Measuring Deliverable Demand Before Signing $279 Billion

Demand does not explain the risk. Second-quarter revenue rose 106% from a year earlier (Form 10-Q, p. 28). The risk lies in how late a supplier learns that demand has weakened: by the time a weaker quarter reaches the order book, someone has already signed the commitments that served it.

Supply and capacity commitments, one quarter

$119B to $279B

Commitments to liquid assets

about 2.0x to 4.2x

Two direct customers, FY2026

36%

H20 charge

$4.5B

[OBSERVED] NVIDIA raised its supply and capacity commitments from $119 billion to $279 billion in a single quarter (Form 10-Q, p. 26). [OBSERVED] Two direct customers bought 22% and 14% of fiscal 2026 revenue (Form 10-K, p. 23), 36% between them on recomputation. [OBSERVED] A $4.5 billion charge for H20 excess inventory and purchase obligations already shows what a policy turn costs (Form 10-K, p. 41). [MODELED] Commitments relative to liquid assets moved from about 2.0 times to about 4.2 times between January and July 2026, a ratio that excludes guarantees, which would raise it, and operating cash flow, which would lower it.

A measurement problem, stated as one

NVIDIA commits capital to supply, cloud capacity, and customer guarantees on the strength of order-based demand signals, and the filings describe no measure of how much of that commitment a downside demand case would leave unrecovered. Orders overstate demand whenever a supplier rations output, because buyers inflate orders to protect their allocation (Lee et al., 1997, p. 551). The proposed system measures deliverable demand instead: what customers can deploy, given power, shell, and readiness, set against every commitment before anyone signs it.

Four decision levels, one integration layer

Decision level	Decision supported	Product
Board and CFO	Approve or defer new commitments	Executive scorecard: coverage ratio, downside unrecovered commitment, concentration, each as a range
Supply and operations	Size and phase supplier commitments	Commitments-versus-demand dashboard by supplier and quarter
Sales and field operations	Sequence allocations and shipments	Site-readiness tracker: energization dates, power, shell
Trade compliance and legal	Anticipate license and tariff exposure	Export-exposure monitor mapping rule events to products and revenue

Table 1. Decision levels and the products that serve them. The products are the proposal; nothing here describes NVIDIA systems.

All four products draw on one governed model, so the scorecard and the dashboard reconcile to the same figures. Every forecast carries an interval. A single point forecast, however carefully built, would repeat the exact error the system exists to correct, because it would present a rationed order book as if it measured settled demand.

Success is a changed decision, not a dashboard

Measure	Baseline	Target
Commitments approved with a stated downside amount	None disclosed	[HYPOTHETICAL] 100% above a set threshold
Quarterly actuals inside the 80% forecast interval	Not measured	[HYPOTHETICAL] 70% to 90%
Data Center revenue from customers sharing utilization data	0%	[HYPOTHETICAL] at least 40%
Warning before a site delay affects a shipment	Not measured	[HYPOTHETICAL] 8 weeks or more
Days from rule publication to a revenue-at-risk range	Charge after the fact	[HYPOTHETICAL] 5 business days or fewer

Table 2. Selected success measures. Targets are the author’s pilot targets and carry no measurement behind them.

THE FINDING WORTH CARRYING

Measurement precedes commitment. A supplier that finances demand as well as supplying it needs a downside figure next to every signature.

What changed between the proposal and the case analysis

The proposal from BIN520 (Foundations of Business Intelligence) framed this as allocation under constraint. Three courses later, the frame widened from allocating scarce supply to sizing the commitments behind it, because the 10-Q moved the risk onto the balance sheet. The analysis did not get more confident; it got more specific about where the uncertainty sits.

What these numbers are not

The coverage ratio measures the speed of a change, not the level of risk. The success targets carry no measurement behind them. And the proposal describes a system NVIDIA can build, not one it has; the company does not publish its internal tools.

Provenance and marks

This post is part of the NVIDIA capstone for the M.S. Business Intelligence program at Full Sail University, from BIN660-O, Business Intelligence Case Studies, taught by Dr. Colton Tapoler. Epistemic marks appear before the claim they govern: [OBSERVED] read from a primary source and verified against it, [MODELED] produced by a stated method from stated inputs, [HYPOTHETICAL] a stated assumption or target with no measurement behind it, [UNVERIFIED] asserted somewhere but not resolved to a locator. AI disclosure: Anthropic’s Claude assisted with research, verification, arithmetic in code, and drafting; the author chose the analysis, checked every figure against its source, and answers for the content.

Non-affiliation notice. The author is not employed by, affiliated with, sponsored by or endorsed by NVIDIA Corporation, Microsoft Corporation, Google, Palantir Technologies, or any other company named here. Companies are named as subjects of an academic analysis prepared by an independent party from public disclosure only. No confidential information was used, and no representation is made on any company’s behalf.