Which banks lead the AI investment race?

Our latest research on which banks are actively investing in AI start-ups, exclusive data on Chinese banks and innovation, a new AI chief at Morgan Stanley, and how to snag the best AI talent.  1,765 words, 6 minute read.

Investment and acquisition of AI start-ups often give banks unparalleled access to cutting-edge AI technologies, market intelligence and expertise they don't have in-house. However, new data from Evident's AI Ventures Dispatch, released today, shows a 37% drop in AI-focused investments by the 50 banks in the Index last year, mirroring a wider downturn in venture capital markets. In this changing environment, we see two notable developments:

European banks gaining ground: As you see, Europe notched the region’s highest share of AI investments on record. Data from the Dispatch shows European banks accounted for 45% of all AI-focused investments in 2023 (up from just 10% in 2015), reaching near-parity with the North American banks that traditionally dominated in this space. This trend validates Europe’s recent gains in the global AI race – underscored by BNP Paribas’ recent participation in Mistral AI's $415 million Series A round.

Who’s leading? Amidst the downturn, there has been a significant reshuffling of the leaderboard on AI ventures in the last year. Some of the traditional leaders such as Goldman Sachs and First Citizens scaled back... Others kept pushing, notably Citigroup jumped from #3 to #1.

What they’re doing? Approaches vary. Some invest far and wide, others appear to have taken a more strategic approach, focusing on a smaller number of investments that might be rolled out inside the bank (Capital One leads the way there).

Why it matters? AI ventures are just one part of a bank’s innovation strategy. But banks that aren’t playing the game are possibly missing out on an opportunity to shape products to best suit their needs, and to reap the financial rewards while doing so.

The Dispatch is a monthly report exclusively prepared for Evident members. Get in touch to find out more about membership, or read an excerpt of The Dispatch: AI Ventures here.

There was lots of noise and some substance produced at Nvidia’s GTC developer conference. I was at “AI Woodstock” in San Jose this week, and here’s how I’d separate them out.

First the noise: quite deafening. The latest tech CEO superstar, walked onto a dimmed stage and, with a wink, seemed to compare himself to another superstar who played the same stage, Taylor Swift. Jensen Huang’s keynote notched up Swiftian streaming numbers, with 12 million viewers and counting.

Huang promised his headline-making Blackwell B200 chip will drive “the new industrial revolution.” Two-and-a-half times more powerful than the previous chip, Blackwell will certainly power the training of ever larger AI models. Impressive at first; seen another way, Nvidia in fact just physically doubled the size of its chip.

So what’s the impact on banks?

Initially, limited.

Most don’t train or host their own large language models. A faster chip will reduce latency – the delay between asking a model a question and getting an answer – and that’ll trickle down as OpenAI, Anthropic and others adopt the new chips, reducing costs and improving model quality.

Something else caught my attention: Nvidia’s new “AI Foundry.”

With their new “Foundry,” Nvidia would directly help companies build customized models, without the prohibitive costs of doing so from scratch. They call it NIMs – Nvidia Inference Microservices.

The B200 can also keep sensitive data safer during training. That matters: Business will be able to train AI models on Nvidia’s servers, letting banks create use cases there that require much tighter security. Think AI for threat assessments, or Blockchain key management.

The day after Jensen spoke, I hosted a fireside chat with Chintan Mehta (CIO and Head of Digital Technology at Wells Fargo). He suggested that Blackwell could rewrite Moore’s Law. "It means more longevity for infrastructure design, which means more runway for software-driven innovation," he said. In other words, more computational power will make it easier to build the cool next-gen services.

I didn’t come away thinking this was AI’s or even Nvidia’s 2007 iPhone moment. Hard to say without hindsight. But what’s clear: The new chip represents another step forwards for generative AI, one of many in the last year. And Nvidia’s new “AI Foundry” may well hasten adoption by helping businesses build AI models to spec.

– Alexandra

China and the U.S. are in an AI Cold War, and this one may now have a kind of Sputnik moment. China passed America in 2022 in terms of the number of AI patents issued for the first time, according to a report from the U.S. National Science Foundation – “a wake-up call,” said one observer.

How’s that playing out in banking? Pretty dramatically, it turns out, when we ran the numbers. In the industry that has been at the forefront of AI adoption, China is charging ahead, according to our exclusive data.

The Industrial and Commercial Bank of China (ICBC), the world’s largest bank by total assets, filed 1,067 patents with the Chinese patent office related to AI – 52 short of Evident AI Index patents leader Capital One. At this rate, ICBC will be tops by the end of this year. And while Bank of America and Capital One are heads above most other U.S. banks – as noted in our recent AI Patents Dispatch – all of the big four Chinese banks have made a concerted effort to innovate and protect their IP, filing an astonishing 1,781 patents between them in 2021.

Why this matters? For the past five years, the Chinese state made AI innovation a top national priority and its banks are in the vanguard. China’s banks aren’t in direct competition with most U.S. or European institutions. As with e-commerce and social media, China is building its own parallel AI ecosystem. Yet the rapid innovation by Chinese banks could mean they’re building strong capabilities and could well expand more ambitiously beyond China.

Read more about how banks are approaching AI patents here.

Aravind Srinivas can’t seem to buy a talent break. Perplexity AI’s CEO last week told Business Insider that a researcher at Meta didn’t come over because his start-up lacked Meta’s stockpile of Nvidia’s highly coveted GPUs. Earlier this month, Perplexity lost another poaching target when Google quadrupled their salary.

These stories reminded us of the checklist a bank research chief shared with us recently. When an AI researcher considers an offer, they ask three questions:

• How many GPUs do you have?
• Who else is working there?
• Can I publish my research?

A “no” on any of them means you’re unlikely to win top talent.

NEW AI HEAD AT MORGAN STANLEY

Morgan Stanley named Jeff McMillan its new firmwide Head of Artificial Intelligence. Before his promotion, McMillan pioneered the bank’s approach to Generative AI, working with OpenAI to develop an LLM-powered financial advisor for its’ wealth management team that went into production in September of last year. This new role puts McMillan at the center of the bank’s AI efforts, working across teams and co-chairing its AI Steering Group.

We asked him how the partnership with OpenAI has evolved over the last year:

“In the beginning we really didn’t understand the complexity of deploying LLMs in a large financial institution. But we figured it out together. What started out as a more consultative business endeavour has really evolved into a tight relationship built on a shared sense of excitement and responsibility for being among the first to deploy this technology.”

• Société Générale hired Christophe Tummers, previously CDO at UBS, as Group Chief Data Officer. He reports to the relatively new COO, Laura Mather, who was previously CIO and CTO at Credit Suisse. The bank has set a goal of generating €500 million in value from Data and AI initiatives by 2026.
  
  
• BNY Mellon announced Roman Regelman, Senior Executive Vice President and Global Head of Securities Services and Digital, will leave the bank in April. The move marks a recent reshuffle at the top of BNY Mellon, which recently hired Gary Delaney as Chief Information Security Officer and Julie Gerdeman as Global Head of Data & Analytics.

Join us for the first virtual roundtable of the year to discuss the latest findings and data from Evident’s April 2024 Innovation Report with AI leaders at two of the world’s largest banks. The roundtable is open to all.