Google TPU 8i vs Google TPU 8t
Side-by-side specifications, performance, and rental pricing for datacenter AI workloads.
TPU 8i: TPU 8, 2026 TPU 8t: TPU 8, 2026
Memory
288 vs 216 GB
72 GB more for the TPU 8i
Bandwidth
8.6 TB/s vs 6.5 TB/s
TPU 8i moves data faster
Google TPU 8i
Full specs →Architecture
TPU 8
Year
2026
Memory
288 GB
Google TPU 8t
Full specs →Architecture
TPU 8
Year
2026
Memory
216 GB
Specifications
TPU 8i TPU 8t
Architecture
TPU 8TPU 8
Launch Year
20262026
Form Factor
——
Memory
288 GB216 GB
Memory Bandwidth
8.6 TB/s6.5 TB/s
TDP
——
Process Node
——
Spec TPU 8i TPU 8t
Architecture TPU 8TPU 8
Launch Year 20262026
Form Factor ——
Memory 288 GB216 GB
Memory Bandwidth 8.6 TB/s6.5 TB/s
TDP ——
Process Node ——
Performance (TFLOPS)
TPU 8i TPU 8t
FP64
No verified data available No verified data available
FP32
No verified data available No verified data available
TF32
No verified data available No verified data available
BF16
No verified data available No verified data available
FP16
No verified data available No verified data available
FP8
No verified data available No verified data available
FP6
No verified data available No verified data available
FP4
10,100 TFLOPS 12,600 TFLOPS
INT8
No verified data available No verified data available
Precision TPU 8i TPU 8t
FP64 No verified data available No verified data available
FP32 No verified data available No verified data available
TF32 No verified data available No verified data available
BF16 No verified data available No verified data available
FP16 No verified data available No verified data available
FP8 No verified data available No verified data available
FP6 No verified data available No verified data available
FP4 10,100 TFLOPS 12,600 TFLOPS
INT8 No verified data available No verified data available
FLOPS by Precision
What actually differs
Both GPUs launched in 2026: the TPU 8i on TPU 8 and the TPU 8t on TPU 8.
Memory is 288 GB against 216 GB, fed at 8.6 TB/s versus 6.5 TB/s. More memory per GPU means larger models fit before you have to shard across cards, which often matters more than raw TFLOPS for inference.
For LLM training and inference, weight the FP8 and FP16 rows and memory capacity most heavily. For scientific and HPC workloads, the FP64 row is the one to read.
Frequently asked questions
Which has more memory, the TPU 8i or the TPU 8t?
The TPU 8i carries 288 GB of memory versus 216 GB for the TPU 8t. Memory bandwidth is 8.6 TB/s for the TPU 8i and 6.5 TB/s for the TPU 8t.
Get Comparison Updates
New GPUs added weekly. Be the first to see how they compare.