The Red Queen Problem

Good morning. Yesterday I graded JPMorgan's capital allocation. Today I want to show you why the grade itself might be based on bad data, because the numbers companies report are systematically misleading about something fundamental: how much of their spending is actually building the future, and how much is just keeping the present from falling apart.

---

THE EVIDENCE

The Red Queen Problem

In Through the Looking-Glass, the Red Queen tells Alice: "It takes all the running you can do, to keep in the same place."

Michael Mauboussin and Dan Callahan wrote a paper at Morgan Stanley in January 2022 called "Underestimating the Red Queen" that applies this line to corporate finance. The argument is simple and, once you see it, hard to unsee: investors overestimate how much of a company's spending goes toward growth because they underestimate how much goes toward maintenance.

A company reports $10 billion in capital expenditures. The market treats that as investment. Analysts model it as growth. The CEO puts it in a slide deck with an upward-sloping arrow.

But part of that $10 billion is just replacing equipment that wore out. Refreshing servers. Patching old software. Retaining employees who would leave without a raise. None of it builds anything new. It prevents decay.

The distinction matters because growth spending and maintenance spending have completely different return profiles. A dollar of growth spending, if deployed well, generates future cash flows that didn't exist before. A dollar of maintenance spending generates zero incremental cash flow. It just preserves what you already have. Conflating the two flatters the company's return on invested capital and misleads anyone trying to figure out what the business is actually worth.

---

THE DATA

The 20% Gap

How big is the gap between what companies report and what maintenance actually costs?

Venkat Peddireddy built the most comprehensive answer in his Columbia University PhD thesis. He studied U.S. companies from 1974 through 2016 and developed a way to estimate true maintenance capital spending. His method starts with depreciation and amortization, then adjusts for write-downs, asset impairments, and losses on asset sales . These are all signals that a company underestimated how fast its assets were losing value.

His conclusion: maintenance capex exceeds reported depreciation by about 20%, on average, across industries. If you're using depreciation as a proxy for maintenance spending, most investors do, because it's the only number the financial statements give you, you're understating the true cost of keeping the business running. That means you're overstating how much of the company's capex is genuinely going toward growth.

Peddireddy also found consequences. Companies that underestimate their maintenance spending experience future write-offs, earnings misses, and poor stock returns. The market catches up, but slowly.

There's a second layer that Mauboussin emphasizes. Capital expenditures aren't the only line item with this problem. Selling, general, and administrative expenses — SG&A — contain a hidden mix of investment and maintenance. When Google was young, its sales and marketing spending was almost entirely growth: signing up new advertisers, building brand awareness, hiring salespeople to cover new territories. Today, a large share of that same spending goes toward retention: keeping existing advertisers from moving to TikTok, replacing salespeople who quit, maintaining relationships that already exist. The budget line hasn't changed. The economic character of the spending has.

---

THE CASE STUDY

Alphabet's $185 Billion Question

In February, Alphabet told investors it expects to spend $175 billion to $185 billion in capex in 2026. That's nearly double the $91.4 billion it spent in 2025, and more than triple the $52.5 billion it spent in 2024. At the midpoint, Alphabet will spend roughly $180 billion on capital in a single year, more than the market capitalization of roughly 450 of the S&P 500's companies.

The market's instinct is to categorize most of this as growth. AI infrastructure. New data centers. Cloud capacity to serve a $240 billion backlog. And much of it is growth. But the Red Queen is already running.

The reported numbers. In 2025, Alphabet's depreciation expense was $21.1 billion, up 38% from $15.3 billion in 2024. That $21.1 billion is the accounting system's estimate of what it cost to maintain the existing asset base for one year. The capex-to-depreciation ratio was 4.3 to 1, which on its face suggests Alphabet is investing far more than it needs to just to stay in place.

But that ratio is misleading, for a reason that goes straight to the heart of Mauboussin's argument.

The accounting change. In 2023, Alphabet changed the estimated useful life of its servers from four years to six years. That single accounting change reduced depreciation expense by roughly $3.9 billion per year. The servers didn't actually last longer. The spreadsheet just said they did. The change boosted Alphabet's net income by about $3 billion annually, and it suppressed the depreciation number that investors use as a rough proxy for maintenance spending.

The tell. Alphabet's CFO told investors on the Q4 call that approximately 60% of capex goes to servers and 40% to data centers and networking equipment. Servers, on Alphabet's books, depreciate over six years. But the AI hardware cycle is moving faster than the accounting assumes. Amazon extended its own server useful lives from five years to six in January 2024, booking a $900 million boost to Q1 net income. One year later, in January 2025, Amazon partially reversed course, shortening the useful life of a subset of its servers back to five years, explicitly citing the pace of AI and machine learning development. That reversal cost Amazon $700 million in 2025 operating income. The GPU-accelerated servers going into these data centers today may be functionally obsolete well before their six-year accounting life expires.

That doesn't make Alphabet's spending foolish. The cloud backlog is real. The AI demand is real. But it means the share of the 2026 capex budget that constitutes genuine net new investment is smaller than the headline number implies. If true maintenance on the existing asset base is $28 billion to $31 billion rather than the reported $21 billion, then of the projected $180 billion in 2026 spending, the net new investment is closer to $149 billion than $159 billion. Still enormous. But 6% less than the market is probably modeling.

That gap compounds over time. Alphabet's CFO warned that depreciation expense will accelerate in 2026 as the 2024 and 2025 capex waves start flowing through the income statement. The Red Queen runs faster every year.

---

THE FRAMEWORK

Three Questions for Every Earnings Report

When you read any company's earnings this week, run them through Mauboussin's framework.

---

ALSO THIS WEEK

Wall Street built a tool to short private credit, and nobody is talking about it. S&P Global and a consortium of banks including JPMorgan, Goldman, and Bank of America launched the CDX Financials Index on Monday — the first credit-default swap index that includes business development companies, the public vehicles that fund a large share of the $3 trillion private credit market. CDS indexes have existed for decades, but they've never covered private credit. The fact that Wall Street felt enough demand to build one now tells you something about where the smart money thinks risk is accumulating. The trigger, reportedly, is a surge in redemption requests from private credit funds as investors worry that AI is disrupting the software companies those funds financed — outsourced customer service platforms, lower-tier coding tools, back-office automation vendors, the kind of mid-market SaaS businesses that private credit loaded up on between 2020 and 2024. If you've been reading about the AI capex boom as a story about winners, this is the other side: the companies being displaced by AI are sitting inside private credit portfolios, and someone just built a derivative to bet on the fallout.

The Texas Stock Exchange just landed its first primary listing, and it tells you something about where capital formation is migrating. Westwood Holdings Group filed a prospectus for an energy ETF that will launch on the TXSE — not on the NYSE, not on Nasdaq, but on a brand-new exchange backed by BlackRock, Citadel, Schwab, JPMorgan, and Goldman Sachs, with $270 million in capital behind it.

Three exchanges now compete for listings in Texas: NYSE Texas, Nasdaq Texas (launched in March), and the TXSE (expected to go live this summer). The capital allocation question here is structural: why are the biggest names in finance pouring money into breaking a duopoly that's existed for decades? Part of the answer is regulatory arbitrage — the TXSE plans listing standards that would disqualify roughly 1,500 companies currently on Nasdaq. Part is geography — one in ten U.S. public companies is headquartered in Texas, representing nearly $2 trillion in market cap. But the deeper story is that listing venue itself is becoming a competitive capital allocation decision, not just an administrative one. That's new. And the fact that the first company to choose the TXSE chose an energy product, not a tech one, tells you who the exchange is really being built for.

TSMC reports tomorrow. After today's discussion, their capex guidance deserves a different kind of scrutiny. The servers Alphabet is scrambling to replace are filled with silicon TSMC is scrambling to build. TSMC is building fabs in Arizona, Japan, and Germany. A leading-edge fab costs north of $20 billion and faces a constant treadmill of process node advancement. The Red Queen runs fast in semiconductors. When TSMC gives you a number, ask: how much is keeping up, and how much is getting ahead?

Tomorrow, I'm covering the $3 trillion in private equity dry powder and what it means for the deal pipeline these banks are fighting over. It's the Deployment.

---

SHARE THE CAPITAL MEMO

Know someone who makes capital allocation decisions for a living? Forward this email. When they subscribe through your referral link, you'll get The Capital Allocator's Checklist, a 10-question framework for evaluating any capital deployment decision, built from peer-reviewed research.

Your unique referral link is below. One referral unlocks the checklist.

Have a good Wednesday.

— Marques

---

Marques Blank is the founder of Blank Capital (fractional CFO and FP&A advisory) and Blank Capital Partners (registered investment advisory). Former Northrop Grumman ($1.6B business unit) and Citibank (securitized credit). CMA, MBA, Series 65.

This newsletter is for informational purposes only and does not constitute investment advice. Research cited: Mauboussin, M.J., & Callahan, D. (2022). Underestimating the Red Queen: Measuring growth and maintenance investments. Consilient Observer: Counterpoint Global Insights, Morgan Stanley Investment Management. · Peddireddy, V.R. (2021). Estimating maintenance CapEx. PhD Thesis, Columbia University. · Mauboussin, M.J., & Callahan, D. (2020). One job: Expectations and the role of intangible investments. Consilient Observer: Counterpoint Global Insights, Morgan Stanley Investment Management. · Mauboussin, M.J., Callahan, D., & Majd, D. (2022). Capital allocation: Results, analysis, and assessment. Consilient Observer: Counterpoint Global Insights, Morgan Stanley Investment Management.