128 TOPS, sparse acceleration.
High-efficiency NPU with 128 TOPS for edge and cloud AI inference.
TRNG, PQC, AES, SHA.