Google AI Platform Dominates Production Use

A recent survey indicates Google's AI platforms, encompassing Gemini and Google Cloud, are the preferred choice for 48% of organizations deploying AI in production, more than double that of OpenAI. This dominance stems from Google's integrated offerings in both models and cloud services. The study also highlights a significant correlation between rigorous tracking of AI infrastructure spending and a better perceived return on investment, with decision-makers being nearly twice as likely to report such tracking.

Further insights reveal that 80% of production AI users report average accelerator utilization at 50% or below, pointing to potential underutilization of resources. Integration with existing data stacks, rather than cost per million tokens, is a more significant factor for 40% of respondents when choosing AI infrastructure. Emerging trends show a strong interest in specialized AI clouds, with 35% likely to evaluate them in the next 12 months.

Uptime and developer productivity are the leading metrics for judging AI infrastructure performance, surpassing AI-specific measures. Dell leads in addressing inference memory concerns, named by 38% of respondents. The survey, conducted in August with 109 respondents from organizations with 100+ employees, also noted a decrease in expectations for cutting back on on-premises compute compared to July.