AI Scale

Cedana has published a reference guide converting AI workload sizes from tokens into the GPU capacity needed to serve them. The piece gives engineers a rule of thumb for estimating how many GPUs are required to process 1 million, 1 billion, or 1 trillion tokens, sparking discussion among developers about AI infrastructure planning and computing costs.

A new report examining how artificial intelligence ventures in China are being financed is drawing attention. The analysis looks at the scale and sources of funding flowing into Chinese AI companies, a topic of growing interest as competition with the United States intensifies and questions mount over how Chinese AI firms are capitalised.

A new index from OneStream finds finance departments are rolling out artificial intelligence more quickly than they can validate or trust the outputs. The report highlights a growing gap between AI adoption and confidence in AI-driven results across finance functions. It adds to ongoing debate about governance and reliability as companies scale AI in decision-making.

Ukraine's former defense minister has unveiled a new interceptor drone designed to counter Russia's fastest attack drones, and says he wants artificial intelligence to help fly it. The project reflects Ukraine's push to scale up air defenses against Russian drone strikes, with AI seen as a way to let interceptors engage targets faster than human operators can react.