The Power of AI Meets the Problems of Scale The financial services industry is embracing AI. Nearly 80 percent of executives, developers, data scientists, engineers, and IT workers indicated that their companies are using AI-enabled applications, according to NVIDIA’s 2022 “State of AI in Financial Services” survey. Across capital markets and consumer finance, a vast majority of those firms reported that AI has already increased annual revenues or reduced costs, some by more than 20 percent. Business Insider estimates that the potential for AI-driven cost savings for banks alone will reach $447 billion by 2023. The ever-growing list of applications includes everything from fraud detection to portfolio optimization, and NVIDIA’s survey showed that firms’ deployment of AI in top use cases is skyrocketing. 1 But almost half of AI projects never make it to production , The AI Enigma and implementation is hindered by familiar challenges. > Lack of infrastructure: AI tooling > Trouble scaling and lagging can be difficult to deploy on legacy performance: Bare-metal infrastructure, and the rise of AI infrastructure makes 91% 23% apps is making IT management scaling AI more difficult, or increasingly difficult. data scientists waste weeks > Data silos and shadow IT: waiting for models to train Investment often happens on non-GPU systems. of financial services firms of executives trust that at a project or team level, > Cost control: Disparate IT are driving critical business their firm can move AI from outcomes with AI research to production on customized bare-metal and unused infrastructure infrastructure, creating pockets raise overhead and cost per of AI outside central IT control. workload. 1 Gartner “P-19019 AI in Organizations,” Claudia Ramos, Erick Brethenoux, 2020 Realizing the Potential of AI in Financial Services | 2