The news: Bank of America (BofA) struggled to deploy NVIDIA’s enterprise AI solution—a bundle of chips and software “designed to build, train and run large-scale AI systems”—per Business Insider.”
Zoom out: It’s no surprise that megabanks run into roadblocks while trying to implement novel AI applications. The central tensions are between banking solutions—which leave no room for missteps related to technology decisions—and AI solutions designed for unregulated or loosely regulated industries. Banks face especially high barriers to adopting new technology due to their scale, creaky technology stacks, and complex data security and compliance requirements. Their technology talent reflects those parameters—not breakneck innovation.
Trendspotting: Megabanks have spent billions of dollars on AI software, model training, supporting infrastructure, and developer and research talent. Spending spans beyond the adoption of new core AI hardware and software to modern infrastructure overall. JPMorgan said in October it spends $2 billion per year on AI. BofA said its 2025 spending on new technology initiatives would reach $4 billion. Royal Bank of Canada, Canada’s largest bank, spends over C$5 billion ($3.64 billion) annually on technology, including AI.
Meanwhile, “AI” has become a mandatory buzzword in bank executives’ public statements. It’s at least a budget category for others and has been at least superficially adopted in many cases. But reality is setting in as the cost of AI leadership in banking becomes more apparent. Capital One has been wary of the costs of its cloud contracts related to running AI workloads as it further adapts its infrastructure. Others are outsourcing: For example, BNY recently partnered with Google Cloud to integrate Gemini’s enterprise platform.
Implications for banks: AI adoption in banking is shifting toward purpose-built software that happens to use AI, often bought from vendors; off-the-shelf AI infrastructure from technology companies; enterprise-wide original research and development of AI models; and heavy investment in modern architecture. Banks that stick to the AI adoption narrative without a specific plan are presumably paying lip service to a technology that they don’t understand.