Google Cloud CEO Thomas Kurian said earlier this month at the Goldman Sachs Communacopia + Technology conference that the average AI server investment pays back in less than two years, and that its custom chips have a payback period only half that of GPU-based solutions, according to IThome.
Custom silicon advantage
Kurian claimed that Google Cloud's custom AI accelerator business is more than twice the size of the second-largest hyperscale data center operator. Most of Google Cloud's infrastructure contracts are long-term, spanning five years, he noted.
The full-stack control from chip to model allows Google to offer products like Gemini Flash 3.8 at half or even lower prices than competitors, Kurian said. He asserted that Google Cloud provides a 2.7x cost-performance advantage in AI training, 80% in AI inference, and 30% in CPU workloads.
Enterprise AI economics
Google's enterprise AI platform, Gemini Enterprise, runs machine learning models at only 40% of the cost of competitors while delivering higher accuracy, according to Kurian. He attributed the adoption by more than 90% of Fortune 100 companies to these factors.
In related news, Google plans to invest at least 13 billion euros in AI infrastructure in Finland, its largest single investment in Europe. The company is also reportedly in talks for a deal exceeding $1.5 billion to acquire talent and license technology from AI coding startup Mechanize.
