The number arrives like an anomaly flagged by a monitoring script: S&P 500 earnings are running 14% above the long-term trend. First time since 1955. Seventy years of data — recessions, wars, productivity revolutions, policy failures — and no cycle has pushed corporate profitability this far from its gravitational center. s silence. The market celebrates. The ledger does not celebrate; the ledger states facts.
I have a professional reflex that makes me distrust headline statistics. In 2017, I traced 450,000 ETH transfers from ICO crowdsales, cross-referencing every transaction against known exchange deposit addresses. The headline said "decentralized community." The wallet clustering data said that 68% of early token ownership flowed through interconnected entities. The narratives and the data diverged.
That experience constructed my methodology. Whenever a number claims to be unprecedented, the first question is not "what does it mean?" It's "how was this measured?" And the second question is: "what's the failure threshold?"
The Measurement Problem
The statistical foundation of the "14% above trend" claim is narrower than it appears. "Long-term trend" is an estimation, not a fact. A linear regression over the 1955-2025 period produces one trend line. A Hodrick-Prescott filter produces another. A trailing moving average produces a third. The deviation shifts significantly based on the anchor point.
The earnings metric itself is equally ambiguous. GAAP net income, operating EPS, and per-share operating cash flow each tell different stories. GAAP earnings include one-time writedowns and structural impairments. Operating earnings strip them out — but often exclude stock-based compensation. Per-share figures are distorted by a decade-long buyback machine that has mechanically retired roughly 20% of outstanding shares. A company that holds net income flat while repurchasing 3% of its shares annually will show 3% EPS growth with zero operational improvement. That's not a hypothesis; that's arithmetic.
Using public data to cross-check the claim: TTM operating EPS for the S&P 500 sits in the $250-270 range as of mid-2025. A log-linear trend fitted from 1955 extrapolates to approximately $220-235 for the current period. The gap of roughly $30-35 per share aligns comfortably with the 14% claim. So the statistic is credible — under the specific assumption that "long-term trend" means an exponential time trend. The conclusion I draw: this measures actual earnings against a time-based trend, not analyst expectations or a cyclical base effect.
That conclusion matters because an earnings series that systematically outruns its own history is consuming future outcomes. Every period of elevated net margins requires a corresponding period of compressed margins to balance. The 1990-2019 average net margin for S&P 500 constituents was approximately 7.5-9.5%. Since 2021, that figure has held between 11-13%. The persistence of this elevated range is itself historically anomalous.
This pattern is not unfamiliar to me. When I audited Aave v1's interest rate model during DeFi Summer, I ran 10,000 simulated liquidation events using Python scripts. The model performed flawlessly under normal conditions. The edge case only appeared at extreme utilization — the boundary condition that the model's authors had optimized away. The 14% overshoot is a similarly optimized boundary. The question is what happens when conditions normalize.
The Four Forces
Decomposing the overshoot: four measurable forces account for the bulk of the deviation. A fifth — the measurement artifact — deserves its own scrutiny.
Force One: The Interest Rate Lock-In Effect
The driver that institutional investors most underestimate is the "rate illusion" embedded in corporate capital structures. During 2020-2021, investment-grade corporations refinanced their balance sheets at historic lows. A wave of fixed-rate issuance locked in coupons between 2-3% for maturities of five to ten years. When the Federal Reserve engineered the fastest tightening cycle since the 1980s — raising the federal funds rate from near zero to 4.25-4.50% by mid-2025 — the marginal cost of new debt soared. But the effective cost of existing corporate debt barely moved.
This creates a structural time lag in monetary policy transmission to the income statement. The Fed's stance has been restrictive in policy terms. The actual interest expense on the S&P 500's existing debt remains calibrated to 2021 pricing. Corporate interest coverage ratios look healthy not because companies are resilient, but because they are riding a refinancing wave that hasn't crested yet.
This is what I call the "survivor dividend." Companies that secured cheap capital in the pandemic era are not more productive. They are better positioned against the yield curve. That's a discrete, one-time advantage — not a compounding one. As the maturity wall approaches and low-coupon debt is replaced at higher coupons, the interest expense drag will appear in the income statement. Moody's and S&P data confirm that roughly 30% of investment-grade debt issued during 2020-2021 comes due between 2026 and 2028. The earnings headwind is scheduled, not speculative.
Logic is the only audit that never expires. The balance sheet audit will catch up with the refinancing floor as surely as the LUNA dashboard caught the reserve depletion I flagged three weeks before that collapse.
Force Two: The Fiscal Pulse
The second force operates at the macro level. The federal government has run deficits of 6-7% of GDP for two consecutive fiscal years. The CBO projected a $1.9 trillion deficit for FY2025 — nearly double the post-war average of approximately 3.5%. This is not a routine level of fiscal expansion. It is a structural injection of demand into the economy, and it lands directly on corporate income statements.
In a restrictive monetary environment, fiscal spending is the counterweight that prevents nominal demand from collapsing. But the distributional detail is critical. The profit share of national income has risen to approximately 25-26% — a post-war record. Labor compensation has fallen to 43-44% of national income, below pre-pandemic levels. The overshoot is not a generalized improvement in prosperity. It is a redistribution of national income toward capital.
My 2024 work on BlackRock's IBIT flows gave me direct visibility into this dynamic from the institutional side. 72% of daily ETF inflows were retained by the custodian — not recycled through trading desks. That's a behavioral inheritance of the post-2008 regime where capital preservation outperformed income generation. Fiscal expansion is a pipeline to that pool of capital.
There is also the industrial policy angle. The CHIPS Act's $52.7 billion in semiconductor subsidies and the Inflation Reduction Act's clean energy credits have begun flowing into corporate income statements. Manufacturing construction spending is running at 20-40% year-over-year growth — the highest level since the 1950s. This "policy-to-profit" pipeline is real, but its sustainability is entirely contingent on policy continuity. Political reversal is a risk the profit share has not priced.
The uncomfortable implication: the earnings overshoot is partially an artifact of government spending. If the deficit narrows — even by a percentage point of GDP — the fiscal pulse weakens, and with it the demand that supports the overshoot. The CBO projects persistent elevated deficits through the decade, which delays the reckoning but does not cancel it.
Force Three: The Asymmetric Pricing Window
The third force is the inflation structure itself. Producer prices have decelerated to approximately 1.5-2% year-over-year, while consumer prices remain at 2.5-3%. This spread — a positive PPI-CPI gap — operates as a direct margin subsidy. Companies pay less for inputs and hold output prices firm. The delta accretes to the bottom line.
The antitrust community has a loaded term for this: "greedflation." The more clinical framing is asymmetric pricing behavior. When costs fall, prices adjust downward with a lag. When costs rise, prices adjust upward immediately. This asymmetry is a well-documented pricing power mechanism, and it is currently delivering measurable, concentrated benefits to S&P 500 constituents.
But the window is finite. As inflation converges to target, pricing power diminishes. And the reverse asymmetry is already forming. Wage contracts signed at 4-5% nominal growth over the past two years will mature and reset. Raw material contracts with elevated floor prices will be renegotiated. When input costs rise while output prices are anchored by competitive disinflation, margin compression becomes structural rather than episodic.
This was the exact pattern my 2021 NFT wash-trading analysis exposed. When 450 interconnected wallets executed circular trades, the floor price of Bored Ape Yacht Club was inflated by approximately 40% of perceived volume. The artifact was not organic demand. It was coordinated pricing mechanics. When coordination breaks down — whether that's wallet clusters or cost-push dynamics — the price reversion is abrupt.
Force Four: The AI Concentration Effect
The fourth and most dangerous force is the concentration of earnings within AI-aligned mega-cap technology firms. Information technology and communication services now represent approximately 40% of S&P 500 earnings — exceeding the dot-com peak in 2000. The AI ecosystem — Nvidia, Microsoft, Alphabet, Meta, Amazon — contributes over 60% of the index's earnings growth. Strip out these names and the remaining 490 constituents sit significantly closer to their long-term earnings trend.
This concentration is visible in capital expenditure data. The four major hyperscalers are projected to exceed $300 billion in combined AI-related capex for 2025. Semiconductor fabrication plants are being built at a pace not seen since the 1950s. The physical infrastructure of the AI narrative is concrete. But the productivity evidence remains missing. Non-farm labor productivity grew at approximately 2.3% in 2024, above the 2010-2019 average of 1.4%. That is marginal improvement, not a J-curve inflection. Total factor productivity — the broadest measure of technological efficiency — has not yet reflected the AI investment levels. The "profit without productivity" gap is the structural weakness.
It is precisely here that my pre-mortem framework becomes essential. I built the LUNA collapse model by tracking a single ratio: stablecoin liquidity depth relative to market cap. The metric I flagged was straightforward — reserves falling below 60% of circulating supply. The warning was published three weeks before the crash. The same logic applies to AI capex: when capital expenditure growth turns negative year-over-year, the narrative's forward fuel is spent.
Force Five: The Measurement Artifact
The fifth contributor to the 14% is statistical illusion. Buybacks. Since 2010, S&P 500 companies have retired roughly 20% of outstanding shares. EPS — the metric against which "trend" is measured — inflates mechanically as the denominator declines. Additionally, the GAAP-to-operating earnings gap is nontrivial. Operating EPS excludes stock-based compensation in some calculations and amortization of intangibles in others. The two measures have diverged persistently since 2018, with operating earnings exceeding GAAP earnings by 8-12% in recent years. The reported "14% above trend" likely captures the operating earnings series — meaning the gap to a pure accounting-based earnings trend is meaningfully smaller.
This is not a criticism of the data. It is a criticism of the interpretation. Any investor reading the overshoot as an unambiguous signal of economic strength is missing the structural composition.
The Contrarian Read
The market's consensus interpretation is that earnings strength confirms economic resilience. The bond market tells a different story. Ten-year Treasury yields persist in the 3.8-4.5% range with a term premium that remains stubbornly positive. If corporate earnings are truly this robust, why do fixed-income investors demand elevated compensation for duration? A plausible answer: the bond market recognizes the earnings overshoot as a liability — a future reversion that the equity market has not yet priced.
There is also a reflexivity problem that mirrors the wash-trading patterns I've documented elsewhere. Strong earnings attract global capital into US equities. Capital inflows strengthen the dollar. A stronger dollar reduces the dollar-equivalent overseas revenue of the multinationals that dominate the index. The mechanism becomes self-limiting. At a dollar index of 104, the feedback loop is primed.
The overshoot also excludes the majority of the economy. ISM Manufacturing PMI hovers in the 48-52 range — contraction or stagnation. Commercial real estate vacancy rates are at cycle highs, with CMBS delinquency trends suggesting the next credit stress may originate there. The top 10 S&P 500 companies now account for a percentage of total index earnings that has exceeded the 2000 peak. This is a "K-shaped" earnings structure: an index that looks healthy because its largest constituents are thriving, while the underlying breadth of profit generation has been deteriorating.
For those who need historical parallels: 1968, 2007, 2017 — earnings overshoots preceding reversion. The one sustained exception was the late 1990s, driven by a genuine technological transformation. The open question is whether AI in 2026 resembles the internet in 1998 or the leverage cycle in 2007. The capex data says transformation. The productivity data says maybe. The valuation data — with the Shiller CAPE at approximately 35, below the 2000 peak of 44 but far above every other historical reading — says the market has already assumed the answer. Assumption is not evidence.
What to Track
The earnings overshoot is a warning, not a signal. The US equity market's vulnerability lies in the concentration of profit drivers: AI capex, fiscal stimulus, interest-rate lock-in, and asymmetric pricing. Each of these can unwind independently; in a bear market, they are likely to unwind together.
Track these thresholds. First: S&P 500 quarterly earnings growth below 3% for two consecutive quarters triggers reversion. Second: net margin below 11% signals the cost-squeeze has begun. Third: the first year-over-year decline in hyperscaler capital expenditure marks the AI narrative's failure event.
Logic is the only audit that never expires. The margin data will tell the truth before the headlines do. The 14% has been earned. The question is who holds the claim when the ledger rebalances. s silence. The data is already speaking. The only question is whether you are listening.