1. Chart data
  2. Data movement bottlenecks constrain AI scaling
Epoch AIScatter plot · 41 points · Epoch AI

Data movement bottlenecks constrain AI scaling

The numbers behind a chart Epoch AI published, redrawn natively. Check the original, study the data, then make your own.

Estimated from the image 41%All chart data
Chart by Epoch AI. Source: x.com

Simplified redraw

Drawn by us from the numbers below. Hover, tap or use the arrow keys.

A simplified redraw: it shows the numbers in our own look, not the look of the original. One accent marks what the chart is about. Values marked with ~ in the table are estimates.

The numbers

First 10 of 41 points are free
GroupSeriesxValue (FLOP)SourceConf.
2010Top-1 ML models by compute2010~100000000000000000Estimate0.80
2012Top-1 ML models by compute2012~100000000000000000Estimate0.30
2013Top-1 ML models by compute2013~1000000000000000000Estimate0.80
2014Top-1 ML models by compute2014~1000000000000000000Estimate0.30
2014Top-1 ML models by compute2014~10000000000000000000Estimate0.30
2014Top-1 ML models by compute2014~100000000000000000000Estimate0.30
2015Top-1 ML models by compute2015~100000000000000000000Estimate0.80
2015Top-1 ML models by compute2015~1e+21Estimate0.30
2015Top-1 ML models by compute2015~1e+21Estimate0.30
2016Top-1 ML models by compute2016~1e+21Estimate0.80

Max data, free with your email for now

See all 41 points and download the files

The full table, the JSON, a CSV and the agent skill file for this chart. These are Max in the plan; while this is a prototype your email unlocks them.

Unlocks all 35 giants. One email a week, leave any time. What you get · Log in

2016Top-1 ML models by compute2016~1e+22Estimate0.30
2016Top-1 ML models by compute2016~1e+23Estimate0.30
2016Top-1 ML models by compute2016~1e+23Estimate0.30
2017Top-1 ML models by compute2017~1e+23Estimate0.30
2017Top-1 ML models by compute2017~1e+24Estimate0.30
2018Top-1 ML models by compute2018~5.5e+23Estimate0.80
2018Top-1 ML models by compute2018~1e+24Estimate0.30
2019Top-1 ML models by compute2019~1e+24Estimate0.30
2020Top-1 ML models by compute2020~1e+24Estimate0.80
2020Top-1 ML models by compute2020~1e+24Estimate0.30
2021Top-1 ML models by compute2021~1e+24Estimate0.30
2021Top-1 ML models by compute2021~1e+24Estimate0.30
2022Top-1 ML models by compute2022~1e+24Estimate0.30
2022Top-1 ML models by compute2022~1e+25Estimate0.30
2022Top-1 ML models by compute2022~1e+25Estimate0.30
2023Top-1 ML models by compute2023~1e+25Estimate0.30
2023Top-1 ML models by compute2023~1e+25Estimate0.30
2024Top-1 ML models by compute2024~1e+25Estimate0.30
2024Top-1 ML models by compute2024~1e+26Estimate0.30
2016Largest feasible training run w/ GPU at high utilization2016~1e+27Estimate0.30
2018Largest feasible training run w/ GPU at high utilization2018~1e+27Estimate0.30
2020Largest feasible training run w/ GPU at high utilization2020~1e+27Estimate0.30
2022Largest feasible training run w/ GPU at high utilization2022~1e+27Estimate0.30
2024Largest feasible training run w/ GPU at high utilization2024~1e+27Estimate0.30
VGG16Top-1 ML models by compute2014~10000000000000000000Estimate0.50
AlphaGo ZeroTop-1 ML models by compute2016~1e+21Estimate0.50
GPT-4Top-1 ML models by compute2023~1e+25Estimate0.50
V100Largest training run w/ GPU at high utilization2017~1e+25Estimate0.50
A100Largest training run w/ GPU at high utilization2020~1e+26Estimate0.50
H100Largest training run w/ GPU at high utilization2022~1e+27Estimate0.50
Latency wallLatency wall2010~1e+33Estimate0.50
Download this chart's data

Max subscribers get the same data by API. API docs

JSONAgent skill file

API · Max

Use this chart via API

On the Max plan you can pull this chart's numbers with the provenance of each one, as JSON or CSV, with its redraw spec, or call it from an agent over MCP. Chart id 682c22933ca3.

Approximate values. Our two readings of this chart agreed only in part (confidence 0.41 of 1), so check the original before you quote a number. The API marks it lowConfidence, and ?min_confidence=0.5 leaves it out.

See the Max planAPI docs
PathReturnsPlan
GET/api/v1/charts/682c22933ca3Title, publisher, date, form, point count; with a Max key, the data linkOpen
GET/api/v1/charts/682c22933ca3/dataEvery point with its provenance and confidence, plus the redraw spec. Add ?format=csv for CSVMax
GET/api/v1/charts/682c22933ca3/render?view=1The draw-only view this page usesOpen
POST/mcpTool get_chart_data with this id, for Claude, Cursor and other agentsMax
curlMax
curl -s https://viral-chart.com/api/v1/charts/682c22933ca3/data \
  -H "Authorization: Bearer $VIRALCHART_KEY"
Claude CodeMax
claude mcp add --transport http viralchart https://viral-chart.com/mcp \
  --header "Authorization: Bearer $VIRALCHART_KEY"

60 requests a minute per key; over that you get 429 with Retry-After. Up to 5 keys per account, made on your account page and shown once. Without Max, the data route answers 402 with a link to the plans. The numbers describe a chart its publisher owns, and every file carries a license note.

Make one like this

This chart's shape, your numbers.

Live preview. Everything stays on your device.

Pick the form

A simplified form: our own look, not the look of the original. Your numbers start from this chart's data.

More data from Epoch AI

All chart data
Estimated from the imageIn August, 25% of math preprints acknowledged AI use, up from 4% in AprilEpoch AI · Line chart · 44 pointsEstimated from the imageWith domestic HBM only, Huawei's AI compute production could stay around 1% of Nvidia's through 2028Epoch AI · Grouped bar chart · 6 pointsEstimated from the imageIncreasing lead times may lead to a substantial deceleration in compute investment scalingEpoch AI · Scatter plot · 12 pointsEstimated from the imageBenchmark ReviewsEpoch AI · Table chart · 15 pointsEstimated from the imageGPT-5 is an incremental update to the frontier, but a major leap from GPT-4Epoch AI · Line chart · 40 pointsEstimated from the imageClaude Opus 5.5 edges ahead on ECI, with Sonnet 5.5 in third placeEpoch AI · Scatter plot · 61 points

Questions people ask

How accurate are these numbers?

They are estimated from the image by two independent passes, with confidence 0.41 on a 0 to 1 scale. Printed numbers read well; values read off an axis can be a few percent off.

Can I reuse this data?

The numbers describe a chart that Epoch AI published. The image, layout and styling belong to them. Use the numbers with a credit to the publisher, and make your picture from your own data.

How do I get the file?

The JSON, CSV and agent skill file are under the email gate on this page. In this prototype they are static files at /api/charts/682c22933ca3.json and /api/charts/682c22933ca3.skill.md.