A single quantum chip with 60 or 70 qubits could outperform a traditional supercomputer that occupies half a football field.

We made these chips programmable via TensorFlow, allowing models to run on CPUs, GPUs, or TPUs with minimal modifications.

We designed a system for both training and inference, not just a chip for training, to handle large problems effectively.

The future data center will be interconnected, allowing for a coherent cluster of high-performance computing resources, which will be the largest of its kind once completed.

The TPU program was initiated to meet the demand for high compute power for inference and training in AI applications.

The exponential improvement of chips allowed us to envision computing as essentially free, enabling widespread use.

Nvidia's key advantage lies in its integrated solution, which includes networking with InfiniBand, CUDA software, and a developer-friendly experience, allowing AI researchers to effectively program GPU clusters as a unified system.

The future of computing will involve running incredibly large neural networks with high performance and low power consumption, requiring different solutions for training and inference workloads.

At Rigetti Computing, we're driving the transition from microchips to systems based on individual atoms or artificial atoms that we build in the lab.