At least ~500K Trainium2
Project Rainier spans multiple U.S. data centers ↗, initially with nearly 500K Trainium2 chips and a path to more than 1M for Claude. That is approximately 150K B200e initially or 300K B200e at 1M chips.
A country-by-country audit of physical NVIDIA data-center GPUs, dated deployments, modeled capacity and announced ceilings.
This report estimates how many physical NVIDIA data-center GPUs may be located in each country by the end of 2027. It combines company and government announcements, regulatory filings, reported orders, dated construction plans and—where explicitly labeled—capacity modeling. Sources sometimes report performance-equivalent rather than physical GPUs, global fleets without country allocation, overlapping customer and operator capacity, or multi-year programs whose later phases extend beyond 2027. Each country row therefore separates firm evidence, a comparable 2027 scenario and longer-term ceilings. The figures are transparent evidence-based estimates, not audited inventories or guaranteed forecasts.
Click any column heading to sort. The default “2027 scenario” is the broad comparison number from the audit; it mixes firm counts, modeled capacity and program ceilings, all explicitly labeled.
| Total rank ↕ | Country ↕ | 2027 scenario GPUs ↕ | GPUs / million ↕ | Per-capita rank ↕ | 2027 population ↕ | Evidence & sources |
|---|
Population is the UN World Population Prospects 2024 medium projection for July 1, 2027. Per-capita values inherit every uncertainty in the GPU numerator.
A separate, model-sensitive view of theoretical dense BF16 Tensor Core throughput. Physical counts above remain the primary inventory view.
| B200e rank ↕ | Country ↕ | Physical GPUs ↕ | B200 equivalents ↕ | B200e / GPU ↕ | Reported or inferred GPU types | Model note |
|---|
Actual TPU and Trainium quantities, kept separate from the NVIDIA rankings because complete country allocation is not public.
| Accelerator generation | Global units through Q2 2026 | B200e / chip | Global B200e | Evidence treatment |
|---|---|---|---|---|
| TPU v4 + v4i | 293,687 | 0.116 weighted | 34,054 | Legacy group; compute uses Epoch's mapping. |
| TPU v5e | 2,357,619 | 0.088 | 206,423 | Google BF16 specification. |
| TPU v5p | 665,300 | 0.204 | 135,721 | Google BF16 specification. |
| TPU v6e | 2,599,160 | 0.408 | 1,060,457 | Google publishes 918 TFLOPS BF16 per chip. |
| TPU v7 | 677,262 | 1.025 | 694,419 | Google publishes 2,307 TFLOPS BF16 per chip. |
| All Google TPUs | 6,593,028 4,382,010–9,108,422 | 0.323 weighted | 2,131,074 1,441,204–2,892,520 | Global modeled shipments; not a U.S. inventory. |
| AWS Trainium2 | 1,400,000 global | ≈0.30 | ≈420,000 global | Company-disclosed units; B200e uses Epoch's compute mapping. |
Project Rainier spans multiple U.S. data centers ↗, initially with nearly 500K Trainium2 chips and a path to more than 1M for Claude. That is approximately 150K B200e initially or 300K B200e at 1M chips.
If 50%–70% of Google's modeled global TPU stock is U.S.-located—the same broad location sensitivity used for the NVIDIA ownership cross-check—it would equal about 1.07M–1.49M B200e. This is an explicit assumption, not a disclosed Google country allocation, and is not used in either ranking above.
GPUs per million = scenario GPUs ÷ projected 2027 population × 1,000,000.Deduplicated from the original source ledger, every country comment and the expanded audit report.