| 1 | Power generation and fuel | MW | firm electricity and the fuel behind it (gas turbines and gas supply; nuclear restarts and uprates; solar with batteries; on-site generation) | 3 | – | 1,280.8 GW us_utility_scale_generating_capacity 86 GW planned_capacity_additions_2026 | NYSE:GEV, ETR:ENR, TYO:7011, NASDAQ:CEG |
|---|
| 2 | Grid, interconnection and transformers | MW connected | a connection that can deliver the power to a site: transmission, substations, transformers and switchgear | 3 | – | 2,061 GW us_active_interconnection_queue_capacity 253 GW queue_natural_gas_capacity | TYO:6501, NYSE:GEV, ETR:ENR, NYSE:ETN |
|---|
| 3 | Data centre shell and cooling | MW of IT load | a building with power, cooling and networking ready for racks, sold as capacity by the megawatt | 1.5 | – | 1.4 % of built colocation capacity us_primary_market_vacancy_rate 7,481.1 MW of colocation capacity us_primary_market_under_construction | NASDAQ:EQIX, NYSE:DLR, private, private (owned by Blackstone, NYSE:BX) |
|---|
| 4 | Wafers and advanced packaging | wafers per month | leading-edge logic wafers and the CoWoS packaging that joins a GPU die to its memory | 2 | – | 89 USD billion per quarter nvidia_data_center_revenue_q2_fy2027 75.2 USD billion per quarter nvidia_data_center_revenue_q1_fy2027 | NASDAQ:NVDA, NASDAQ:AMD, NASDAQ:AVGO, NASDAQ:MRVL |
|---|
| 5 | Memory and storage (HBM, DRAM, NAND) | GB and HBM stacks | high-bandwidth memory stacked on each GPU, server DRAM, and the NAND behind enterprise SSDs | 1.5 | – | 56.4 % of global HBM revenue hbm_market_share_sk_hynix_1q26 50 % of global HBM revenue hbm_market_share_sk_hynix_2q26 | KRX:000660 (ADS: NASDAQ:SKHY), KRX:005930, NASDAQ:MU, NASDAQ:SNDK |
|---|
| 6 | GPUs, servers, racks and networking | GPUs | accelerators assembled into servers and racks with the network that ties them into a cluster | 0.75 | – | 2.8 USD million per GB200 NVL72 rack (low end of reported range) gb200_nvl72_rack_price_low 400,000 USD per 8-GPU HGX B200 server (low end of reported range) hgx_b200_8gpu_server_price_low | TWSE:2317, TWSE:2382, TWSE:6669, NASDAQ:SMCI |
|---|
| 7 | GPU-hours | GPU-hour | rented accelerator time under multi-year contracts (neoclouds) or used internally (hyperscalers) | 1 | – | 130 USD bn (2026 capital expenditures incl. principal payments on finance leases, low end of range) meta_capex_guidance_2026_low 145 USD bn (2026 capital expenditures incl. principal payments on finance leases, high end of range) meta_capex_guidance_2026_high | NASDAQ:MSFT, NASDAQ:GOOGL, NASDAQ:AMZN, NASDAQ:META |
|---|
| 8 | Tokens | tokens | model output sold by the million tokens, trained and served on rented or owned GPU-hours | 0.5 | – | 3,200 trillion tokens per month (3.2 quadrillion) google_tokens_processed_per_month 480 trillion tokens per month google_tokens_processed_per_month_prior_year | NASDAQ:GOOGL, private, private, NASDAQ:META |
|---|
| 9 | Applications and end demand | tokens consumed per user per day | products that turn tokens into revenue, split by how much compute each consumes | 0.25 | – | 3,200 trillion tokens per month across Google products and APIs (floor: 'over 3.2 quadrillion') google_tokens_processed_per_month 950 million monthly active users of the Gemini app gemini_app_monthly_active_users | NASDAQ:MSFT, private, NASDAQ:GOOGL, private |
|---|