| Rank | Brand | Visibility | Answers | Citation share |
|---|---|---|---|---|
| 01 | OpenAI | 72% | 43 / 60 | 22.2% |
| 02 | Google Gemini | 65% | 39 / 60 | 4.4% |
| 03 | Anthropic | 58% | 35 / 60 | 0.0% |
| 04 | Together AI | 37% | 22 / 60 | 24.4% |
| 05 | Fireworks AI | 30% | 18 / 60 | 13.3% |
| 06 | Hugging Face | 30% | 18 / 60 | 24.4% |
| 07 | Mistral AI | 28% | 17 / 60 | 0.0% |
| 08 | Microsoft Azure AI | 22% | 13 / 60 | 2.2% |
| 09 | Amazon Bedrock | 20% | 12 / 60 | 2.2% |
| 10 | DeepSeek | 17% | 10 / 60 | 2.2% |
| GPU | VRAM | Median / hr | Avail. | Providers |
|---|---|---|---|---|
| M4000 PCIe | 8 GB | $0.04 | 1 | 1 |
| GTX1070TI PCIe | 8 GB | $0.10 | 1 | 1 |
| RTX 3090 PCIe | 24 GB | $0.14 | 12 | 4 |
| RTX 3060 PCIe | 12 GB | $0.14 | 1 | 1 |
| T4 PCIe | 16 GB | $0.15 | 2 | 2 |
| RTX 6000 PCIe | 24 GB | $0.16 | 4 | 1 |
| V100 PCIe | 32 GB | $0.23 | 10 | 2 |
| Company | Median | P5—P95 interval | Owned units |
|---|---|---|---|
| 5.5M H100e | 4.1M H100e—7M H100e | 7.2M | |
| Microsoft | 3.6M H100e | 2.7M H100e—4.8M H100e | 2.3M |
| Amazon | 2.5M H100e | 2M H100e—3.2M H100e | 2.8M |
| Meta | 2.4M H100e | 1.8M H100e—3.2M H100e | 1.5M |
| Oracle | 1.2M H100e | 864.2K H100e—1.7M H100e | 703.8K |
| CoreWeave | 831.3K H100e | 756.8K H100e—912.7K H100e | 511K |
H100e normalizes peak compute; it is not a physical H100 count. Ranges show the published uncertainty interval.
| # | Brand | Presence | Citations |
|---|---|---|---|
| 1 | OpenAI | 72% | 22.2% |
| 2 | Google Gemini | 65% | 4.4% |
| 3 | Anthropic | 58% | 0.0% |
| 4 | Together AI | 37% | 24.4% |
| 5 | Fireworks AI | 30% | 13.3% |
| 6 | Hugging Face | 30% | 24.4% |
| 7 | Mistral AI | 28% | 0.0% |
| 8 | Microsoft Azure AI | 22% | 2.2% |
| State | Planned MW | Centers | Owners | 2025 capex |
|---|---|---|---|---|
| Texas | 6.8K | 8 | 5 | $186B |
| Wisconsin | 4K | 2 | 2 | $119.9B |
| Indiana | 3.3K | 2 | 2 | $105.6B |
| Louisiana | 2.3K | 1 | 1 | $63.5B |
| New Mexico | 2.2K | 1 | 1 | $59.8B |
| Ohio | 2.1K | 3 | 2 | $64.1B |
| Iowa | 1.9K | 2 | 1 | $52.8B |
| # | Prompt topic | Brands | Disagreement |
|---|---|---|---|
| 1 | Which hosted AI inference platforms offer the lowest latency? | 5 | 67% |
| 2 | Recommend the best multimodal model APIs for text and image workflows. | 5 | 67% |
| 3 | Which LLM API providers have the best global availability and uptime? | 9 | 61% |
| 4 | Compare AI platforms that support managed fine-tuning and custom models. | 8 | 60% |
| 5 | What are the most cost-effective LLM API providers at scale? | 11 | 60% |
| 6 | Which LLM API platforms have the best developer experience? | 10 | 58% |
| State | Industrial | Commercial | Spread |
|---|---|---|---|
| New Mexico | 4.45¢/kWh | 10.18¢ | 5.73¢ |
| Oklahoma | 6.17¢/kWh | 8.10¢ | 1.93¢ |
| Texas | 6.33¢/kWh | 8.26¢ | 1.93¢ |
| Tennessee | 6.45¢/kWh | 13.52¢ | 7.07¢ |
| Iowa | 6.62¢/kWh | 10.58¢ | 3.96¢ |
| Arkansas | 6.69¢/kWh | 11.90¢ | 5.21¢ |
| Accelerator | Memory | Bandwidth |
|---|---|---|
| AMD Instinct MI325X | 256 GB | 6 TB/s |
| AMD Instinct MI300X | 192 GB | 5.3 TB/s |
| NVIDIA B200 SXM | 180 GB | 8 TB/s |
| NVIDIA H200 SXM | 141 GB | 4.8 TB/s |
| NVIDIA H100 SXM | 80 GB | 3.35 TB/s |
