AI chip · NVIDIA
NVIDIA B200 (Blackwell)
Full-power Blackwell data-center GPU (~1000W TDP), 192 GB HBM3e. The core building block of the GB200 superchip.
Deployed in 3 named data centers · Part of 1 rack design.
Market position
B200 is Blackwell's HGX-form-factor chip: 2.5x H100 dense FP8, 192 GB HBM3e, NVLink 5. Meta Prometheus, Microsoft Fairwater and CoreWeave started their Blackwell fleets here rather than jump straight to GB200 NVL72 rack-scale.
Efficiency and power
How much work you get per watt and per dollar, and how much power a full rack draws.
- Perf per watt
- 4.50 FP8 TFLOPS/W4,500 TFLOPS ÷ 1000 W = 4.50
- Rack power
- 14.3 kW per NVIDIA HGX B200 (8-GPU baseboard) (~1788 W per chip)14.3 kW ÷ 8 chips (whole-system, includes CPU, memory, NICs, cooling)
- Cooling
- hybridDepends on the rack design
What fits in 192 GB
Which open-source LLMs run on one NVIDIA B200 (Blackwell), by precision. Weights only: add roughly 15% headroom for real serving. If a model does not fit at FP16, try INT8 or INT4 (smaller quality trade-off than most people expect).
| Model | Params | FP16 | INT8 | INT4 |
|---|---|---|---|---|
| Llama 3.1 8B dense | 8 B | 16 GB ✓ | 8 GB ✓ | 4 GB ✓ |
| Llama 3.1 70B dense | 70 B | 140 GB ✓ | 70 GB ✓ | 35 GB ✓ |
| Llama 3.1 405B dense | 405 B | 810 GB ✗ | 405 GB ✗ | 203 GB ✗ |
| Llama 3.3 70B dense | 70 B | 140 GB ✓ | 70 GB ✓ | 35 GB ✓ |
| DeepSeek V3 MoE (671B total, 37B active per token) | 671 B | 1342 GB ✗ | 671 GB ✗ | 336 GB ✗ |
| DeepSeek R1 MoE (671B total, 37B active per token) | 671 B | 1342 GB ✗ | 671 GB ✗ | 336 GB ✗ |
| Qwen 2.5 7B dense | 7 B | 14 GB ✓ | 7 GB ✓ | 4 GB ✓ |
| Qwen 2.5 72B dense | 72 B | 144 GB ✓ | 72 GB ✓ | 36 GB ✓ |
| Mixtral 8x7B MoE (46.7B total, 12.9B active per token) | 46.7 B | 93 GB ✓ | 47 GB ✓ | 23 GB ✓ |
| Mixtral 8x22B MoE (141B total, 39B active per token) | 141 B | 282 GB ✗ | 141 GB ✓ | 71 GB ✓ |
| Gemma 2 27B dense | 27 B | 54 GB ✓ | 27 GB ✓ | 14 GB ✓ |
| Command R+ dense | 104 B | 208 GB ✗ | 104 GB ✓ | 52 GB ✓ |
| Kimi K2 MoE (1T total, 32B active per token) | 1000 B | 2000 GB ✗ | 1000 GB ✗ | 500 GB ✗ |
Math: FP16 = params × 2 bytes; INT8 = params × 1 byte; INT4 = params × 0.5 bytes. MoE models sum every expert (full weights on disk), not the per-token active subset.
Common questions
How much does NVIDIA B200 (Blackwell) cost?
NVIDIA B200 (Blackwell) doesn't have a public launch list price. Vendors like this one usually price through direct sales rather than a public sheet. See list price per TFLOP for chips that do publish.
How much memory does NVIDIA B200 (Blackwell) have?
192 GB of HBM3e, running at 8,000 GB/s. Straight from the vendor datasheet. See chips with the most memory for context.
How much power does one NVIDIA B200 (Blackwell) draw?
1000 W at the chip. A full server draws more once you add CPU, memory, networking and cooling: see the rack power number above. Compare to other chips on perf-per-watt.
Which data centers use NVIDIA B200 (Blackwell)?
3 named data centers run them, including xAI Colossus 2, Stargate Abilene and Stargate Doña Ana County / Project Jupiter (New Mexico). See who has the most NVIDIA B200 (Blackwell) for the ranked list.
Which open-source LLMs fit on one NVIDIA B200 (Blackwell)?
Of 13 open-source LLMs we track, 7 fit at FP16, 9 at INT8, and 9 at INT4 (for example Llama 3.1 8B, Llama 3.1 70B and Llama 3.3 70B). Full table above with each model's memory need.
Key facts
- Class
- Compute SoC (AI accelerator)
- Designer
- NVIDIA
- Safety-critical?
- No (data-center inference / training)
- Record as of
- 2026-09-27
- Most recent source
- 2026-06-01 (across 7 sources on this page)
Specifications (15 fields, click to expand)
Straight from the vendor datasheet. Dense throughput shown first; sparse (2:4) numbers in parentheses where the vendor publishes them. Full datasheet linked below.
- Process node
- TSMC 4NP
- Transistors
- 208 B
- CUDA cores
- 33,792
- Tensor cores
- 1,056 (5th gen (Blackwell))
- TDP
- 1000 W
- Memory
- 192 GB HBM3e
- Memory bandwidth
- 8,000 GB/s
- PCIe
- Gen 5 x16 (1,800 GB/s)
- FP16 (dense)
- 2,250 TFLOPS (4,500 sparse)
- FP8 (dense)
- 4,500 TFLOPS (9,000 sparse)
- FP4 (dense)
- 10,000 TFLOPS (20,000 sparse)
- INT8 (dense)
- 4,500 TOPS (9,000 sparse)
- Form factor
- SXM6
- Announced
- 2024-03-18
- Released
- 2024-11-01
Source: vendor datasheet
Generation
Foundry & process
- Fabbed at
- TSMC (all chips TSMC makes →)
- Process node
- TSMC 4NP
- Transistors
- 208 B
Rack designs that use it
Whole-rack systems buyers order in bulk (NVL72, HGX, TPU pods) that ship with NVIDIA B200 (Blackwell) inside.
- ×8NVIDIA HGX B200 (8-GPU baseboard)platform
Compare with
Benchmarks
Published performance numbers by workload. Vendor datasheet figures where noted; independent measurements otherwise.
Data centers running NVIDIA B200 (Blackwell)
Who has the most? →Named data-center campuses with NVIDIA B200 (Blackwell) on site. Counts shown where the operator has published them; other rows are described in general terms.
- xAI Colossus 2 Data Centeras of 2026-06-01partially energizedreported
Blackwell generation for Colossus 2 expansion
- Stargate Abileneas of 2026-03-01partially energizedreported
Nvidia Blackwell chips filling ~0.3 GW of energized capacity across 4 of 8 buildings; ramping toward 1.2 GW
- Stargate Doña Ana County / Project Jupiter (New Mexico)as of 2025-01-21under constructioninferred
Planned Blackwell capacity (Nvidia default for new Stargate builds; specific SKU pending disclosure)
Inferred from Stargate program's consistent Blackwell posture per OpenAI + Nvidia partnership; upgrade when G42/OpenAI confirms
Sources
- Blackwell architectureNvidia
Adoption rows appear as we confirm each chip-in-robot pairing from a public source. See every chip we track for the full catalog or the NVIDIA page.