Two data visualizations exploring GPU hardware efficiency and U.S. data center electricity consumption — drawn from real datasets including LBNL 2024, the IEA Energy & AI report, and the MDPI Kappa-Energy Index paper.
Real-time health map for each visualization on this page. Live status updates automatically as data loads.
| Visualization | Status | Type | Notes |
|---|---|---|---|
| Visualization 01 — GPU Architecture Trade-off | Live | database.php | Connecting to database.php... |
| Visualization 02 — U.S. Electricity Over Time | Live | database.php | Connecting to database.php... |
| Visualization 03 — GPU Specs Dashboard | Live | database.php | Connecting to database.php... |
| Visualization 04 — IEA Demand Snapshot | Live | database.php | Connecting to database.php... |
| Visualization 05 — Electricity Surge + GPU Efficiency | Live | database.php | Connecting to database.php... |
| Visualization 06 — Regional Concentration Trends | Live | database.php | Connecting to database.php... |
Each point represents a live cluster record from database.php. The chart compares cluster power capacity against estimated AI-chip density, with bubble size reflecting the scale of installed accelerators.
| Cluster | AI Chips | Power (MW) | Chips / MW | Status | Country | Owner |
|---|---|---|---|---|---|---|
| Loading live cluster data… | ||||||
Hover over any bubble to see the cluster name, power draw, chip count, and country. Clusters in the top-left are the most efficient — high chip density at low power. Clusters in the bottom-right are power-hungry but chip-sparse. The ideal zone (dashed green box) marks clusters that pack the most AI compute per megawatt.
This visualization now uses the live datacenter_components table from database.php. It tracks how storage, networking, infrastructure, conventional servers, and AI servers contribute to total electricity demand.
| Year | Infrastructure (TWh) | Network (TWh) | Storage (TWh) | Conventional Servers (TWh) | AI Servers (TWh) | Total (TWh) | AI Share |
|---|
Hover over any year to see a full breakdown of electricity use by component for that year. The bright green band at the top is AI server load — watch how it grows from near-zero and begins pulling the total upward from 2020 onward. The dashed white line shows the overall total.
Interactive analytics across GPU hardware generations using only the live `gpu_specs` table from `database.php`. Filter by manufacturer, release year, and sort metric.
SOURCE — database.php · gpu_specs
Bubble size = memory bus width. Ideal: high shaders, large memory.
Horizontal bars sorted by GPU clock (MHz).
| Manufacturer | Product | Year | Mem (GB) | Bus Width | GPU Clock | Mem Clock | Shaders | Chip |
|---|---|---|---|---|---|---|---|---|
| Loading… | ||||||||
Use the filters at the top (Manufacturer, Year Range, Sort By) to narrow down the dataset. The bubble chart plots each GPU by memory size (X) vs. shader count (Y) — bubble size reflects memory bus width. The bar chart ranks GPUs by your chosen metric. Hover over any point or bar for full specs. The data table below shows all filtered records with raw numbers.
Live electricity-demand comparisons drawn from `regional_data_annex` plus the latest datacenter component mix from `datacenter_components`.
SOURCE — database.php · regional_data_annex + datacenter_components
World total electricity consumption pulled from `regional_data_annex`.
Top regions by 2030 base-case total electricity consumption.
This panel is driven directly by `regional_data_annex`, so region rankings update with the live database instead of a fixed snapshot.
| Dataset | 2022 | 2026 | Notes |
|---|
The three stat cards at the top show global electricity demand now vs. the 2030 base case — watch them count up from zero as the data loads. The bar chart on the left shows world totals by year (green = actual, blue = 2030 projection). The chart on the right compares the top 5 regions side-by-side: 2024 actual vs. 2030 base case. Hover any bar for exact TWh values. The regional cards below rank the biggest electricity consumers by projected 2030 demand.
Regional electricity growth and GPU hardware efficiency proxies, both rendered directly from live `regional_data_annex` and `gpu_specs` records.
SOURCE — database.php · regional_data_annex + gpu_specs
TWh · actual and IEA base-case projections
| Region | 2022 TWh | 2024 TWh | 2030 TWh | 2024–2030 Growth | Share (2024) |
|---|
NVIDIA architectures from Volta to Hopper — TDP, compute throughput, and efficiency per watt
| GPU | Architecture | Year | TDP (W) | FP16 TFLOPS | VRAM (GB) | Mem BW (TB/s) | TFLOPS/W | Tier |
|---|
Stacked area using live actual rows plus 2030 base-case projections from `regional_data_annex`.
X: TDP (W) · Y: FP16 TFLOPS/W · Bubble size: memory bandwidth
This section pairs two views. The regional electricity table shows how data center demand compares across regions from 2022 → 2030. The GPU efficiency bubble chart below it plots individual GPU models by estimated power draw (X) vs. compute efficiency (Y) — bigger bubbles indicate wider memory bandwidth. The core tension: GPU hardware got ~4× more efficient per watt from V100 to H100, but AI workloads grew 10× in the same window — hardware gains alone cannot offset demand growth.
Three interactive charts showing where global electricity demand is concentrated and how it shifts by 2030. Start with the constellation (circle size = 2024 demand), then follow growth in the growth ladder, and see proportional share change in the ribbon chart. Hover or click any region to highlight it across all three views.
SOURCE — database.php · regional_data_annex
Source figures come directly from live `regional_data_annex` records in `database.php`.
| Metric Group | Label | Region | Year | Value | Units | Status | Source |
|---|
Circle area scales to live 2024 electricity demand by region. Hover or click each region for details.
Bars compare 2024 actual values with 2030 base-case values for the world and the largest regions in the database.
Loading live growth comparison…
The same live regional totals in a proportional 2030 base-case ribbon.
Loading live concentration summary…
Loading live growth summary…
Loading live regional comparison…
Adjust the three key levers — facility efficiency, GPU generation, and grid carbon intensity — to see how total energy footprint and carbon output change. All calculations use real data ranges from the live database.