Definition
A Tensor Processing Unit (TPU) is an application-specific integrated circuit (ASIC) developed by Google to accelerate machine-learning workloads. Rather than being a general-purpose processor, a TPU is purpose-built for the dense matrix multiplications that dominate neural-network training and inference, trading flexibility for throughput and power efficiency on that narrow task.
Systolic array architecture
The heart of a TPU is a systolic array: a grid of multiply-accumulate (MAC) units wired directly to their neighbours so data flows through the array in lock-step, like blood pumping through a heart. Google's first-generation TPU used a 256 x 256 array of 65,536 ALUs, letting it perform tens of trillions of multiply-add operations per second without constantly shuttling intermediate values back to memory. This dataflow design is what gives the TPU its efficiency advantage over a CPU for matrix-heavy code.
Why it matters for sovereignty
TPUs live mostly inside Google's cloud, which makes them the opposite of self-hosted compute: powerful, but rented and centrally controlled. For builders who want to keep AI workloads under their own roof, understanding the TPU model helps clarify the trade-offs of on-device alternatives. The same systolic-array idea reappears, scaled down, inside the NPU on a laptop and, scaled up, inside data-centre GPUs.
For the convergence of hash-intensive and inference-intensive compute under one roof, see how D-Central frames the distributed compute model and the broader move toward sovereign AI.
In Simple Terms
A Tensor Processing Unit (TPU) is an application-specific integrated circuit (ASIC) developed by Google to accelerate machine-learning workloads. Rather than being a general-purpose processor, a…