Skip to content
Slow Variables

Ask the data

Answers come from the same store as the site; every number is checked against the record it cites.

Methods · Training input

Public text data exhaustion year (Epoch projection)

emerging40/95 confidence, limited or vaguegrade Bleadingrefreshed irregularly: A projection from a 2024 paper; it changes only when Epoch republishes the analysis, so the two-year freshness rule does not apply.

1What this measures

Epoch's central projection of the year in which language-model training fully utilises the effective stock of public human-generated text, from the 2024 paper.

Why it matters. Bottleneck 88 (data exhaustion) as a dated estimate: when the readily available data is spent, continued scaling becomes a business decision about synthetic and licensed data.

Proxy types
benchmark
Unit
year
Cadence
annual
Valve
invention to product

2How we track this

Normal band
≥ 2028 year
Fast band
≤ 2026 year
Falsifying
—

Normal = exhaustion in 2028 or later (the paper's compute-optimal case); fast = 2026 or earlier (the overtrained case, which is what the AI 2027 scenario's compute path implies). Between is emerging.

Applied to epoch_blog.public_text.exhaustion_year.pt.

3Tracker interpretation

A projection, not a reading; the 80% interval runs 2026 to 2032, and the estimate moved four years later between Epoch's 2022 and 2024 papers.

4Evidence

Latest point
2028 yearas of 2024-06-06
public_text
Value the bands apply to
2028 yearas of 2024-06-06

1 observations. Hollow points are disputed (see counterevidence). Every point links to its observation.

5Status and reasoning

emergingsince 2026-09-10 · evaluate

Epoch's 2024 projection puts full utilisation of public human-generated text in 2028 for compute-optimal training (80% interval 2026-2032), inside the normal band (2028 or later); the 2022 paper had said 2024. One research group's estimate, tier 6 and single-source, so the status is capped at emerging; refreshed when Epoch republishes. Initial seed.

6Timeline notes

  • 2024-06-06 public_text · 2028 yearas of 2024-06-06

7Counterevidence

What cuts against this reading

One research group's estimate; synthetic data, licensed corpora and multimodal data all extend the stock; the estimate has already been revised once by four years.

8Update history

  1. 2026-09-10unmeasured to emergingconf — → 40 · evaluate

    Epoch's 2024 projection puts full utilisation of public human-generated text in 2028 for compute-optimal training (80% interval 2026-2032), inside the normal band (2028 or later); the 2022 paper had said 2024. One research group's estimate, tier 6 and single-source, so the status is capped at emerging; refreshed when Epoch republishes. Initial seed.

    obs:c43909ea

9Confidence

40 / 95 — limited or vague evidence

Confidence is independent of status: 90–95 multiple strong independent sources; 70–89 good evidence, some ambiguity; 50–69 mixed or hard to operationalise; below 50 limited or vague.

10Related