制。面对充满极端分歧和极高预期的美股市场，盲目跟风狂欢或盲目恐慌都不可取。

审视美债收益率变化：密切监控10年期美债收益率的波动，当无风险收益率持续走高时，会直接挤压高估值AI科技股的估值水分。警惕盈利预期的水分：不要盲目迷信分析师给出的大幅盈利增长预期，重点甄别企业是靠真实的业务收入增长还是单纯的短期降本增效。做好仓位管理与分散：由于标普500指数高度集中在少数巨头手中，且整体股权风险溢价处于历史低位，切忌过度押注单一资产，可适度关注中小盘股及国际市场的对冲机会。

简洁时间轴 [00:00] 市场关于美股AI泡沫的争论与核心疑问 [02:54] 标普500指数的估值基准与指标选择 [04:51] 席勒市盈率与滚动市盈率的真实历史水位 [07:37] 前瞻市盈率与惊人低位的PEG指标解析 [10:07] 中小盘股及海外国际市场的估值横向对比 [12:12] 深度拆解美股估值背后的四大结构性隐患 [16:33] 股权风险溢价与当前市场面临的风险总结

[Source](https://www.bilibili.com/video/BV1mBpg6iEbx)

中文简报

美股估值到底处于何种水平：标普500究竟是贵还是便宜？主持人Richard Coffin通过对比市场上相互矛盾的估值指标，评估了标普500指数是否处于历史性高估状态。尽管席勒市盈率（Shiller CAPE）等滞后指标已超过40倍，逼近互联网泡沫时期的极值，但分析师激进预测的27.3%年化盈利增长率，却使其远期PEG比率仅为0.69，在表面上显得十分便宜。然而，这种表面上的低估严重依赖于分析师极度的乐观偏见、市场对“科技七巨头（Magnificent 7）”的高度集中、企业削减成本带来的不可持续的利润率扩张，以及掩盖了实际现金消耗的资本化AI基础设施支出。最终，随着10年期美债收益率上升将股权风险溢价压缩至约4%，当前市场定价实际上要求企业展现出近乎完美的执行力，几乎不容许任何差错。标普500指数的席勒市盈率已攀升至40倍以上，接近互联网泡沫时期的历史高位，但由于该指标平滑了过去10年的历史数据，在同比盈利增长约30%的时期受到了严重扭曲。 (4:51) 将华尔街激进的5年期27.3%年化盈利增长预测考虑在内后，标普500指数的远期PEG比率降至0.69；尽管远期市盈率高达约19倍，却在表象上释放出数十年未见的低估信号。 (7:37) 尽管标普400中盘股和标普600小盘股的远期市盈率相对更便宜，分别约为15倍和14倍，但MSCI ACWI ex USA中的国际股票估值优势最为显著，远期市盈率仅为12.5倍，且伴随着强劲的增长预测。 (10:00) 美国股权风险溢价已压缩至约4%，接近数年低点，高企的10年期美债收益率降低了持有高波动性股票的相对吸引力，并对重资本投入的AI基础设施项目构成威胁。 (15:00)

Reading summary

《The Plain Bagel》主持人理查德·科芬指出，标普500指数的估值看似自相矛盾：基于历史回溯指标显得昂贵，基于未来增长预测又看似便宜，但支撑这种“便宜”的华尔街盈利预期存在严重结构性缺陷。指标悖论：席勒市盈率与远期PEG的分歧 (4:51) 标普500指数在通胀居高、美债收益率上升和油价波动的背景下屡破新高，但不同估值模型呈现出截然相反的结论。从回溯性指标来看，席勒市盈率（CAPE）已突破40倍，处于接近互联网泡沫顶峰的历史极端高位；不过，该指标平滑过去10年通胀调整后利润的计算方式，严重低估了近期标普500成分股连续三个季度超过25%（三季度近30%）的高速盈利增长，且滚动市盈率（TTM P/E）仍明显低于2000年峰值。转至前瞻性指标时，市场面貌发生反转。Yardeni Research的数据显示，标普500的一年远期市盈率约为19倍，虽高于16倍的长期均值，但远低于互联网时期的25倍。结合分析师给出的未来5年高达27.3%的年化复合盈利增长预期，远期PEG比率骤降至约0.69（低于1.0通常被视为低估），使该指数在表面上呈现出数十年罕见的廉价特征。横向对照：中小盘与海外市场的相对估值 (10:00) 若将标普500与美国其他市值梯队比较，标普中盘400指数和标普小盘600指数的远期市盈率分别为约15倍和14倍，绝对倍数明显低于标普500（近20倍）。然而一旦引入增长预期进行PEG调整，标普中盘的PEG为0.80，小盘为0.96，而标普500凭借极高的增长预测反而以0.69成为三者中最低。放眼全球市场，估值分化更为显著。涵盖22个发达市场与24个新兴市场的MSCI ACWI ex USA指数，其远期市盈率仅为12.5倍。尽管海外市场盈利预期同样呈现强劲上修态势且预计年化增速与标普500相当，但其估值依然维持着深度的折价状态。华尔街预测偏差与指数集中度隐患 (12:12) 低远期估值的背后潜藏着多项结构性漏洞。首先是分析师的系统性乐观偏差：Yardeni Research追踪预测轨迹的“曲线图”（squiggles）显示，分析师在盈利上升周期中往往给出夸张预期，随后随着时间推移逐季下调目标。其次是指数高度集中的问题，“七巨头”（Mag 7）占据标普500近三分之一权重，其高估值推升了整体水平并掩盖了其余493家公司的表现；未来如OpenAI等处于巨额现金消耗阶段的未上市AI企业若挂牌，还可能拖累整体净利。此外，企业利润与收入增长存在明显脱节。标普500营业收入预期增速仅为15.6%，远落后于27%以上的净利润预测，这意味着利润增长高度依赖AI带来的成本削减。然而随着亏损的AI基础供应商逐步转向盈利模式并上调服务费用，企业的成本削减空间将面临收窄极限。会计折旧风险与被挤压的股权风险溢价 (15:00) 巨额AI基础设施开支的处理方式也构成了潜在的会计隐患。目前硬件采购和数据中心建设主要作为资本支出（Capex）计入资产负债表，而非直接反映在利润表中的运营成本，这在短期内粉饰了当期利润。如果高算力芯片由于技术迭代导致物理或商业淘汰速度快于会计折旧年限，企业未来或将面临剧烈的资产减记，导致现有盈利失真。更为关键的是无风险利率的引力效应。由于10年期美债收益率上升，衡量股票投资回报相对无风险资产补偿的股权风险溢价（ERP，即盈利收益率减去10年期美债收益率）已降至约4%的历史低位区间。这不仅降低了投资者持有权益资产的动力，也对极度依赖外部资本维持高额基建投入的AI科技生态构成估值重压。理查德·科芬强调，剖析估值并非建议投资者择时离场，也未否定AI生产力创造财富的潜力，但这凸显出当前市场对“近乎完美执行”的定价极其脆弱，留给投资者的安全边际十分有限。

Source quotes

[0:00] Richard Coffin: Wait a minute, that's me [0:00] Richard Coffin: tech bros flew too close to the sun [4:51] Richard Coffin: dot-com crash 2.0 [7:37] Richard Coffin: We're not done beating this dead horse till I say so. [12:12] Richard Coffin: Now I know what you're thinking, by God, Richard, we've been all over the place now. Just give it to me straight. Is the S&P 500 expensive or not? [12:12] Richard Coffin: squiggles [15:00] Richard Coffin: delivers us to Valhalla [15:00] Richard Coffin: And sure, maybe I didn't actually clear anything up for anyone, but hopefully it gives you a more nuanced understanding of where we are today... Then, hey... We had fun together, didn't we? What a fun journey that was together, huh? Be safe out there.

Original transcript

[0:00] The number one question in stock investing right now, is AI going to kill us? Followed closely thereafter by is AI in a bubble? The AI bubble is bursting. South Korea's AI bubble just popped. The AI bubble just burst. The AI bubble video bubble is about to pop. Wait a minute, that's me. Regardless, get my point. Many people are worried that the stock market, specifically in the United States, is in a bubble. Tech companies have come to dominate major indices while spending hundreds of billions of dollars on infrastructure for AI companies that are ultimately unprofitable, all while facing headwinds from rising oil prices, treasury yields, and heightened inflation. Meanwhile, the stock market has continued to reach new all-time highs, which is why many people have come to the conclusion that the market surely is in a bubble. Surely, valuations are stretched. But are they really? Don't get me wrong, there are a lot of individual AI companies walking around with pretty questionable valuations. SpaceX is a prime example, breaking into the top 10 list of largest market caps in the US, despite its revenues not even cracking the top 200 list. And there are plenty of red flags, plenty of reasons why the current stock market run-up is not sustainable. But whether the market as a whole is in a bubble or not is actually quite difficult to determine because... It depends on what you look at. And some people will argue, hey, listen, yes, AI stocks are expensive, but consider the massive potential they face. The businesses that are in the space with high valuations now might be set to reap the biggest reward from the changes to come. Now, no one knows what the future holds. We're going to end up somewhere between the two extremes of the utopia, where everyone retires rich with an optimist robot folding their laundry, and Armageddon, where AI takes over and there's not even an economy left because tech bros flew too close to the sun. you But truly, there are just too many variables to accurately predict what happens from here. So I thought instead, let's narrow our scope. Focus on one aspect here. Is the market, namely the S&P 500, expensive based on historical standards? And well, even here, that question proves more difficult to answer than you might expect. Given, again, it depends on where you look. Charts like this one definitely suggest that things are overpriced. It could convince most people that we are in a bubble. suggest that things are actually pretty attractive. And regardless of your view, it's important to try and understand why people are on the other side. For example, still buying when so many people think we are in a bubble. So today, let's go through the exercise of trying to answer this question, of comparing all these different metrics to better understand why the market looks the way it does. And while it might not give us an explicit yes or no answer, it will give us a more nuanced S&P 500, the index we'll be using today, is naturally 500 of the largest U.S. stocks by market capitalization, but not necessarily the 500 largest U.S. companies, given that there's other inclusion criteria, including time and earnings requirements, which is why companies like SpaceX are currently not in the index. for today's exercise. But how exactly do we value an index? Because the level alone doesn't tell us a whole bunch. Yes, the S&P 500 has doubled over the past four years, but to put that into perspective, we have to compare it to some fundamental metric of how the underlying companies are performing or how they're expected to perform in the future. And now normally with an individual company, you could do something like a discounted cash flow, where you estimate all future cash flows a company is going to create, you discount them based on the timing, and then you add that all together and make a few adjustments to come to some estimated with an index, that's not really feasible. Not only do we have 500 companies to do that for, but the amount of estimates we'd be throwing in here would make such a large margin of error that ultimately it would render any value we come to pretty useless. So instead we can do a relative valuation where we compare the index level to again some singular market cap weighted average of some fundamental metric across all 500 companies. Today we're going to use the most common one, Earnings, which is used commonly for the price to earnings or P/E ratio of individual stocks, where you take the price of a company's stock and divide it by the earnings per share, the profit they have after accounting for all of their expenses. And if you take this metric for each of the 500 different companies and you calculate a market cap weighted average of the bunch, then you'll come to a P/E for the S&P 500. So, how does that look? Well, again, it depends on what you're talking about. If you look at this PE for example, the Shiller's PE or Cyclically Adjusted Price to Earnings Ratio, CAPE for short, [5:00] probably convince most people that we're in for the dot-com crash 2.0. We are currently above 40 times a level not seen since the dot-com bubble, and the metric is fast approaching a new all-time high. Which, yeah, is pretty alarming. But one important point is that the CAPE ratio, as the name suggests, is cyclically adjusted. There's actually an earning smoothing factor that reduces the year-over-year volatility. What that means is that to calculate the CAPE ratio, we actually take the price and divide it by the average inflation-adjusted earnings over the past 10 years, with the idea being that this will help to ensure that short-term factors like where we are in the business cycle don't unnecessarily skew our rating of a company's valuation. It gives us a more long-term view of the stock's value. However, that also means that we're putting less weight on more recent earnings, which matters if we encounter a period of rapid earnings growth, because it will understate the impact of that earnings growth and might overstate the valuation of these stocks. Something that we are currently experiencing. In Q3 of this year, for example, the S&P 500 saw its earnings grow by nearly 30%, the third consecutive quarter of growth above 25 year over year For reference the 25 average earnings growth for the S 500 has been 6 Now let not argue that this is sustainable Certainly that figure will come down at some point. A lot of the short-term growth is related to the AI build-out, and at some point, as with the internet build-out, we're not going to have three to four data centers popping up a week in the US. But if you subscribe to this idea that AI is this revolutionary technology that will unlock a company's ability to grow more rapidly into the future, then the CAPE ratio might again overstate a company's valuation. And you might want to consider something like a more simplified trailing PE, which only takes earnings from the past 12 months. Naturally, you'll see the figure is much more volatile since we aren't smoothing things, which is why it skyrocketed here in 2008. But today, the measure, while certainly more elevated, is well below what we saw during the peak of the dot-com bubble. So already, we're starting to see a bit of a different picture. Yes, things do seem expensive, but not as far stretched as they were during the dot-com bubble. They at least have earnings behind these companies to justify some of this performance we've seen. But so far we only looked at trailing valuations pricing a company based on what it done in the past In truth most investors invest in stocks based on what they expected to do in the future So we should also consider forward-looking P.E. ratios, which take the price of a stock and divide it by the next 12 months of earnings. Now here we run into a similar problem as with the discounted cash flow model, which is that we're talking about the future, and no one knows what the future holds for many of these companies. Nonetheless, people do try to estimate the earnings a company will see over the next several years. And we can take what's called the consensus estimates for all of these companies. Consensus just meaning the average of all the analyst forecasts for these earnings. Now, there's no one single source for consensus estimates. So depending on where you pull it, you might get a slightly different figure. But I'm going to be using the numbers from Yardini Research, which is a platform run by the famous economist Ed Yardini, who actually has a financial model named after him. This is not a paid promotion, but you're at any research. Excellent free resource with a lot of charts published daily about the stock market. And you can see here that they have the forward PE for the S&P 500 based on consensus estimates. And if you take a look, you'll see that the metric is indeed heightened at around 19 times, above the long average of 16 times However it is again below the dot peak of 25 times and it actually been on the decline over the past year Finally since I know you guys are having a blast here, we're going to look at one more forward metric, which is the forward peg ratio, the forward PE for these companies divided by their five-year estimated earnings growth rate, with this ratio giving us an idea of whether a company's valuation is justified. Any ratio above one is generally considered on the more expensive side, whereas ratios below one are generally considered cheaper. You're growing faster than the valuation might otherwise suggest. And this is going to be the most controversial metric here because it actually suggests that the S&P 500 is cheap. And not just cheap, but the cheapest it's been in decades. So trailing valuations certainly are elevated, but this suggests that analysts are expecting the growth of earnings for these companies to actually outpace these higher valuations. And sure enough, if you look at the consensus five year earnings growth estimate, it has surged over the past two years and currently sits at 27.3%. So the S&P 500 is actually cheap. End of story, right? Wrong. We're not done beating this dead horse till I say so. So far, we're not done being [10:00] We've only categorized things as expensive or cheap based on their historical performance. We can also compare the index's valuation levels to other indices to get an idea of how they stand in the group. For example, we can look at the S&P 600 and the S&P 400, which are the S&P indices of U.S. stocks in the small cap and the mid cap range, respectively. So small and medium sized companies. times versus the S&P at nearly 20 times. But again, if we look at the peg ratio, the S&P 500 is actually on the low end. This chart is a little crowded, but you can see the S&P 400 has a peg of 0.8, the 600 is at 0.96, and the 500 is the lowest. at 0.69. Finally, of course, we can compare the S&P 500 to other international markets to see how the US stands against other countries. Here we'll use the MSCI All Country World Index, x the United States, which, as the name implies, is an international index with 22 developed and 24 emerging countries. capturing 85% of global equities outside the U.S. Now, while the U.S. market has gotten a lot of attention over the past several years, this index has actually itself done quite well over 2025 and this year to date. But how does its valuation look? Well, with the forward PE, you can see that the index is actually fairly cheap by historical standards at 12.5 times, with a big part of this being that earnings estimates for this index are themselves surging, growth over the next few years. And now we don't have the same five-year earnings average growth rate that we do with the U.S. to compare peg ratios, but you can see that international markets are actually expected to grow earnings at a similar pace to the S&P 500, despite trading at a pretty meaningful discount. Now I know what you're thinking, by God, Richard, we've been all over the place now. Just give it to me straight. Is the S&P 500 expensive or not? Correct. But to further our understanding of the valuations here, we need to finish things off with a few important caveats, constraints, and considerations. The first biggest one, again, being that when we're talking about forward valuations, we're relying on forecasts. which we all know how great people are at those. In fact, Yardini Research has a graph that illustrates this perfectly called squiggles. This might be a bit confusing at first, but this line here is actual operating earnings per share for S&P 500 companies. And each squiggle represents estimates for one data point over a 25 month period leading up to the release of S&P 500. of that data point that extends a bit beyond the actual date, because there's a delay between when a quarter finishes and when the results are announced. But you get the point, each of these lines is for one data point being estimated over time. And one thing you'll notice beyond the fact that the estimates move quite a bit, is that analysts are often too optimistic when things are going well. Most of these squiggles during the upward slope in the purple line themselves have downward slopes. So forecasts can only get us so far. The second thing of course to mention is S&P 500 concentration. Despite covering 500 different companies, the index's performance is largely determined by just a handful of companies. For example, you might have heard of the Magnificent 7, which includes Google, Amazon, Apple, Meta, Microsoft, Tesla, and NVIDIA. They the entire index, and you can see that they have above average valuations compared to the other 493 constituents. We also don't yet have stocks available for the AI poster kids, OpenAI or Anthropic, but those companies are likely to skew things even further, especially given that insider reports suggest highly unprofitable operations. So when those companies come to market, that's going to impact valuations further. Now the third thing I'll highlight is that this massive earnings are forecasting seems to be coming predominantly from cost savings rather than rapid revenue growth. This chart here breaks it down quite well. You can see that revenue growth is actually only forecasted to grow 15.6%, which is still a lot to be sure, but it suggests that most of the earnings growth is coming again from cost savings, which can only move so much. We also have to consider that if these cost savings are coming from AI adoption, that at some point they'll likely be diminished by rising AI usage prices. We know that most AI companies are unprofitable. At some point, they're going to have to have a more sustainable pricing model. And again, the current high growth we have seen is likely not sustainable long-term. The CAPE ratio was actually invented to help see past these periods of highlighting here is that costs may be underrepresented in the earnings figures we're using here. Most of the spending in AI that we've seen [15:00] some predominantly in the form of capital expenditures. buying chips, investing in data centers, things of that sort that don't actually appear on the income statement as an expense, but rather go into the balance sheet as an investment, something that will be monetized at a later date that isn't considered an upfront cost, which is important because a lot of these companies have been spending all of their earnings on AI investments and some question the assumptions going into these accounting practices. For example, there are concerns that the chips these companies are buying will have a shorter accounting assumptions call for. It's why you need to treat the P/E ratio with a bit of skepticism. We're effectively trying to value the entire U.S. stock market off of A, a metric that could be skewed, and B... oftentimes one year of performance, which may or may not offer a good gauge of all future performance from here on out. So we've said a lot today, and at the same time, nothing at all. But there is a takeaway here, that being that analysts are expecting pretty meaningful earnings growth from AI here. And if things work out, even given the current stock valuations we have, stocks could experience solid performance or even good performance from here. But the performance we've seen today is pretty abnormal from a historical basis and there is a lot of upside already priced into stock valuations today but which leaves investors with a lot of risk. I'll treat everyone here with one more metric, the equity risk premium, which takes the inverse of the PE ratio called the earnings yield, effectively your earnings return that you achieve buying a stock at its current price, and you subtract the 10-year treasury yield. It essentially shows the marginal return investors receive for investing in stocks rather than just in treasury bonds and collecting that premium today is only around 4%, which is not as low as it was during the dot-com bubble. but it's still the lowest it's been in some time. What that means is that investors are taking on equity risk, investing in the stock market rather than just treasury yields, for a lower marginal equity return effectively. Now this is because the 10-year treasury yield has spiked recently, but that itself is also a risk faced by the market, given that higher treasury yields tend to act as a gravity for stock market valuations. As yields rise, it can pull investor capital out of the stock market, which can be quite meaningful for AI companies that are currently relying on investor capital to fund their growth. Now, the flip side argument is that AI faces a remarkable upside here. That's not going to be captured by any of these metrics. And none of this is to argue that people should be trying to time the market. If anything, the hope is to communicate just how complicated it is to answer even a very basic single facet of what's going on here. other AI execs are going to deliver us to Valhalla. And sure, maybe I didn't actually clear anything up for anyone, but hopefully it gives you a more nuanced understanding of where we are today, Then, hey... We had fun together, didn't we? What a fun journey that was together, huh? Be safe out there.