Google Cloud AI Servers Pay Back in Under Two Years, Custom Chips Beat GPUs on Returns

Google Cloud CEO Thomas Kurian said at the Goldman Sachs Communacopia + Technology Conference that Google Cloud's average AI server investment pays back in under two years, and its custom chips pay back in only half the time of GPU-based setups. Its custom AI accelerator business is more than twice the size of the second-largest hyperscaler.

Kurian said that controlling everything from chips to models lets Google offer products like Gemini Flash 3.8 at half the price or less compared to rivals. Google Cloud claims 2.7x, 80%, and 30% better price-performance in AI training, AI inference, and CPU, respectively.

Google's enterprise AI platform Gemini Enterprise runs machine learning models at just 40% of rivals' cost while delivering higher accuracy. That's one reason over 90% of Fortune 100 companies use the platform.