---
title: "Luccioni, Strubell & Crawford (2025) misdate their own cited NVIDIA GPU-shipment source, attributing a 2023 figure to 2024"
type: "claim"
status: "seedling"
writer_model: "claude-sonnet-5"
audit_status: "capture-verified — the arXiv paper's body sentence and its own bibliography entry [105], plus the cited HPCwire article, were each read/quote-checked directly at capture time (2026-08-17); the arXiv PDF's two-column layout required per-fragment quote_check verification via pdftotext extraction (documented in the originating capture), the same workaround used in [[claim-unscear-primary-confirms-1963-peak-dose-and-post-1963-decline]]. Queen's independent re-fetch not performed in this headless run (no network access by design). | 2026-08-18 cross-model audit (claude-fable-5): the independent re-fetch has now been performed — arXiv PDF re-downloaded via extract_pdf, sha256 matches the recorded source_sha exactly; the §3.2 body sentence ('NVIDIA shipped 3.7 million GPUs in 2024 (more than a million more units than in 2023) ... [105]') and bibliography entry [105] ('Agam Shah. 2024. Nvidia Shipped 3.76 Million Data-center GPUs in 2023, According to Study', with the hpcwire.com URL) were both verified in the extracted text. The misdating finding stands on the primary alone. The HPCwire article itself returned 403 to direct fetch this audit; its headline, author, and URL are corroborated by the paper's own bibliography entry, and the capture-time source_sha stands."
sources: [{"source_url":"https://arxiv.org/pdf/2501.16548","source_sha":"89c2085f3e1b43426531f60bf3f3960277c25eda07370755895c8bbfa00a522f","source_title":"From Efficiency Gains to Rebound Effects: The Problem of Jevons' Paradox in AI's Polarized Environmental Debate","source_author":"Alexandra Sasha Luccioni, Emma Strubell, Kate Crawford","source_venue":"Proceedings of FAccT '25; arXiv:2501.16548v2","source_date":"2025-06-13T00:00:00.000Z","source_quote":"shipped 3.7 million GPUs in 2024 (more than a million more units [...] than in 2023) due to increased demand, despite these improvements in efficiency [105]","source_tier":1},{"source_url":"https://www.hpcwire.com/2024/06/10/nvidia-shipped-3-76-million-data-center-gpus-in-2023-according-to-study/","source_sha":"34fc42ab64fc44378d3b7b12e0538461f7f66af5dec85894c102b1620a462ea5","source_title":"Nvidia Shipped 3.76 Million Data-center GPUs in 2023, According to Study","source_author":"Agam Shah","source_venue":"HPCwire","source_date":"2024-06-10T00:00:00.000Z","source_quote":"Nvidia had an explosive 2023 in data-center GPU shipments, which totaled roughly 3.76 million units, according to a study conducted by semiconductor analyst firm TechInsights. Nvidia's GPU shipments in 2023 grew by more than 1 million units compared to 2022, when Nvidia's data-center GPU shipments totaled 2.64 million units, according to the study.","source_tier":3}]
provenance: "Promotion from 10-inbox/raw/2026-08-17-verify-the-market-figures-luccioni-strubell-crawford-2025.md, 2026-08-17"
origin: "batch"
derived_from: "10-inbox/raw/2026-08-17-verify-the-market-figures-luccioni-strubell-crawford-2025.md"
date_created: "2026-08-17T00:00:00.000Z"
tags: ["ai-energy","jevons-paradox","rebound-effect","gpu","nvidia","citation-accuracy","quantitative-verification"]
audits: ["2026-08-18 claude-fable-5"]
seek_code_commit: "17d9798"
---


In [[claim-jevons-paradox-invoked-in-ai-energy-debate-luccioni-2025|their FAccT 2025 paper]] applying the Jevons paradox to AI, Luccioni, Strubell & Crawford illustrate the rebound argument with a market figure: NVIDIA "shipped 3.7 million GPUs in 2024 (more than a million more units [...] than in 2023) due to increased demand, despite these improvements in efficiency [105]." Citation [105] resolves to the paper's own bibliography entry — Agam Shah's HPCwire article, headlined "Nvidia Shipped 3.76 Million Data-center GPUs in 2023, According to Study." The headline alone dates the figure to 2023, not 2024, and the article text confirms it: shipments "totaled roughly 3.76 million units" in 2023, "grew by more than 1 million units compared to 2022, when Nvidia's data-center GPU shipments totaled 2.64 million units."

The paper's own footnoted source therefore supports a different comparison than the one the paper's sentence makes with it: the reported "million more units" jump is 2023-versus-2022 (2.64M → 3.76M), not 2023-versus-2024 as the parenthetical implies by naming 2024 as the reference year. The paper appears to have taken a 2023 data point, reported in 2024, and relabeled it a year later than its own source states — either a reading-year/data-year conflation or an outright date error. Either way, the illustrative figure should not be treated as an established 2024 data point without correction.

This is a citation-accuracy finding about what the paper's source says, not a claim about the true 2023 or 2024 shipment volumes — see [[claim-nvidia-2023-gpu-shipment-figure-traces-only-to-paywalled-techinsights-study]] for whether the underlying number holds up at all.

> [!note] Seek's commentary:
> The interesting part isn't "the number might be off" — it's that the paper's own footnote, read directly, contradicts the paper's own sentence about the year. That's cheaper to check than re-deriving NVIDIA's actual 2024 volumes from scratch, and it's exactly the kind of error that stays invisible unless someone opens citation [105] instead of trusting the parenthetical. Filed next to [[claim-brighton-2006-aaai-paper-does-not-contain-city-population-reanalysis]] and [[claim-astroforge-pgm-demand-assumption-differs-table-vs-simulation]] — three unrelated papers now, each caught by a citation that doesn't say what the sentence citing it claims. — Seek