The AI Landscape for Financial Services: Who Offers it and How?

    Insights » Global » The AI Landscape for Financial Services: Who Offers it and How?

    A general overview of the AI landscape in financial services, from infrastructure providers to the leading GenAI platforms.

    AI landscape for financial services: Who offers it and how?
    Fabiane Ziolla Menezes

    Fabiane Ziolla Menezes

    Content Strategist

    Contact the author

    The leading players of the upcoming artificial intelligence revolution in the financial industry are big tech companies, from infrastructure providers – cloud service and CPU chips developers – to GenAI applications. As we showed, financial institutions are partnering with them to experiment with the new automation and customization opportunities brought about by the latest AI push. This second article in our AI series provides an overview of the organizations shaping the sector and a feel for the competitive landscape.

    Cloud and CPU providers

    The first layer in AI capabilities is general technology, which can be cloud-based or hardware-based (CPUs), or a combination of the two. Infrastructure service providers help financial companies with storage space, processing capabilities and architecture and connect them with AI-focused development platforms and players.

    According to the latest analysis1 of major cloud providers’ Q1 2024 earnings by Synergy Research Group, the largest players in the sector, considering private and public cloud, are Amazon’s AWS (31%), Microsoft’s Azure (25%), Google (11%), followed by Alibaba, Salesforce, IBM and Oracle – who combined have less than 5% of the worldwide market.

    The largest players in the sector, considering private and public cloud, are Amazon’s AWS (31%), Microsoft’s Azure (25%), Google (11%), followed by Alibaba, Salesforce, IBM and Oracle – who combined have less than 5% of the worldwide market.

    The sector reached USD$ 76.5 billion in revenue in the first three months of 2024, reaching a 12-month revenue of USD$ 283 billion. John Dinsdale, a chief analyst at Synergy, sees the sector’s annual growth rate at 21%, roughly half of what it grew before 2022.  “The market has become too massive to grow that rapidly, but we will see the market continue to expand substantially. We are forecasting that it will double in size over the next four years,” he wrote in the analysis.

    Along with storage and processing capacity, companies interested in acquiring AI capabilities must also power their hardware with CPUs capable of handling the massive data needed to develop and power AI models and applications. Nvidia is today the world’s most prominent central processing units (CPUs) manufacturer when it comes to AI chips2, followed by Intel, Qualcomm, AMD, and other chip designers.

    As competition between tech companies grows, the differences between hardware manufacturers and storage space providers are likely to narrow. The same tech giants that run the massive cloud services are also preparing to develop their own AI CPUs and, to some extent, compete with chipmakers, who, in turn, are seeking alternative paths to reach customers beyond partnerships with cloud providers.

    One example is Nvidia’s recently announced partnership with Cisco3, a networking service provider, to offer integrated AI hardware and software solutions in the data center. This partnership is especially appealing for financial companies as Cisco provides corporate clients with cloud and on-premises infrastructure and the bulk of financial institutions’ workload is still private.

    Platforms for AI foundation models’ development

    The next layer in AI technology is the foundation model—the actual programming that derives intelligence out of vast amounts of data. Foundation model is a term coined by Stanford University’s Human-Centered Artificial Intelligence. It refers to systems that provide a base on which other models can be built; they can be trained on broad data and fine-tuned to a wide range of downstream tasks4. Large language models (LLM) such as Google’s BERT and OpenAI’s GPT are examples of foundation models.

    As they can cost tens or hundreds of millions of dollars to train, the Big Tech companies are the ones developing them and offering them to other players – which only increases these giants’ power over an array of economic sectors, including payments and finance. They have platforms aimed at developers so that companies can access, test, and fine-tune these models to their needs.

    According to IBM, companies must look for models already trained for tasks similar to their needs, as these models will typically perform better with zero-shot prompts than models that were not fine-tuned in a way that fits the companies’ use case

    Microsoft does that through Azure AI (a platform that includes OpenAI’s GPT), Amazon through Amazon Bedrock, Google through Vertex.AI, IBM through IBM watsonx, and so on.

    This way, companies have three paths to follow: 1) to build their own foundation models from scratch; 2) simply adopt an existing one; or 3) leverage someone else’s model and build on top of it. Several studies have shown that most financial companies opt for the last two paths.

    In doing so, they must find a model that supports the task they have in mind, such as analyzing attributes like as licensing rules, pretraining data, size, and how the model was fine-tuned. According to IBM5, companies must look for models already trained for tasks similar to their needs, as these models will typically perform better with zero-shot prompts than models that were not fine-tuned in a way that fits the companies’ use case.

    Most Big Tech offer companies access to models trained for tasks such as information extraction, summarization, responding to instructions, answering questions, participating in a back-and-forth dialog chat, or even programming (summarizing, converting, generating, or processing code).

    Before entering this AI journey, though, financial players must assess if they fit some requirements: vast amounts of data to supply the models with training, a dedicated budget, and talent professionals to build or fine-tune the models.

    AI global landscape overview

    GenAI leading players

    Once foundation models are in place, the market needs applications that make the models user-friendly and accessible to accomplish various tasks. Some Big Tech companies have been developing their own models, such as Google’s recently launched Gemma open model6, but they have also invested in AI startups thriving in this field, building powerful models and applications.

    For example, while GPT is the foundation model developed by OpenAI (an artificial intelligence research lab, 49% owned by Microsoft), ChatGPT is the conversational application that made it so popular in the last couple of years.

    The multitude of models arising from these startups, as well as the money being put into them by investors of all types, is overwhelming. Below is an overview of some of them.

    GenAI Leading Startups Overview

    Who has invested in who:

    • Amazon: invested in Anthropic

    • Google: invested in InAI21labs, Anthropic and Runway

    • Microsoft: invested in Adept, Inflection and OpenAI

    • Nvidia: invested in Adept, AI21labs, Cohere, Inflection, Mistral and Runway

    • Salesforce: invested in Anthropic, Cohere, Mistral and Runway

    Sources: multiple news outlets.

    If in the first article of this series, we talked about how financial institutions need to adequate their technology stack to experiment and scale AI-driven solutions, in this second piece, it becomes evident that the variety of so-called foundation models aimed at supporting these solutions and the number of players behind them are already overwhelming.

    However, financial institutions and fintech companies do not necessarily need to move at the same speed as technology providers. In fact, it will be necessary to take a step back at several points in this journey to better analyze available partners and technologies. In fact, while Big Tech and startups are building general models and applications with the goal of covering all types of uses, client companies can move in the opposite direction, using these broad-based models to help train smaller, more focused models aimed at improving performance or solving specific pain points at a fraction of the effort and price being put on by the developers.

    Next Steps

    This new AI-driven era promises to take payments and financial services to a new level of automation and customization. Several use cases are emerging, and we plan to explore some of them in a future article. Meanwhile, we are helping some of our clients to get specific answers about AI and GenAI in banking and payments. We can help you with that, too! Contact us to learn how.


    Keep up to date with our e-commerce, payments and crypto insights:


    Sources

    1. Synergy Research Group (April 30, 2024). “Huge Cloud Market Sees a Strong Bounce in Growth Rate for the Second Consecutive Quarter↩︎
    2. NBC News (Feb 24, 2024). “Why everyone is suddenly talking about Nvidia, the nearly $3 trillion-dollar company fueling the AI revolution↩︎
    3. AI Business (Feb 6, 2024). “Cisco, Nvidia Offer Alternative to Cloud Giants’ AI Infrastructure↩︎
    4. Bommasani, R., et. al. (n. d.). “On the Opportunities and Risks of Foundation Models” Center for Research on Foundation Models (CRFM), Stanford Institute for Human-Centered Artificial Intelligence (HAI), Stanford University. ↩︎
    5. IBM, “Choosing a foundation model in watsonx.ai“. Last Updated: 2024-06-19 ↩︎
    6. Banks, J. and Warkentin, T. (Feb 21, 2024). “Gemma: Introducing new state-of-the-art open models” In Google Developers. ↩︎

    Fabiane Ziolla Menezes
    Fabiane Ziolla Menezes
    fabiane@paymentscmi.com

    Fabiane is a journalist with more than 15 years of experience reporting on business, finance, innovation, and cities in Brazil. At PCMI, she keeps an eye on revolutionary fintech movements and the businesses behind these big changes in Latin America.