The mechanism worth arguing about is how a sector outlook becomes a coupon. Nothing in the published executive summary names a corporate issuer, sets a threshold, or announces an action; the only ratings printed on the page are sovereign, with the US at Aa1 Stable [3], China at A1 Negative [4] and the European Union at Aaa Stable [7]. So the honest description is that Moody's has adopted the equity market's premise, capital spending on compute and infrastructure far outpacing the revenue AI applications generate [2], as the opening frame of a credit document [9]. That is a long way from a rating change and a short way from the assumption sheet an analyst fills in when a data center borrower arrives asking for money against a long-dated lease.
The gap itself is unquantified. What arithmetic there is says something on its own. Across the two executive summaries there is exactly one dollar figure, and it belongs to digital finance rather than to AI: industry plans suggesting more than $300 billion of technology markets spend by 2030 [8], which spread across 2026 through 2030 averages a bit over $60 billion a year [11]. The AI side gets "far outpaces" and no denominator [12]. A ratio with neither term disclosed stays commentary. It becomes underwriting the moment someone writes down contract coverage.
What actually prices, for a landlord, is the commitment. Moody's calls infrastructure the critical bottleneck and points at the surge in data-center construction and long-term capacity commitments, alongside specialist chip shortages and grid and power limits that are reshaping who gets access [5]. The monetization doubt, though, is aimed a level up, at whether the leading providers can convert capability into revenue while open-source models from China close the gap [4]. Uneven productivity capture across and within sectors [6] means no lender gets to treat AI benefit as a sector-wide assumption, and the regulatory divergence Moody's flags, the EU AI Act on one side and China's licensing framework on the other, raises the tenant's cost of running the workload rather than the landlord's cost of housing it [7].
This reads differently depending on where you look. Contracted cash flow from investment-grade offtakers may simply absorb an ARR debate, in which case spreads never notice; or the doubt lands on a tenant's own rating first and reaches the landlord as counterparty risk rather than sector risk; or, and this is probably wrong but it is the more interesting version, the constraint that gets priced in 2026 is interconnection rather than income, because a half-built hall with no power is a credit event whatever the AI revenue line does [5].
Moody's has not put a number on the gap it just made the premise of the year. The full outlook sits behind registration, free to all users only for a limited time [10], which tells you the intended reader is someone building assumptions rather than someone quoting a market. The thesis fails if the 2026 documents that follow keep the bubble language in the outlook and out of the issuer files. Until then, there is one dollar figure in two summaries, and it is about tokenized Treasurys [8].
