Chapter 16.4
The Economics of the Build-Out
The build-out is a multi-trillion-dollar bet that levers a long-dated, illiquid capital structure against the shortest-lived collateral ever financed at this scale — and whether it is a supercycle or a balance-sheet reckoning turns on three macro numbers: the cost to energize a watt, the debt the sector can carry against depreciating silicon, and the revenue that has to show up before the depreciation does.
What you'll decide here
- Whether you underwrite the sector at the ~$15-20M/MW AI-optimized build cost that 2026 actually delivers, or the legacy ~$7-11M/MW number that still anchors stale models — because the gap is the difference between a financeable plan and a funding hole.
- How much of the build you can finance with contracted, take-or-pay-backed debt against a depreciating GPU asset, versus merchant exposure that a single demand miss strands — the split that sets the sector's true debt capacity.
- Whether to treat power oversubscription (provisioning more IT against a fixed interconnection than peak draw would allow) as a yield strategy or a tail risk — it is both, and the workload mix decides which.
- Which macro depreciation life you believe the aggregate fleet obeys, because at sector scale the 2-3 vs 5-6 year gap is the single largest swing in whether reported industry earnings are real or borrowed from a future write-down.
- What actually protects returns when the cycle turns — contracted revenue, a low-cost firm power position, conservative leverage, and design-for-flexibility — versus the levers (utilization optimism, residual optimism, circular financing) that amplify the downside instead.
This is the sector-macro altitude. Chapter 1.8 scored a single asset — does this specific facility, financed this specific way, clear its cost of capital over its economic life. This chapter asks the larger and more contested question: is the industry as a whole over- or under-building, and does the aggregate capital stack survive the depreciation clock it is racing? A single asset can clear its hurdle in a sector that is collectively over-built; a single asset can fail in a booming one. The two altitudes are independent, and conflating them is the most common analytical error in the 2026 discourse.
This chapter runs the decision-and-consequence frame at sector altitude rather than firm altitude. We trace the capex wave and the build cost per MW that sets its denominator; we confront the macro depreciation problem — the GPU-ROI-decay argument from Chapter 1.8 aggregated across the whole fleet, where it becomes a question about the quality of reported industry earnings; we map the debt wave and the new financing structures funding the gap, including power oversubscription as a yield strategy; and we close on the revenue-vs-capex gap and what actually protects returns when the cycle turns. The through-line: almost every number here is contested, and the contested ones are exactly the ones that decide whether this is a supercycle, a digestion, or a bubble. We flag them as we go and bind them to the dated forecast register in Appendix D.
The capex wave and the cost to energize a watt
The magnitudes are without precedent in industrial history. Top-4 US hyperscaler capex runs ~$600B in 2026 (estimates span $600-725B by analyst, on differing scope definitions), up roughly a third year-on-year, against a global data-center capex approaching ~$1T in 2026 at a ~21% CAGR through 2029 (Dell'Oro / CreditSights, 2026). The long-horizon framings are larger still: ~$6.7T of global data-center capex by 2030 (of which ~$5.2T is AI-capable), the midline of a $3.7T-constrained-to-$7.9T-accelerated scenario band (McKinsey, 2025). These are forecasts, not committed builds — the error bars are wide and scenario-dependent — but the central case is a sustained, multi-year capital deployment larger than the build-out of the Interstate Highway System, the Apollo program, and the early commercial internet combined.
The denominator that turns those headline trillions into a financeable plan is the cost to build and energize a watt, and it has roughly doubled in five years. The legacy data-center number — ~$7-11M/MW for a conventional shell — no longer describes an AI facility. An AI-optimized build runs ~$15-20M/MW all-in (Goldman Sachs models ~$15M/MW; market benchmarks put the global average near ~$11.3M/MW across all build types, with AI-optimized facilities at $15-20M+), driven by four cost factors a legacy model omits: electrical density (the power chain to feed 130-600 kW racks), liquid cooling (DLC plant and facility water that air-cooled halls never needed), redundant power trains, and the scripted functional testing an AI cluster demands before handoff (Goldman Sachs; iRecruit / Construct Elements market data, 2026). Underwrite the sector at the stale number and you under-state the build by 40-100% — which is precisely how a financing gap appears where a naive model saw none.
The cost stack inverts the legacy intuition, and that inversion is the macro story. In a conventional data center the building and power plant dominate; in an AI factory the silicon dominates everything — IT/servers run ~60-64% of capex (~$17.50/W), power+cooling+electrical ~29-30% (~$7-10/W), and the building shell only ~7% (~$1.90/W), for an all-in capital intensity near $27.50/W (Epoch AI; domain synthesis, 2026). GPUs alone are roughly a third of all data-center capex (Dell'Oro, 2026). The consequence at sector scale is the same as at firm scale, only larger: the asset's economic life is the GPU's economic life, not the concrete's, and the entire industry's reported profitability hinges on how fast that silicon decays.
The depreciation problem, at sector altitude
Chapter 1.8 is the canonical home for the firm-level depreciation debate; here we take it up a level, because aggregated across the fleet the same argument stops being an accounting choice and becomes a question about the quality of reported industry earnings. The mechanism is plain: every year an operator extends the assumed useful life of its AI fleet, it moves cost off the current income statement, so reported operating margin rises even though nothing about the physical asset improved. When the largest operators do this in opposite directions in the same window — Meta extended server life from 4.0 to 5.5 years (+$2.9B income); Amazon went the other way, 6 to 5 years (−$700M) (company filings, 2025) — no outside party should pretend the number is settled.
The bull case is the training-to-inference cascade: a GPU retired from frontier pre-training is not scrap, it cascades down to post-training, then inference serving, then batch and internal workloads, earning revenue at each step. If the cascade holds, economic life stretches toward book life and the accounting is honest. The bear case is that the cascade is finite, that each new generation is so much more efficient per token that the old part becomes uneconomic to run against grid power, and that the residual market is thin. The residual evidence is genuinely mixed: H100s retained ~60-83% of value at 18 months, but rental rates fell 64-75% from their $8-10/hr peak, and the implied residual after three years is ~20-40% (Hashrate Index / CNBC synthesis, 2025). A high residual underwrites the cascade defense; a low one validates the short-life bears. Both can be partly true at once.
At sector scale the stakes are systemic rather than per-firm. The sharpest bear version — that the industry is systematically under-depreciating its fleet, booking long lives to flatter earnings while the assets decay on the short schedule — puts ~$176B of understated depreciation across 2026-2028 against an industry AI-asset D&A line approaching ~$400B/yr (Burry / secondary analyses, 2025-2026). The honest engineering-economics posture is not to pick a side but to recognize that depreciation policy is now the central earnings-quality battleground for the entire sector — and that the whole long-life defense rests on a deep, liquid secondary GPU market that can absorb cascaded hardware at a stable residual. If that market is thin, the cascade is a story rather than a cash flow.
| Outcome | Depreciation life that holds | Revenue vs capex | Utilization / demand | What it means for the debt wave |
|---|---|---|---|---|
| Supercycle | 5-6 yr — cascade holds, residuals stable | Revenue catches the gap; Jevons demand absorbs supply | Fleet runs above breakeven; reasoning/agentic demand multiplies | Contracted debt is well-collateralized; ABS performs |
| Digestion | Mixed — life shortens for frontier parts, holds for cascaded | Gap persists but narrows; over-build absorbed over 2-3 yr | Merchant fleets soften; contracted fleets stay full | Refinancing stress on merchant builds; contracted debt fine |
| Bubble / reckoning | 2-3 yr — cascade thin, residuals collapse | Revenue never arrives at scale; $600B+/yr gap widens | Utilization collapses below the ~70% breakeven cliff | GPU-backed debt under-collateralized; correlated default risk |
Financing and the debt wave
The defining structural feature of the 2026 build-out is that it has outgrown self-funding. Against a multi-year build estimated near $2.9T (2025-2028) with a ~$1.5T financing gap beyond hyperscaler operating cash flow (Morgan Stanley, 2025), the sector reached for external capital — and the form that capital took is the macro story. Epoch AI's parse of the five largest cloud companies' filings found capex growing ~70%/yr against cash flow growing ~23%/yr, with aggregate free cash flow crossing zero around Q3 2026 (Epoch AI, 2026). When the richest balance sheets in the world stop self-funding, the marginal megawatt is financed by the credit market — and the credit market is now exposed to AI-infrastructure risk in a way it was not in 2023.
The instruments are GPU-collateralized debt, delayed-draw term loans (DDTLs), bankruptcy-remote SPVs, and asset-backed securitization. Data-center ABS issuance ran ~$25-27B in 2025 and is projected toward $30-40B/yr in 2026-2027 (JPMorgan / Morgan Stanley, 2026), with GPU-backed ABS specifically forecast to grow from ~$8B (2025) toward ~$25B by 2028 — CoreWeave landed the first ~$8.5B investment-grade-rated GPU-collateralized deal. Morgan Stanley estimates ~$130B of US data-center securitized-credit net issuance across 2026-28, ~75% of it ABS. The strategic catch is structural and it rhymes with the depreciation debate: this debt underwrites a depreciating, deflating asset against a thin secondary market. The same residual-value uncertainty that makes the accounting contested is now wired into the capital structure, and securitization distributes that risk into pensions and asset managers — a new systemic linkage between AI capex and broad credit markets.
The visible test case makes the tension concrete. CoreWeave reported FY25 revenue $5.13B and a 60% adjusted-EBITDA margin, but a −$1.17B net loss, ~$21-25B of debt, interest near 46% of EBITDA, and a ~$66.8B backlog (~13x revenue) concentrated in a few anchor tenants (company filings, 2026). The 'circular financing' critique — vendor equity stakes and residual backstops that let a buyer finance the purchase of the vendor's own chips — is a real structural risk, not a talking point: it couples the financing to the same demand and residual assumptions the equipment depends on, so a residual shock hits collateral, covenants, and revenue at once. The deal mechanics are engineered in Chapter 2.5; the point here is macro: the sector's true debt capacity is set by how much of the revenue is contracted and how durable the residual is, not by the headline collateral value.
The revenue-vs-capex gap
This is the question that decides the whole thing: does the revenue arrive before the depreciation does? The most-cited framing is Sequoia's — David Cahn's '$600B question': the annual end-user revenue that has to materialize to justify the capex deployed, a gap that has widened, not closed, as capex accelerated through 2026 (Sequoia, 2024-2026). The pressure is visible in the ratio of capex to demonstrated end-demand: pure-play AI ARR remains small against the capital deployed — OpenAI at roughly $20B run-rate is on the order of ~3% of 2026 capex — even as hyperscaler cloud-AI revenue grows fast off a real base (domain synthesis / company filings, 2026). The bears read this as a bubble; the bulls read the small-but-fast-growing revenue as the early innings of Jevons-paradox demand. Both are looking at the same numbers.
The counterweight to the gap is genuine and it is the bull case's strongest card: the cost to serve a model of fixed quality has fallen ~10x per year (a16z's 'LLMflation'; Epoch's median measurement nearer 50x/yr), and that deflation, via Jevons, multiplies token volume — aggregate AI spend rose ~320% over two years even as token prices fell ~280x (a16z; Epoch AI, 2024-2026). Reasoning and agentic workloads are a structural demand multiplier on top, reshaping the decode-heavy future that Chapter 16.3 treats in full. If demand keeps absorbing the supply, the gap closes from the revenue side and the build-out is vindicated. The gap is not evidence of a bubble by itself — it is the open question whose resolution defines which of the three scenarios the sector lands in. Underwrite it with an explicit demand-growth curve, never a flat extrapolation, and never assume today's token price holds.
Deep dive: why the revenue gap, the depreciation gap, and the financing gap are the same gap wearing three hats
It is tempting to treat the $600B revenue gap, the ~$176B understated-depreciation question, and the ~$1.5T financing gap as three separate worries. They are not — they are three measurements of a single underlying uncertainty: will AI demand resolve high enough, fast enough, to fill the asset before the silicon decays? If demand arrives, revenue closes the revenue gap, utilization stays above the ~70% breakeven cliff, the cascade keeps residuals high enough to validate the longer depreciation life, and the contracted cash flows service the debt — all three gaps close together. If demand disappoints, all three open together: revenue misses, utilization falls below breakeven, residuals collapse so the short-life bears are vindicated, and the GPU-backed debt is suddenly under-collateralized against falling revenue. This is the correlation that matters most at the macro altitude.
The systemic danger is precisely that the tails are correlated, and that securitization has distributed the resulting credit risk into the broader financial system. A demand miss does not produce a gentle, sector-contained digestion; it produces a simultaneous revenue, residual, and refinancing shock, transmitted through ABS held by pensions and asset managers, against offtaker concentration (a large share of neocloud backlogs sits with a handful of names). The bull case is symmetric and equally correlated: if Jevons demand shows up, the same linkages amplify the upside. The engineering-economics posture is to stop modeling the three gaps independently and price them as one correlated event — because that is how they will actually arrive. → structural resolution scenarios in Chapter 16.5; the firm-level stress tests in Chapter 1.8.
What actually protects returns
At the macro altitude, what protects returns is a small set of positions that pay off across all three correlated tails at once, not a clever capital structure. Four matter, and they are the inverse of the levers that amplify the downside.
- Contracted revenue over merchant exposure. The sector's true debt capacity is set by the take-or-pay-backed share of revenue, not the headline collateral value. A contracted, diversified offtaker base survives a demand miss; a merchant fleet into a soft GPU-rental market is the canonical bubble casualty. The hedge is real termination economics and tenant diversification; the failure mode is a backlog concentrated ~13x revenue in a few anchor tenants.
- A low-cost, firm power position. Energy is the largest controllable lifetime opex line, and on a power-bound build a cheap, firm, long-dated supply is worth more to ROI than most capex optimization. It is also the asset that lets you run power oversubscription as a yield strategy rather than a liability. → siting and speed-to-power economics in Chapter 3.2.
- Conservative leverage and depreciation. The operator that books the short life up front and limits leverage has already priced the bear case into its balance sheet; the one running maximum leverage against an optimistic residual is the one a residual shock detonates. Circular financing is the sharpest version of the amplifying lever — avoid it as a structural matter, not a reputational one.
- Design-for-flexibility. The cheapest hedge that pays off across every correlated tail: a hybrid procurement posture (colo anchor plus neocloud overflow) so a demand miss sheds opex instead of stranding capex, and the reserved floor loading, water, and electrical headroom that let the asset absorb a density step instead of being stranded a generation behind. You cannot manufacture a take-or-pay contract or a deep residual market after the cycle turns; you can buy flexibility cheaply before it does. → the irreversibility framing in Chapter 1.8.