Chart of the Day | Coatue Summary

1. Key Themes

Theme: AI capex is on track to exceed the scale of the largest US capital pools

AI capex projected to outsize mortgage originations

The chart puts 2028E AI capex across hyperscalers, labs, and neoclouds at $2.8T, above the $2.1T of 2026E US single-family mortgage originations. The spending is large enough to be compared with the core plumbing of the US financial system.

"AI capex in 2028E will be roughly… ~1.5x US mortgage originations"

AI capex dwarfs government and corporate spending benchmarks

AI capex is also compared with the defense budget and with the rest of the S&P 500's investment, which shows how concentrated and unprecedented the spending is.

"~3x US national defense budget" and "~3.5x S&P 500 non-AI capex"

Chart data: AI capex $2.8T (2028E) vs. US national defense budget $1.0T (FY26) vs. S&P 500 non-AI capex $0.8T (2025)

Theme: Financing infrastructure is the next bottleneck and opportunity

AI needs a standardized financing architecture

Coatue argues that spending at this scale requires an institutional financing layer like the one that standardized and scaled the housing market, which implies a major build-out in structured finance, securitization, and credit products tied to compute.

"AI capex will soon need its own Fannie & Freddie to standardize, finance and scale"

Headline question: "Will compute financing rival the mortgage market?"

2. Contrarian Perspectives

Compute financing could become an asset class on par with housing finance

The consensus framing treats AI capex as a corporate budget item funded from hyperscaler cash flows. Coatue instead frames it as a capital-markets problem needing GSE-style standardization, with mortgage-market scale as the reference point. That implies debt, securitization, and intermediary infrastructure will matter as much as the chips and data centers.

"AI capex will soon need its own Fannie & Freddie to standardize, finance and scale"

Coatue also hedges the claim in the footnote, noting that the comparison is illustrative and that the capex figures are not a debt-financing forecast.

"Comparisons are for illustrative purposes only and reflect different periods and categories of expenditure. AI capex estimates do not represent projected debt-financing demand."

3. Companies Identified

Fannie Mae & Freddie Mac (Fannie & Freddie)

Description: US government-sponsored enterprises that standardize, guarantee, and securitize mortgages.
Why mentioned: Used as the analogy for the financing infrastructure AI capex may need.
Quote: "AI capex will soon need its own Fannie & Freddie to standardize, finance and scale"

Hyperscalers, labs, and neoclouds (as a category)

Description: The cloud giants, AI labs, and newer GPU-cloud providers driving AI infrastructure spending.
Why mentioned: Their combined capex forms the $2.8T 2028E figure.
Quote: "AI capex: hyperscalers, labs, neoclouds (2028E)"

Coatue Management

Description: Investment firm and author of the newsletter and underlying public markets deck.
Why mentioned: Source of the analysis.
Quote: "See more from our latest public markets update deck."

4. People Identified

No individuals are named in this article.

5. Operating Insights

Build for the financing layer, not just the compute layer

If AI capex scales to mortgage-market size, the companies that standardize contracts, underwriting, and risk-sharing for compute assets could capture outsized value. Founders in fintech, structured credit, and infrastructure services should consider this white space.

"AI capex will soon need its own Fannie & Freddie to standardize, finance and scale"

Use scale benchmarks to frame the market for investors and customers

Comparing a new market to established capital pools (mortgages, defense, S&P 500 capex) is a clear way to communicate magnitude and legitimacy.

"AI capex in 2028E will be roughly… ~1.5x US mortgage originations … ~3x US national defense budget … ~3.5x S&P 500 non-AI capex"

6. Overlooked Insights

AI capex is already multiples of the rest of corporate America's investment

The chart implies that the AI buildout exceeds the capex of the entire S&P 500 outside AI by roughly 3.5x. This suggests a major concentration of capital allocation, with implications for macro growth, supply chains, and the cyclicality of one sector.

"S&P 500 non-AI capex (2025) $0.8" versus "AI capex: hyperscalers, labs, neoclouds (2028E) $2.8"

The comparison mixes time periods and categories

The bars span 2025, 2026E, FY26, and 2028E, so the multiples are directional rather than precise. Investors should treat them as illustrative framing rather than apples-to-apples forecasts.

"2026E and 2028E figures represent estimates and are subject to change; there is no guarantee that such forecasts or projections will be realized."