---
title: "Powering Intelligence: How AI Is Rewiring Australia's Energy Future"
output:
  flexdashboard::flex_dashboard:
    orientation: columns
    vertical_layout: fill
    theme: flatly
    css: www/styles.css
    social: menu
    source_code: embed
    self_contained: true
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(
  echo = FALSE,
  message = FALSE,
  warning = FALSE
)

library(tidyverse)
library(plotly)
library(scales)
library(htmltools)
library(jsonlite)

source("data/load_data.R")

iea_global <- prep_iea_global()
aus_elec <- load_australia_electricity()
aus_mix <- prep_aus_elec_mix()
aemo_fy30 <- prep_aemo_scenarios_fy30()
aemo_hist <- prep_aemo_history_forecast()
aemo_regional <- prep_aemo_regional_long()
iea_regional <- load_iea_regional()

tc <- tc_colors

apa_source <- function(citation) {
  paste0('<span class="source">Source: ', citation, ".</span>")
}
```

Introduction {data-icon="fa-bolt"}
=======================================================================

<div class="intro-hero">

<p class="intro-kicker">Editor topic</p>
<h2 class="intro-headline">AI and humanity: studying in the age of generative AI</h2>

<p class="intro-deck">
Every ChatGPT query, every generated image, and every AI-assisted essay runs on data centres that consume real electricity. The International Energy Agency (IEA) estimates global data-centre demand reached <strong>415 TWh in 2024</strong> (1.5% of world electricity) and could exceed <strong>945 TWh by 2030</strong> as AI adoption accelerates.
</p>

<p class="intro-deck">
Australia's grid is greening <e2><80><94> renewables reached <strong>39%</strong> of generation in 2025 <e2><80><94> even as data-centre load on the National Electricity Market is forecast to <strong>triple to 12 TWh by FY30</strong>. The five charts below follow a deliberate arc: global AI energy demand, then Australia's electricity transition, rising generation, data-centre load on the grid, and where that load concentrates.
</p>

<div class="intro-stats">
<div class="stat-card"><span class="stat-value">415 TWh</span><span class="stat-label">Global data-centre use, 2024</span></div>
<div class="stat-card"><span class="stat-value">945 TWh</span><span class="stat-label">IEA forecast, 2030</span></div>
<div class="stat-card"><span class="stat-value">39%</span><span class="stat-label">Australian renewables, 2025</span></div>
<div class="stat-card"><span class="stat-value">12 TWh</span><span class="stat-label">Australia data-centre load, FY30</span></div>
</div>

<div class="intro-explore"><strong>How to explore:</strong> Use the button controls above each chart. Hover for exact values. Click legend items to show or hide series.</div>

</div>

Chart 1 {data-icon="fa-globe"}
=======================================================================

<div class="story-lead"><strong>1 <e2><80><94> Global context.</strong> Use the <strong>buttons top-right</strong> to switch between the worldwide outlook and regional growth to 2030.</div>
<div class="interaction-hint">Interactive: click <strong>Global outlook</strong> or <strong>Regional growth</strong> above the chart.</div>

```{r chart1}
iea_base <- iea_global |> filter(scenario == "Base Case")
iea_sens <- iea_global |> filter(year == 2035)
regional <- iea_regional |>
  mutate(
    region = reorder(region, increment_twh),
    label = paste0("+", increment_twh, " TWh (+", pct_increase_vs_2024, "%)")
  )

p1 <- plot_ly() |>
  add_trace(
    data = iea_base, x = ~year, y = ~consumption_twh,
    type = "scatter", mode = "lines+markers", name = "Consumption (TWh)",
    line = list(color = tc$blue, width = 3), marker = list(size = 9),
    hovertemplate = "<b>%{text}</b><br>Consumption: %{y} TWh<extra></extra>",
    text = ~paste0(scenario, " (", year, ")"), visible = TRUE
  ) |>
  add_trace(
    data = iea_sens, x = ~year, y = ~consumption_twh,
    type = "scatter", mode = "markers", name = "2035 sensitivity (TWh)",
    marker = list(size = 12, symbol = "diamond", color = ~scenario,
      colors = c("Lift-Off" = tc$coral, "High Efficiency" = tc$teal, "Headwinds" = tc$grey)),
    hovertemplate = "<b>%{text}</b><br>Consumption: %{y} TWh<extra></extra>",
    text = ~scenario, visible = TRUE
  ) |>
  add_trace(
    data = iea_base, x = ~year, y = ~share_global_elec_pct,
    type = "scatter", mode = "lines+markers", name = "Share of global elec. (%)",
    yaxis = "y2", line = list(color = tc$coral, dash = "dot", width = 2),
    marker = list(size = 7, color = tc$coral),
    hovertemplate = "Share: %{y}%<extra></extra>", visible = TRUE
  ) |>
  add_trace(
    data = regional, x = ~region, y = ~increment_twh,
    type = "bar", name = "Regional growth (TWh)",
    marker = list(color = tc$light_blue),
    text = ~label, textposition = "outside",
    hovertemplate = "<b>%{x}</b><br>Additional demand to 2030: %{y} TWh<extra></extra>",
    visible = FALSE, xaxis = "x2", yaxis = "y3"
  ) |>
  tc_layout(
    "Chart 1: Global data-centre electricity demand is accelerating",
    xaxis = list(title = "", tickvals = c(2024, 2030, 2035), domain = c(0, 1)),
    yaxis = list(title = yaxis_title("Consumption (TWh)", tc$blue)),
    yaxis2 = list(
      title = yaxis2_title("Share (%)", tc$coral),
      overlaying = "y", side = "right", range = c(0, 5),
      tickfont = list(color = axis_label_col(tc$coral)), automargin = TRUE
    ),
    xaxis2 = list(title = "", domain = c(0, 1), anchor = "y3", visible = FALSE),
    yaxis3 = list(title = "Additional TWh to 2030", domain = c(0, 1), anchor = "x2", visible = FALSE),
    legend = legend_inset(),
    margin = MARGIN_BTNS,
    updatemenus = list(
      view_buttons(
        active = 0,
        buttons = list(
          list(label = "Global outlook", method = "update",
            args = list(list(visible = c(TRUE, TRUE, TRUE, FALSE)),
              list(xaxis = list(visible = TRUE, title = ""),
                xaxis2 = list(visible = TRUE), yaxis3 = list(visible = FALSE)))),
          list(label = "Regional growth", method = "update",
            args = list(list(visible = c(FALSE, FALSE, FALSE, TRUE)),
              list(xaxis = list(visible = FALSE),
                xaxis2 = list(visible = TRUE, title = ""), yaxis3 = list(visible = TRUE))))
        )
      )
    )
  )

chart_div(
  p1,
  paste0("<strong>Multivariate:</strong> electricity consumption (TWh), share of global electricity (%), and regional additions. ",
    apa_source("International Energy Agency. (2025). <em>Energy and AI</em>")),
  chart_num = 1,
  aria_label = "Chart 1. Global data centre electricity demand with scenario sensitivity cases."
)
```

Chart 2 {data-icon="fa-leaf"}
=======================================================================

<div class="story-lead"><strong>2 <e2><80><94> Fuel composition (not volume).</strong> This chart shows <em>which</em> fuels supply Australia's electricity (% shares by coal, gas, wind, solar) <e2><80><94> the structural shift before we ask how much total power the grid produces.</div>
<div class="interaction-hint">Interactive: click <strong>2000-2010</strong>, <strong>2010-2020</strong>, or <strong>All years</strong> above the chart.</div>

```{r chart2}
mix_colors <- c("Coal" = tc$coal, "Gas" = tc$gas, "Wind" = tc$wind, "Solar" = tc$solar)

p2 <- plot_ly(
  data = aus_mix, x = ~year, y = ~share_pct, color = ~source, colors = mix_colors,
  type = "scatter", mode = "lines", stackgroup = "one", line = list(width = 1.2),
  hovertemplate = "<b>%{fullData.name}</b><br>Year: %{x}<br>Share: %{y:.1f}%<extra></extra>"
) |>
  tc_layout(
    "Chart 2: Australia's electricity mix is shifting from coal to renewables",
    xaxis = list(
      title = "Year", dtick = 5,
      rangeselector = list(
        y = 1.02, yanchor = "bottom",
        buttons = list(
          list(count = 10, label = "2000-2010", step = "year", stepmode = "backward"),
          list(count = 20, label = "2010-2020", step = "year", stepmode = "backward"),
          list(step = "all", label = "All years")
        )
      ),
      rangeslider = list(visible = FALSE)
    ),
    yaxis = list(title = "Share of generation (%)", range = c(0, 100)),
    legend = legend_inset(),
    margin = MARGIN_CHART2
  )

chart_div(
  p2,
  paste0("<strong>Multivariate:</strong> fuel-source shares (% of generation) <e2><80><94> composition, not total volume. ",
    apa_source("Ritchie, H., Roser, M., & Rosado, P. (2025). <em>Energy dataset</em>. Our World in Data")),
  chart_num = 2,
  aria_label = "Chart 2. Stacked area of Australian electricity generation shares by fuel."
)
```

Chart 3 {data-icon="fa-chart-line"}
=======================================================================

<div class="story-lead"><strong>3 <e2><80><94> Scale and headroom.</strong> Unlike Chart 2's fuel shares, this chart measures <em>how much</em> electricity Australia generates (TWh) and whether renewables are keeping pace <e2><80><94> or whether per-capita use is still rising as AI load arrives.</div>
<div class="interaction-hint">Interactive: click <strong>Renewables %</strong> or <strong>Per capita</strong> top-right above the chart.</div>

```{r chart3}
aus_dual <- aus_elec |> filter(year >= 2000)

p3 <- plot_ly(data = aus_dual) |>
  add_trace(
    x = ~year, y = ~electricity_generation, name = "Generation (TWh)",
    type = "scatter", mode = "lines+markers",
    line = list(color = tc$blue, width = 2.5), marker = list(size = 5),
    hovertemplate = "Generation: %{y:.1f} TWh<extra></extra>"
  ) |>
  add_trace(
    x = ~year, y = ~renewables_share_elec, name = "Renewables share (%)",
    type = "scatter", mode = "lines+markers", yaxis = "y2",
    line = list(color = tc$teal, dash = "dot", width = 2.5),
    marker = list(size = 5, symbol = "diamond"),
    hovertemplate = "Renewables: %{y:.1f}%<extra></extra>", visible = TRUE
  ) |>
  add_trace(
    x = ~year, y = ~per_capita_electricity, name = "Per-capita use (kWh)",
    type = "scatter", mode = "lines+markers", yaxis = "y2",
    line = list(color = tc$orange, width = 2), marker = list(size = 4),
    hovertemplate = "Per capita: %{y:,.0f} kWh<extra></extra>", visible = FALSE
  ) |>
  tc_layout(
    "Chart 3: Generation is rising as renewables take a larger share",
    xaxis = list(title = "Year", dtick = 5),
    yaxis = list(title = yaxis_title("Generation (TWh)", tc$blue), automargin = TRUE),
    yaxis2 = list(
      title = yaxis2_title("Renewables (%)", tc$teal),
      overlaying = "y", side = "right", range = c(0, 110),
      tickfont = list(color = axis_label_col(tc$teal)), automargin = TRUE
    ),
    legend = legend_inset(),
    margin = MARGIN_BTNS,
    updatemenus = list(
      view_buttons(
        active = 0,
        buttons = list(
          list(label = "Renewables %", method = "update",
            args = list(list(visible = c(TRUE, TRUE, FALSE)),
              list(yaxis2 = list(
                title = yaxis2_title("Renewables (%)", tc$teal),
                range = c(0, 110), overlaying = "y", side = "right",
                tickfont = list(color = axis_label_col(tc$teal)), automargin = TRUE
              )))),
          list(label = "Per capita", method = "update",
            args = list(list(visible = c(TRUE, FALSE, TRUE)),
              list(yaxis2 = list(
                title = yaxis2_title("Per capita (kWh)", tc$orange),
                range = c(9000, 11000), overlaying = "y", side = "right",
                tickfont = list(color = axis_label_col(tc$orange)), automargin = TRUE
              ))))
        )
      )
    )
  )

chart_div(
  p3,
  paste0("<strong>Multivariate:</strong> absolute generation (TWh) paired with renewables share (%) or per-capita use (kWh) <e2><80><94> scale and headroom, not fuel mix. ",
    apa_source("Ritchie, H., Roser, M., & Rosado, P. (2025). <em>Energy dataset</em>. Our World in Data")),
  chart_num = 3,
  aria_label = "Chart 3. Dual axis chart of Australian electricity generation and renewables or per capita use."
)
```

Chart 4 {data-icon="fa-server"}
=======================================================================

<div class="story-lead"><strong>4 <e2><80><94> AI on the grid.</strong> Data-centre demand could reach <strong>12 TWh by FY30</strong> under AEMO's Step Change scenario <e2><80><94> toggle to compare FY30 sensitivity cases.</div>
<div class="interaction-hint">Interactive: click <strong>Timeline</strong> or <strong>FY30 scenarios</strong> top-right above the chart.</div>

```{r chart4}
scenario_cols <- c(
  "Progressive Change" = tc$grey, "Step Change" = tc$blue,
  "Green Energy Industries" = tc$teal, "Rapid AI Uptake" = tc$coral
)

p4 <- plot_ly() |>
  add_trace(
    data = aemo_hist, x = ~financial_year, y = ~consumption_twh,
    type = "scatter", mode = "lines+markers", name = "Consumption (TWh)",
    line = list(color = tc$blue, width = 3),
    marker = list(size = ~if_else(period == "Historical", 9, 11),
      symbol = ~if_else(period == "Historical", "circle", "diamond"),
      color = ~if_else(period == "Historical", tc$blue, tc$coral)),
    hovertemplate = "<b>%{x}</b><br>Consumption: %{y} TWh<extra></extra>", visible = TRUE
  ) |>
  add_trace(
    data = aemo_hist, x = ~financial_year, y = ~nem_grid_share_pct,
    type = "scatter", mode = "lines+markers", name = "NEM grid share (%)",
    yaxis = "y2", line = list(color = tc$coral, dash = "dot", width = 2),
    marker = list(size = 7, color = tc$coral),
    hovertemplate = "NEM share: %{y}%<extra></extra>", visible = TRUE
  ) |>
  add_trace(
    data = aemo_fy30, x = ~scenario, y = ~consumption_twh,
    type = "bar", name = "FY30 scenarios (TWh)",
    marker = list(color = ~scenario, colors = scenario_cols),
    text = ~paste0(consumption_twh, " TWh"), textposition = "outside",
    hovertemplate = "<b>%{x}</b><br>FY30: %{y} TWh<extra></extra>",
    visible = FALSE, xaxis = "x2", yaxis = "y3"
  ) |>
  tc_layout(
    "Chart 4: Australian data-centre demand - history and scenarios",
    xaxis = list(title = "", categoryorder = "array",
      categoryarray = aemo_hist$financial_year, domain = c(0, 1)),
    yaxis = list(title = yaxis_title("Consumption (TWh)", tc$blue)),
    yaxis2 = list(
      title = yaxis2_title("NEM share (%)", tc$coral),
      overlaying = "y", side = "right", range = c(0, 14),
      tickfont = list(color = axis_label_col(tc$coral)), automargin = TRUE
    ),
    xaxis2 = list(title = "", domain = c(0, 1), anchor = "y3", visible = FALSE),
    yaxis3 = list(title = "FY30 consumption (TWh)", domain = c(0, 1),
      anchor = "x2", visible = FALSE, range = c(0, 18)),
    legend = legend_inset(),
    margin = MARGIN_BTNS,
    updatemenus = list(
      view_buttons(
        active = 0,
        buttons = list(
          list(label = "Timeline", method = "update",
            args = list(list(visible = c(TRUE, TRUE, FALSE)),
              list(xaxis = list(visible = TRUE), xaxis2 = list(visible = FALSE),
                yaxis3 = list(visible = FALSE)))),
          list(label = "FY30 scenarios", method = "update",
            args = list(list(visible = c(FALSE, FALSE, TRUE)),
              list(xaxis = list(visible = FALSE), xaxis2 = list(visible = TRUE),
                yaxis3 = list(visible = TRUE))))
        )
      )
    )
  )

chart_div(
  p4,
  paste0("<strong>Multivariate:</strong> data-centre consumption (TWh), NEM grid share (%), and FY30 scenarios. ",
    apa_source("Oxford Economics Australia. (2025). <em>Data centre energy demand</em> (Prepared for AEMO)")),
  chart_num = 4,
  aria_label = "Chart 4. Australian data centre electricity consumption timeline and FY30 scenarios."
)
```

Chart 5 {data-icon="fa-city"}
=======================================================================

<div class="story-lead"><strong>5 <e2><80><94> Local geography.</strong> Demand is concentrating in <strong>Sydney</strong> and <strong>Melbourne</strong> <e2><80><94> filter by region to see who bears the load.</div>
<div class="interaction-hint">Interactive: click <strong>All</strong>, <strong>Sydney</strong>, <strong>Melbourne</strong>, or <strong>Other</strong> top-right above the chart.</div>

```{r chart5}
regional_cols <- c("Sydney" = tc$blue, "Melbourne" = tc$coral, "Other regions" = tc$grey)

p5 <- plot_ly(
  data = aemo_regional, x = ~financial_year, y = ~consumption_twh,
  color = ~region, colors = regional_cols, type = "bar",
  text = ~paste0(consumption_twh, " TWh"), textposition = "inside",
  insidetextfont = list(color = "white", size = 10),
  hovertemplate = "<b>%{fullData.name}</b><br>%{x}: %{y:.1f} TWh<extra></extra>"
) |>
  tc_layout(
    "Chart 5: Data-centre demand is concentrating in Sydney and Melbourne",
    barmode = "stack",
    xaxis = list(title = "Financial year"),
    yaxis = list(title = "Electricity consumption (TWh)", range = c(0, 40)),
    legend = legend_inset(),
    margin = MARGIN_BTNS,
    annotations = list(
      list(x = "FY25", y = 1.0, yref = "paper", text = "Melbourne: 20%",
        showarrow = FALSE, font = list(size = 9, color = tc$coral), xanchor = "center"),
      list(x = "FY50", y = 1.0, yref = "paper", text = "Melbourne: 41%",
        showarrow = FALSE, font = list(size = 9, color = tc$coral), xanchor = "center")
    ),
    updatemenus = list(
      view_buttons(
        active = 0,
        buttons = list(
          list(label = "All", method = "restyle", args = list(list(visible = c(TRUE, TRUE, TRUE)))),
          list(label = "Sydney", method = "restyle", args = list(list(visible = c(TRUE, FALSE, FALSE)))),
          list(label = "Melbourne", method = "restyle", args = list(list(visible = c(FALSE, TRUE, FALSE)))),
          list(label = "Other", method = "restyle", args = list(list(visible = c(FALSE, FALSE, TRUE))))
        )
      )
    )
  )

chart_div(
  p5,
  paste0("<strong>Multivariate:</strong> regional consumption (TWh) for FY25 and FY50. ",
    apa_source("Oxford Economics Australia. (2025). <em>Data centre energy demand</em> (Prepared for AEMO)")),
  chart_num = 5,
  aria_label = "Chart 5. Stacked bar chart of Australian data centre consumption by region."
)
```
Closing {data-icon="fa-flag-checkered"}
=======================================================================

<div class="closing-hero">

<p class="closing-kicker">The human stakes</p>
<h2 class="closing-headline">What AI's energy cost means for Australia</h2>

<div class="closing-grid">

<div class="closing-card closing-gain"><h3>What may be gained</h3><p>AI can democratise access to knowledge, accelerate research, and support students who use generative tools to learn <e2><80><94> provided the energy system keeps pace. Australia has a genuine advantage: a rapidly greening grid and a stable regulatory environment that attracts hyperscale investment.</p></div>

<div class="closing-card closing-loss"><h3>What may be lost</h3><p>The efficiency gains that offset earlier waves of digitalisation may not hold. The IEA warns that in a Lift-Off scenario, global data-centre demand could exceed <strong>1,700 TWh by 2035</strong>. Benefits could concentrate in Sydney and Melbourne while grid, water, and land costs are borne more broadly.</p></div>

<div class="closing-card closing-act"><h3>How to prepare and act</h3><ul class="closing-list"><li><strong>Ask where your data lives.</strong> Using an AI tool is also an energy choice.</li><li><strong>Demand transparency.</strong> Universities and tech firms should publish energy use data.</li><li><strong>Plan the grid now.</strong> AEMO forecasts data centres at <strong>12% of NEM supply by FY50</strong>.</li></ul></div>

</div>

<p class="closing-final">
The invisible cost of intelligence is becoming visible. These five charts are a pitch for a story Australia needs to tell <e2><80><94> before the servers outgrow the conversation.
</p>

</div>

<div class="closing-section acknowledgements">

<h2 class="section-heading">Acknowledgements</h2>

<p>Data visualisations created in R (plotly, flexdashboard). Datasets: Our World in Data Energy Dataset; International Energy Agency (2025) <em>Energy and AI</em>; Oxford Economics Australia (2025) report prepared for AEMO. Manually extracted IEA and AEMO figures documented in project CSV files and referenced in chart captions.</p>

</div>

<div class="closing-section references">

<h2 class="section-heading">References</h2>

<p>International Energy Agency. (2025). <em>Energy and AI: Energy demand from AI</em>. https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai</p>

<p>Oxford Economics Australia. (2025). <em>Data centre energy demand: Final report</em> (Prepared for the Australian Energy Market Operator). https://www.aemo.com.au/-/media/files/stakeholder_consultation/consultations/nem-consultations/2024/2025-iasr-scenarios/final-docs/oxford-economics-australia-data-centre-energy-consumption-report.pdf</p>

<p>Ritchie, H., Roser, M., & Rosado, P. (2025). <em>Energy dataset</em>. Our World in Data. https://github.com/owid/energy-data</p>

<p>Rosado, P. (2020). <em>Energy mix</em>. Our World in Data. https://ourworldindata.org/energy-mix</p>

</div>