The Economics of the Build-Out

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 they are routinely conflated in the 2026 discourse.

At sector altitude the questions are the capex wave and the build cost per MW that sets its denominator; 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; the debt wave and the new financing structures funding the gap, including power oversubscription as a yield strategy; and the revenue-vs-capex gap and what protects returns when the cycle turns. 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 magnitudes are without precedent in industrial history. Top-4 US hyperscaler capex is guided to ~$710–740B in 2026 (the four companies' own guidance as of 22 Jul 2026: Amazon ~$200B, Microsoft ~$190B, Meta $125–145B, Alphabet $195–205B; CreditSights' Oracle-inclusive top-5 estimate is ~$750B), up more than 60% 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 McKinsey's $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 ~$15-25M/MW all-in at the facility level — shell, fit-out, power chain, and cooling, but excluding the IT/servers themselves (JLL's 2026 Outlook puts AI-optimized builds at $15-25M/MW; Goldman Sachs models ~$15M/MW; market benchmarks put the global average near ~$11.3M/MW across all build types), 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 also inverts the legacy intuition. In a conventional data center the building and power plant dominate; in an AI factory the silicon dominates everything — servers run ~56% of capex (~$21.2/W, with the accelerators inside that line), the facility — MEP, shell, and construction labor — ~30% (~$11.4/W), and network and cluster infrastructure ~13% (~$4.9/W), for an all-in up-front intensity — silicon included — near $37.9/W (Epoch AI, May 2026). Mind the scopes before subtracting: Epoch's ~$11.4/W facility line is leaner than the ~$15-25M/MW AI-optimized facility band above, and its ~$37.9/W all-in sits visibly above the ~$30/W (~$30B/GW) canon of Chapter 6.3 — a vintage-and-scope gap flagged on that claim rather than silently averaged. 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.

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. 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 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). You do not have to pick a side to see 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.

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 securitization ran ~$26B in 2025 — ~$15B of ABS plus ~$11B of CMBS — with gross supply projected toward $30-40B/yr in 2026-2027 (CREFC, 2026), and total data-center ABS forecast to grow from ~$8B (2025) toward ~$25B by 2028 — while CoreWeave's first ~$8.5B investment-grade-rated (A3) GPU-collateralized deal was a delayed-draw term loan (DDTL 4.0), not an ABS issuance. 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: 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; at this altitude the point is that 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.

This is the question that decides the outcome: does the revenue arrive before the depreciation does? The most-cited framing is Sequoia's — David Cahn's '$600B question' — now nearer ~$840B/yr on 2026 capex: 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 $25B run-rate (mid-2026; an internal ~$30B full-year 2026 target) is on the order of ~3.5% of 2026 capex — even as hyperscaler cloud-AI revenue grows fast off a real base (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 rather than a flat extrapolation, and do not assume today's token price holds.

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.