Inside a Capital One data center in Virginia, a single PostgreSQL cluster reconciles roughly 100 million card transactions every business day, with the ledger settling before East Coast traders pour their second coffee. American finance now runs on databases the way it once ran on paper, and the systems behind that shift are no longer back-office plumbing. They are the product.
How US banks moved from mainframes to managed engines
For decades, the largest US banks ran on DB2 and IMS on IBM mainframes, with COBOL stitching everything together. That stack still exists at JPMorgan, Wells Fargo, and Bank of America, but it now sits beside Oracle Exadata, Microsoft SQL Server, PostgreSQL, MongoDB, and cloud-native services from AWS, Google, and Snowflake. The Federal Reserve’s research on payments system modernization documents how core ledger workloads are being split into transactional engines and analytical warehouses, with FedNow settlement records flowing through both layers in near real time.
Capital One famously closed its last on-premises data center in 2020 and moved its workloads to AWS, with relational data running on Aurora PostgreSQL and event data on DynamoDB. Goldman Sachs runs much of its securities ledger on Snowflake and BigQuery for analytics, with kdb+ powering its time-series price history. Stripe, Plaid, and Block sit on a mix of MongoDB, MySQL, and CockroachDB. The point is that no single database wins. American finance picks engines for specific workloads, then ties them together with change-data-capture pipelines that synchronize records in seconds rather than overnight. The point of this portfolio is resilience as much as performance, since regulators now expect each workload to keep running even when one engine fails.
That portfolio approach is now standard across the top 25 US banks. A typical institution runs a mainframe core for the system of record, a relational store for customer accounts, a document store for case management, a graph database for fraud rings, a time-series store for trading data, and a warehouse for reporting. Stitching them together is the new craft, and the engineering teams that do it well ship features faster than peers who still rely on weekend ETL windows.
Where these systems are actually used in US finance
Card networks are the heaviest day-to-day workload. Visa and Mastercard process more than 1.5 billion US authorizations on a peak shopping day, and each one touches a transactional database within milliseconds. Real-time payments add a second tier of pressure. The Federal Reserve’s FedNow service crossed 1 million instant payments per day in early 2026, and every participating bank has had to build or buy a 24/7 ledger engine that never closes for batch processing.
Risk and compliance run on a different shape of data. Banks store trade history, KYC records, and transaction monitoring events in column-oriented warehouses, then run pattern queries to spot money laundering and fraud. Deloitte’s financial services insights note that the average top-20 US bank now feeds more than 200 source systems into a central data platform, with daily volumes north of 50 terabytes. Wealth managers at Morgan Stanley and Charles Schwab use graph databases to map household relationships across accounts. Insurance carriers run actuarial models against Snowflake and Databricks. Even the smallest community banks rely on FIS, Fiserv, or Jack Henry cores that sit on Oracle or SQL Server under the hood.
For US fintechs, the database choice often defines the product. A neobank like Chime runs on cloud-native Postgres and Kafka. A buy-now-pay-later firm like Affirm depends on low-latency decision databases to approve a loan in under 300 milliseconds. The TechBullion state of US fintech 2025 coverage shows that funding rounds increasingly call out database architecture as a competitive moat.
Benefits the US market is already seeing
The first benefit is speed. Cloud database engines have cut typical query times from seconds to milliseconds and have collapsed weekly batch reports into live dashboards. Bank executives now expect to see liquidity positions and credit exposure refreshed throughout the trading day, not at end of day. Citi, for example, has cut its overnight risk reporting window by more than half since moving parts of its data platform to Snowflake.
The second benefit is cost flexibility. Pay-as-you-go pricing lets a smaller US bank run analytics that used to require an eight-figure Oracle license. Community banks now access fraud models hosted by vendors like Q2, Alloy, and Hawk AI without buying any database hardware. The third benefit is integration. Modern engines speak SQL, JSON, and streaming protocols, which makes it simpler to connect to Plaid, Stripe, Yodlee, and the open banking aggregators Akoya and MX. That matters because the CFPB’s Section 1033 rule on personal financial data rights takes effect for large banks in April 2026, and only banks with clean data layers can answer those data-portability requests at scale.
Risks regulators and CIOs are watching
Concentration risk is the loudest concern. If a single cloud provider has an outage, dozens of US banks could lose access to their core data at the same moment. The OCC and the Federal Reserve have both issued guidance asking large banks to document fallback plans for their cloud database providers. The 2023 outages at AWS and the 2024 CrowdStrike incident left scars across the industry, and boards now ask quarterly questions about geographic redundancy and recovery testing.
Data security is the second risk. Customer financial records are among the most sensitive data sets in the US economy, and a database breach typically costs an American firm more than $5 million on average, according to recent breach studies. The CISA guidance on cybersecurity for financial services calls out encryption at rest, key management, and least-privilege access as the most common gaps. Insider threats and misconfigured permissions still account for a large share of incidents at US banks and brokerages.
Talent is the third risk and the quietest one. The Bureau of Labor Statistics projects strong growth for database administrators and data engineers, but US banks compete with hyperscalers and AI labs for the same people. A bank that cannot hire enough Postgres or Snowflake engineers will struggle to keep its ledger systems healthy, no matter what software it licenses. TechBullion’s coverage of cloud finance modernization tracks how regional banks are using managed services to bridge the gap.
Long-term opportunities for US finance
The next decade points toward three openings. Real-time everything is the first. FedNow, RTP from The Clearing House, and continuous settlement for securities push every bank toward 24/7 database operations and away from batch windows. Firms that finish that migration early will sell instant-payment products that slower competitors cannot match, and they will price liquidity products against live balances rather than yesterday’s snapshot.
Open data is the second opening. With CFPB Section 1033 live for the largest banks in 2026, customer-permissioned data sharing becomes a regulatory baseline, not an experiment. Banks that store data in clean, well-modeled tables will export it to Akoya, MX, and Plaid with minimal friction. Banks that still rely on screen scraping and undocumented mainframe extracts will struggle, and they will pay vendors for the privilege of compliance.
The third opening is AI inside the database itself. Snowflake Cortex, Databricks Mosaic, and Oracle’s vector features bring large-language-model queries to financial data without exporting it. A US bank can ask, in plain English, why a fraud model rejected a transaction, and get an answer grounded in the same warehouse where the data lives. That puts pressure on standalone analytics vendors and reshapes how teams at JPMorgan, Capital One, and PNC build their next decision systems. For more context on related software trends, see TechBullion’s tracking of AI in financial services. The signal for 2026 and beyond is direct. The database is no longer a back-end choice. It is the seam where compliance, customer experience, and AI all meet, and the US banks that treat it that way will set the pace for the rest of the industry through the rest of the decade.
