The Peer Benchmark That Broke My Own Argument

A problem statement that names a company without naming a comparison is an assertion about that company alone. This post records the test that was run against an earlier draft of my own argument, and the fact that the argument failed it.

What the earlier draft said

An earlier draft of the solution plan held that client inventory was elevated and treated that as evidence for the proposed system. The claim was never locatored against a comparison, only against the client’s own prior periods, which is a growth observation wearing the clothes of a level observation.

The test was straightforward. Take the seven listed semiconductor firms nearest the client in scale or in product adjacency, compute inventories divided by annual revenue from the balance sheet and the income statement of each most recent annual filing, and see where the client lands.

Second lowest in the set

Company	Filing	Inventories / revenue	Ratio
Broadcom	FY2025 10-K, FYE 2 Nov 2025	$2,270M / $63,887M	3.55%
TSMC	FY2025 20-F, FYE 31 Dec 2025	NT$288,109.5M / NT$3,809,054.3M	7.56%
NVIDIA (client)	FY2026 10-K, FYE 25 Jan 2026	$21,403M / $215,938M	9.91%
QUALCOMM	FY2025 10-K, FYE 28 Sep 2025	$6,526M / $44,284M	14.74%
Marvell	FY2026 10-K, FYE 31 Jan 2026	$1,388.0M / $8,194.6M	16.94%
Intel	FY2025 10-K, FYE 27 Dec 2025	$11,618M / $52,853M	21.98%
AMD	FY2025 10-K, FYE 27 Dec 2025	$7,920M / $34,639M	22.86%
Texas Instruments	FY2025 10-K, FYE 31 Dec 2025	$4,804M / $17,682M	27.17%

Table 1. Inventory to revenue across the peer set. Observed. Every ratio recomputed from the printed figures. Locators: Broadcom revenue p. 48 and inventory p. 47; TSMC revenue p. F-7 and inventories p. F-5 agreeing to Note 13 p. F-35; NVIDIA revenue Note 16 p. 78 and inventories p. 53; QUALCOMM revenue p. F-4 and inventories p. F-3; Marvell revenue p. 65 and inventories p. 64; Intel revenue p. 60 and inventories p. 62 agreeing to Note 6 p. 79; AMD revenue p. 58 and inventories p. 60 agreeing to Note 3 p. 69; Texas Instruments revenue p. 27 and inventories p. 29.

THE RETRACTION

Client inventory is not elevated against this set. It is lower than five of the seven comparators and roughly a third of the highest. An argument built on inventory level would fail on its own evidence, and no such argument is made anywhere in the finished plan.

A high provision rate on a low inventory level

[OBSERVED] Client inventory provisions of $4.0 billion are 1.85 percent of fiscal 2026 revenue on recomputation (Form 10-K, Note 9, footnote 1, p. 67). AMD recorded approximately $440 million of net inventory and related charges on fiscal 2025 revenue of $34.639 billion, which is 1.27 percent on recomputation (Form 10-K, Note 3, p. 70; Item 7, p. 53).

Intel discloses no absolute charge, only year over year increases of $878 million in consolidated inventory reserves (Form 10-K, Item 7, p. 26), so no rate can be computed for Intel without inventing the base, and none is computed anywhere in this work.

The client writes down more per dollar of revenue while holding less per dollar of revenue. That combination is not a stocking problem. It is a commitment problem.

A line no comparator carries

[OBSERVED] Accrued liabilities at 25 January 2026 include $2.739 billion for excess inventory purchase obligations, against $2.095 billion a year earlier (Form 10-K, Note 9, p. 68). The accrued liabilities notes or balance sheet faces of all seven comparators were read for an equivalent line.

AMD itemises accrued marketing, compensation, customer-related and other accrued expenses (Note 3, p. 69). Marvell itemises variable consideration, income taxes, technology licences, lease liabilities, restructuring, interest, deferred revenue and royalties (Accrued liabilities note, p. 101). Intel, Texas Instruments, QUALCOMM, TSMC and Broadcom present a single other-accrued caption with no component note at all. None of the seven discloses an accrual for excess or non-cancellable purchase obligations.

Each of them discloses purchase commitments off balance sheet, in the commitments note, which is where a commitment sits when the company does not expect to pay for supply it cannot place.

WHAT THE NEGATIVE MEANS

A confident negative across a full peer set is evidence, not an absence of it. The client has recognised, as a present liability, an expected loss on supply already committed. Its comparators have not. The distinguishing exposure is commitment, not inventory, and commitment is precisely what a constrained allocation system governs.

What the comparison does not show

[INFERENCE] The client operates at a scale and a growth rate none of the comparators match, and a ratio computed across firms of unlike size carries that difference silently. The peer set was selected by me for scale or product adjacency and is not a defined index.

The ratios also compare a single balance sheet date against a full year of revenue. That is the conventional construction and it understates inventory for any company growing quickly within the year, which includes the client. The direction of the bias is stated here rather than corrected, because correcting it would require quarterly averages that three of the seven do not publish in comparable form.

The comparison is offered as a check on a claim this project had made, not as a demonstration that the client is mismanaged. It retired one argument and produced two better ones, which is the whole of what it was asked to do.

Provenance and marks

This post is part of the NVIDIA capstone for the M.S. Business Intelligence program at Full Sail University. It originates in BIN520-O, the second course of the program, and therefore predates the other capstone posts published here. 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, [INFERENCE] reconstructed from indirect evidence, [CONVENTION] a customary figure with no published benchmark behind it, [UNVERIFIED] asserted somewhere but not resolved to a locator.

Non-affiliation notice. The author is not employed by, affiliated with, sponsored by or endorsed by NVIDIA Corporation. NVIDIA is named here as the subject of an academic analysis prepared by an independent party, from public disclosure only. No confidential information of NVIDIA Corporation was used, and no representation is made on its behalf.