Today’s university students are the first cohort in history to complete an entire degree alongside generative artificial intelligence. A student who enrolled in 2023 sat their first exams the same week ChatGPT reached a hundred million users, and will graduate into a workforce being re-drawn in real time. For this generation, AI is not a distant future to prepare for - it is a classmate, a tutor, a threat and a tool, all at once.
These five charts re-cut real, independently sourced data into a single story: the money pouring in, who controls it, the capabilities being conquered, what is being built, and how far the technology has already spread into everyday life. Hover any element, and click a legend entry to isolate a series.
Before AI reached our essays and exam halls, it reached investors. The line below traces global private investment in AI companies (venture and private-equity money, not Big Tech’s internal budgets) year by year. For most of the 2010s it crept along; then ChatGPT’s launch in late 2022 lit a fuse, and investment vaulted to a record US$290 billion in 2025. The post-ChatGPT stretch is drawn in red.
Data: Stanford AI Index Report (2026) via Our World in Data; Description: Inflation-adjusted to constant 2021 US$.
▸ Interact: hover any point for that year’s figure, or click a legend entry to isolate a segment.
The headline number hides a power story. Instead of dollars, this chart shows each region’s share of the global total so the band widths always add to 100%. One band swallows the rest. The United States’ share climbed from 70% in 2013 to 83% in 2025, while China’s slid from nearly a third at its 2018 peak to under 4%. The tools a student in Melbourne writes, codes and studies with are overwhelmingly shaped by a single country, which makes whose values and assumptions get built in a part of the human story too.
Description: Regional shares computed from the absolute figures. Data: Stanford AI Index Report (2026) via Our World in Data.
▸ Interact: click a region in the legend to add or remove its band.
Money buys capability, and capability has a finishing line, that’s us. For each skill, the grey dot marks the year its benchmark was introduced and the coloured dot marks the year AI first beat the average human. Read each row as a race. Speech, handwriting and image recognition were won years ago (by 2015–2018). But the rows that matter most to a student, maths, code and complex reasoning, are still unfinished (red, open dots); as of 2023 the machine had not yet passed us. The question is how many of those open dots will close before you graduate.
Description: Parity years derived from normalised benchmark scores. Grey = benchmark introduced; green = year AI beat humans; open red = not yet (as of 2023). Data: Kiela et al. (2023), via Our World in Data.
▸ Interact: hover either end of a row for the exact years.
Where did all that capability get pointed? Rather than a running total, this chart shows the number of large-scale AI models built in each individual year, stacked by domain, so the height of each bar is the field’s annual output. The surge is recent and lopsided: 124 new large-scale language models were built in 2025 alone more than double the next-biggest domain. Language is the raw material of study (reading, writing, explaining, arguing), so the boom landed squarely on the skills universities exist to teach.
Description: Annual additions derived from cumulative counts. Data: Epoch AI (2026), via Our World in Data. “Large-scale” = training compute above 10²³ FLOP.
▸ Interact: click “Language” in the legend to drop it and compare the rest.
This is not a story about the future. Each dot is one European country’s share of adults who used a generative AI tool in 2025, the first year the figure was measured. The dashed line is the 36% European average; stems show how far each country sits above (blue) or below (orange) it. The spread is wide; from 17% in Turkey to 56% in Norway; but the direction is unmistakable, and among students the share runs higher still. The question for universities is no longer whether AI belongs in education. It already does.
Description: Deviation from the cross-country mean. Blue = above average, orange = below. Data: Eurostat (2026), via Our World in Data.
▸ Interact: hover any dot for the precise national figure.
All five visualisations are my own designs, built in R with the plotly package. The underlying data come from four independent, reputable providers (cited below) and are included as local CSV files for full reproducibility. Rather than reproduce the providers’ own charts, each visual applies an original transformation, that is, regional shares of investment, a derived race-to-parity timeline, annual (not cumulative) model counts, and deviation from the average and uses chart forms (line, 100%-stacked area, dumbbell, stacked bars, diverging lollipop) chosen to suit the message. Multi-series charts use the colour-blind-safe Okabe–Ito palette, and axes begin at zero where magnitudes are compared, to avoid exaggeration.
Eurostat. (2026). Use of generative AI tools by individuals [Data set]. Our World in Data. https://ourworldindata.org/grapher/share-europe-using-generative-ai
Kiela, D., Thrush, T., Ethayarajh, K., & Singh, A. (2023). Test scores of AI systems relative to human performance [Data set]. Our World in Data. https://ourworldindata.org/grapher/test-scores-ai-capabilities-relative-human-performance
Maslej, N., Fattorini, L., Perrault, R., Gil, Y., Parli, V., Brynjolfsson, E., Manyika, J., … Walsh, T. (2026). Private investment in AI by region [Data set]. Our World in Data. https://ourworldindata.org/grapher/private-investment-in-artificial-intelligence
Rahman, R., Owen, D., & You, J. (2026). Cumulative number of large-scale AI models by domain [Data set]. Our World in Data. https://ourworldindata.org/grapher/cumulative-number-of-large-scale-ai-models-by-domain
Generative AI (ChatGPT) was used to help locate open datasets, correct errors in the R/plotly code, and edit the content. All data were sourced from the independent providers cited above; no data were generated or simulated. All the data has been reviewed and all the final work is my own and I take responsibility for the final submission..