Cloud Adoption Update for Financial Market Infrastructure Providers 2H25

Overview

In the second half of 2025, Financial Market Infrastructure Providers (FMIs) continued to accelerate the migration of critical workloads across the trade lifecycle to AWS. Exchanges, Central Counterparty, Clearing Houses (CCPs), and Central Security Depositories (CSDs) have been leveraging cloud services to drive efficiency and unlock innovation. Generative AI (GenAI) and agentic AI played an increasingly critical role as engines of innovation. We expect this trend of rapid growth in GenAI adoption to accelerate in 2026, driven by the growing breadth of use cases in production, coupled with AWS’s rapid pace of innovation (key announcements from re:Invent 2025 here).

Trading and Post Trade

FMIs continue to move mission critical trade and post-trade systems to AWS. At re:Invent 2025, London Stock Exchange Group (LSEG) discussed how they will use AWS bmsfe-2 Outposts to power the FX trading infrastructure for their PriceStream venue. PriceStream links over 100 liquidity providers and over 1,200 liquidity taker clients with fair and equal access for over 150 currency pairs. With Outposts, PriceStream’s matching engine runs more than 2x faster across all latency thresholds compared to on-premises, with a 6x latency improvement at median latency point.

Also at re:Invent 2025, Nasdaq shared how they engineered multiple layers of resiliency and reliability into the operation of their mission critical matching engines on AWS Outposts. Nasdaq discussed its approach to disaster recovery for trading systems by utilizing a multi-site, active/active deployment and by highlighting failure domains, static stability, and disconnected operations.

The Japan Exchange Group (JPX) announced that they are conducting a Proof of Concept (POC) for its matching engine on AWS. JPX will test the performance of arrowhead, its cash equity trading system, jointly developed by the Tokyo Stock Exchange and Fujitsu, on AWS bmsfe-2 Outposts.

LSEG shared how they migrated the core clearing and settlement workloads for London Clearing House (LCH) to AWS to enhance resiliency. As part of the migration, LCH created a deployment model that enabled it to meet the Bank of England’s resiliency mandate, which required LCH to sustain operations independently for 5 days in the event of a complete cloud services provider outage. To meet this requirement and optimize resiliency, LCH deployed a multi-region strategy across two AWS regions and built a tertiary environment on AWS Outposts. LSEG also discussed how LCH leveraged Amazon’s Experience Based Accelerators (EBAs) to migrate over 12 applications to AWS over the last 3 years, including the critical and highly-regulated Collateral Management System (CMS). Since 2022, LSEG has conducted 11 EBAs, involved 13 cross-functional teams, identified 5 repeatable migration plans and 26 process improvements, while removing over 50 critical blockers.

Boerse Stuttgart modernized its data platform by moving three on-premises Oracle databases to Amazon RDS for Oracle. The data platform ingests up to 10 billion quotes a day and contains over 63 terabytes of mission critical data, which Boerse Stuttgart leverages to deliver accurate, real-time information to customers and ensure transparency across the trading process. Boerse Stuttgart worked with AWS and App Associates, an AWS Partner, to migrate its workloads to AWS over a single weekend. The team now saves two days a month on maintenance, conducts failover testing in minutes rather than days, and has increased platform performance by 15 percent.

The Options Clearing Corporation (OCC), one of the eight Systemically Important Market Utilities (SIFMUs), wrote a blog detailing how they architected an identity governance framework using AWS IAM Identity Center. OCC’s approach reduced operational complexity while strengthening security controls.

Data Distribution and Analytics

FMIs and market data providers continue to distribute market data to customers on AWS, while also leveraging AWS to enhance their offerings and prepare for a world with more GenAI workflows. In partnership with RoZetta, SIX Group migrated its historical tick data platform, Ticks by SIX, to AWS. RoZetta ingested and normalized hundreds of terabytes of tick data, including 10 years of history and more than 30 million instruments. Working alongside AWS, RoZetta developed the Ticks by SIX pilot in 3 months, including data transformation and enhancement pipelines, data governance controls, and a secure, high-performance framework. The new platform ingests 10 terabytes of data a month and enables SIX to deliver customized, large datasets to its customers within hours, a process that previously took days. It also enables customers to receive data within 2 hours of market close, as well as a set of pre-calculated analytics such as time bars.

BMLL published an AWS case study detailing how it has built and run its market data platform on AWS since 2014. BMLL uses Amazon S3 to store over 20 petabytes of data – “one of the world’s largest harmonized datasets of historical trading information.” BMLL saves $3.5 million a year in storage costs by using S3 Intelligent-Tiering and S3 Glacier.

LSEG published a blog detailing how the firm uses Amazon S3 to host its Tick History Data and Tick History PCAP datasets. With AWS, LSEG saved up to 80% on data storage costs and can dynamically scale to handle microbursts of message traffic seamlessly. To streamline data access and analytics, LSEG’s S3 Direct customers can query this data directly or download it on-demand over a congestion-free global network. This query-in-place capability eliminates the need to transfer massive datasets to separate analytical systems, reducing complexity and enabling faster time-to-insight on petabyte-scale tick data repositories.

At re:Invent 2025, LSEG discussed their data monetization and distribution journey to AWS, with a focus on how they built the Matrix Signals Platform (MSP). MSP is a market intelligence and analytics platform that sits on top of over 75 petabytes of historical PCAP data and consumes 5-10 terabytes of real-time market data a day from 575 venues. MSP enables LSEG to do real-time anomaly detection on 90 million data streams, 274+ billion messages a day, and trillions of historical tick messages. By leveraging AWS’s analytics tools and GenAI services, LSEG can dynamically provision capacity to meet hot spots of trading activity globally, correlate spikes in message activity with news flow, and track the end-to-end data path for each message across their network. LSEG reduced the cost of storage and compute by 97% and 82%, respectively, and reduced the time to signal from 3 days to a window of 30-45 minutes.

LSEG announced it is collaborating with AWS to transform its market-leading “Real-Time Full Tick” and “Real-Time Optimized” data processing capabilities. The firm will leverage AWS’s services in the collection, routing, and distribution of its real-time financial data. This collaboration will use LSEG’s private cloud and will enable financial institutions to access LSEG’s critical market data with greater flexibility, speed, and resilience.

S&P Global, FactSet, and IDC were launch partners of Amazon Quick Suite’s Quick Research function. Amazon Quick Suite is AWS’s AI-powered workspace helping organizations derive answers from their enterprise data and move swiftly from insights to action. Quick Research enables business professionals to tackle complex business problems by completing weeks of data discovery, analysis, and insights tasks in minutes. Customers now have access to S&P Global’s, FactSet’s, and IDC’s specialized datasets for their research needs.

GenAI & Agentic AI

FMIs are increasingly deploying GenAI across their workstreams to enhance efficiency and launch new products and services. JPX used Amazon Bedrock to enhance the usability and searchability of timely disclosure documents. Every year, companies listed on the JPX file over 140,000 timely disclosure documents, totaling over 1.1 million pages. These documents cover a range of topics and are often difficult to parse and extract key information from rapidly. JPX used Amazon Bedrock to generate new searchable tags for data contained within each filing, which has improved the searchability of the documents. JPX also linked the generated tags to other systems, such as the JPxData Portal. This enabled them to eliminate the challenge of data linkage and data duplication, while also driving cost savings.

JPX also leveraged GenAI to enhance the efficiency and output of its SCRIPTS Asia division. SCRIPTS Asia transcribes audio from earnings briefings and Investor Relations (IR) events of listed companies, creates a database of event details including speaker information, and provides this data to institutional investors, financial institutions, and information vendors. Translating this information into English has been a perennial challenge because of the enormous volume of translation work, the significant cost burden, and the high-precision translations required for the global financial industry. JPX used Amazon Bedrock to streamline the translation into English. As a result, SCRIPTS Asia is now able to produce high-quality translations 10x as fast and about 35x cheaper compared to manual creation. This enabled SCRIPTS Asia to translate more events to English that it previously couldn’t because of time and cost issues.

LSEG built its AI Content Creation (AICC) platform on AWS to enhance its Worldcheck solution. Worldcheck is a global database containing data on high-risk individuals and organizations. It helps organizations and governments derisk their financial actions and contains millions of profiles (PEPs, sanctions, criminals, adverse media) with data from tens of thousands of data sources in over 60 languages, which are curated by over 200 analysts. LSEG worked with AWS to optimize content curation by leveraging LLMs to automate low-level data extraction and summarization. With GenAI, LSEG was able to process updates to the Worldcheck database in minutes instead of hours. This enabled analysts to spend more time on strategic analysis instead of routine tasks, improved the team’s content capacity eith no additional headcount, and led to earlier detection of issues, increasing client trust.

LSEG used Amazon Bedrock to build a GenAI-powered system that significantly improved the efficiency and accuracy of market surveillance operations. As part of the project, LSEG created a prototype that automatically predicts the probability of news articles being price-sensitive. During a six-week evaluation period, the system showed 100% accuracy in identifying non-sensitive news, and 100% recall in detecting price-sensitive content. The system’s ability to process news articles instantly and provide detailed justifications helps analysts focus their attention on the most critical cases while maintaining comprehensive market oversight.

At re:Invent, Coinbase discussed how it leveraged GenAI to enhance customer support operations, compliance processes, and developer productivity. Coinbase worked with AWS to build a GenAI-powered assistant that handles more than 65% of customer queries. The tool has saved Coinbase over 5 million employee hours every year and has resulted in customer interactions being solved in a single session (<10 minutes to resolution). To meet Coinbase’s high bar for quality of service, every customer response generated by the conversational chatbot is assessed for relevancy, accuracy, potential bias, hallucinations, and more. Coinbase also built internal GenAI-powered tools that provide human support agents with real-time insights to help answer complex queries. These tools led to a 10% improvement in customer content resolution, as well as an improvement in resolution times and CSAT scores. Coinbase also uses GenAI to extract insights from all customer support tickets to continuously improve products and services.

To streamline compliance investigations for KYC (Know Your Customer), KYB (Know Your Business), and TMS (Transaction Monitoring Systems), Coinbase built customized ML models to predict high-risk situations, and are using GenAI to automate and accelerate complex investigations. Powered by Coinbase’s Compliance Auto-Resolution Engine (CAR), intelligent agents extract information from a wide variety of sources, and then compose a narrative summary and investigation report, which enables human compliance agents to quickly and confidently make well-informed decisions.

Coinbase also discussed how they are accelerating their software development lifecycle (SDLC) with GenAI. Currently, GenAI writes more than 40% of the code at Coinbase, resulting in over 75,000 employee hours saved. The exchange has seen 3x increase in bugs found, and an 86% reduction in costs for AI-based Quality Assurance testing.

The Options Clearing Corporation (OCC) is also using GenAI to streamline software development processes and has reduced time spent documenting software tests by as much as 80%. The OCC also optimized the process of checking and updating code, helping developers to deploy solutions more efficiently. The firm has also simplified the creation of developer release notes, technical summaries, and procedure documents, which has enabled teams to organize and share information more quickly.

FINRA shared it is using GenAI to enhance its software development lifecycle (SDLC) and enhance code generation. Currently, over 1000 active users are leveraging FINRA’s GenAI-powered Integrated Development Environment (IDE), leading to a 30% improvement in overall code quality, with 80% positive engineer sentiment.

Crypto.com detailed how it used user and system feedback to improve and optimize instruction prompts continuously. By using a feedback-driven approach, crypto.com was able to create more effective prompts that adapt to various subsystems while maintaining high performance across different use cases. As a result, crypto.com saw a 34% improvement in task effectiveness, transforming a basic prompt with 60% accuracy into a robust classification system with 94% accuracy on challenging cases.