Data Center GPU/Accelerator Shipments
Estimated quarterly data center AI accelerator shipments by company (thousands of units). Sources: industry estimates, TrendForce.
Estimated quarterly data center GPU/accelerator shipments (thousands of units). Sources: industry estimates, TrendForce, JPMorgan research.
AI Accelerator Architecture Comparison
Side-by-side comparison of AI training and inference accelerators — specs, memory, performance, and pricing
| Product | Architecture | Node | Transistors | Memory | Mem BW | FP16 TFLOPS | FP8 TFLOPS | TDP | Interconnect | Launched | Price Est. |
|---|---|---|---|---|---|---|---|---|---|---|---|
 H100 SXM | Hopper | 4nm (TSMC) | 80B | 80GB HBM3 | 3.35 TB/s | 1,979 | 3,958 | 700W | NVLink 4.0 (900 GB/s) | 2023-Q1 | $25,000–$40,000 |
| B200 | Blackwell | 4nm (TSMC) | 208B | 192GB HBM3E | 8 TB/s | 5,000 | 10,000 | 1000W | NVLink 5.0 (1.8 TB/s) | 2024-Q4 | $30,000–$50,000 |
| GB200 NVL72 | Blackwell | 4nm (TSMC) | 208B (x72) | 13.5TB HBM3E | 576 TB/s | 162K | 1440K | 120kW | NVLink 5.0 (full rack) | 2025-Q1 | $2M–$3M (rack) |
 MI300X | CDNA 3 | 5nm/6nm (TSMC) | 153B | 192GB HBM3 | 5.3 TB/s | 1,307 | 2,614 | 750W | Infinity Fabric (896 GB/s) | 2023-Q4 | $10,000–$15,000 |
| MI325X | CDNA 3.5 | 5nm/6nm (TSMC) | 153B | 256GB HBM3E | 6 TB/s | 1,307 | 2,614 | 750W | Infinity Fabric (896 GB/s) | 2024-Q4 | $15,000–$20,000 |
| MI350X | CDNA 4 | 3nm (TSMC) | 185B | 288GB HBM3E | 8 TB/s | 5,000 | 10,000 | 1000W | UALink / Infinity Fabric 4.0 | 2025-Q2 | TBD |
 Gaudi 2 | Gaudi | 7nm (TSMC) | ~25B | 96GB HBM2E | 2.46 TB/s | 432 | 864 | 600W | 24x 100GbE RDMA | 2023-Q2 | $12,000–$15,000 |
| Gaudi 3 | Gaudi | 5nm (TSMC) | ~50B | 128GB HBM2E | 3.7 TB/s | 1,835 | 3,670 | 900W | 24x 200GbE RDMA | 2024-Q4 | $15,000–$20,000 |
 Google TPU v5e | Custom ASIC | 7nm (TSMC) | ~20B | 16GB HBM2E | 820 GB/s | 197 | 394 | 200W | ICI (Inter-Chip Interconnect) | 2023-Q3 | Cloud-only |
| Google TPU v6e (Trillium) | Custom ASIC | 5nm (TSMC) | ~35B | 32GB HBM3 | 1.6 TB/s | 460 | 920 | 300W | ICI v2 | 2024-Q4 | Cloud-only |
Specifications from official datasheets. Prices are estimated street prices or reported CSP costs. GB200 NVL72 figures are for the full 72-GPU rack system. Sources: NVIDIA, AMD, Intel, Broadcom datasheets and press releases.
AI Chip Product Roadmap
Product availability timeline for AI accelerators across NVIDIA, AMD, Intel, and Broadcom/Google
Legacy
In Production
Announced
Roadmap
2020
2021
2022
2023
2024
2025
2026
2027
2028
NVIDIAA100 (7nm)
H100 (4nm)
H200 (4nm)
B100/B200 (4nm)
GB200 NVL72 (4nm)
Rubin (R100) (3nm)
Feynman (2nm)
AMDMI250X (6nm)
MI300X (5nm)
MI325X (5nm)
MI350X (3nm)
MI400 (2nm)
IntelGaudi 2 (7nm)
Gaudi 3 (5nm)
Jaguar Shores (Intel 18A)
BroadcomGoogle TPU v5e (7nm)
Google TPU v6e (Trillium) (5nm)
Google TPU v7 (Ironwood) (5nm)
Meta MTIA v2 (5nm)
Product availability timeline based on official announcements and analyst roadmaps. “Announced” products have confirmed specs but are not yet shipping. Sources: company keynotes, product launches, investor presentations.