What technology stack does JPMorgan Chase run its AI on?
JPMorgan Chase runs AI on a hybrid estate it is moving to the cloud. Global CIO Lori Beer oversees more than 65,000 technology employees, over 7,000 applications and nearly an exabyte of data on a $19.8 billion budget, per American Banker in October 2026. Dimon's April 2024 letter said about 70% of data runs in public or private cloud, with multiple providers including AWS and Google Cloud to avoid lock-in and $2 billion going into four private-cloud data centers in 2024; AWS said the bank ran nearly 1,000 applications on its cloud by end-2024, including chase.com. The model layer is LLM Suite, a proprietary abstraction layer through which third-party models from OpenAI, Anthropic and others are swapped in, with more than 200 applications built on it. Software is increasingly written by agents: Beer says 'most code is now assisted or written with agentic coding capabilities', coding assistants cut the coding portion of development 10% to 20%, Chase card refactors mainframe code with generative AI in a sandbox, and an AI Threat Modeling Co-Pilot made threat modeling 20% more efficient. The 2026 Innovation Week rule: match the right model to the right task on security, resiliency and cost.[1][2][3][4][5][6][7]
Data as of Oct 16, 2026 · JPMorgan profile updated Oct 13, 2026 · 22 sources cited on this page · 140 banks profiled on this site · Markdown
What are the layers of JPMorgan Chase's AI and technology stack?
As executives, the bank's blog and partners describe them. GPU vendors and data-center locations are not disclosed.
| Layer | What is on the record | Source | Date |
|---|---|---|---|
| Budget and estate | $19.8 billion technology budget; more than 65,000 technology employees; over 7,000 applications; nearly an exabyte of data; about $12 trillion of digital payments a day | American Banker | October 1, 2026[1] |
| Cloud | About 70% of data in public or private cloud; multiple providers including AWS and Google Cloud to avoid lock-in; $2 billion for four private-cloud data centers in 2024 | Dimon's April 2024 letter via Bank Automation News | February 14, 2025[2] |
| AWS footprint | 100 applications in 2022 to nearly 1,000 by end-2024; chase.com and core banking, deposits and payments applications migrated; UK consumer bank built on AWS from 2021 | AWS via Bank Automation News | February 14, 2025[2] |
| Model layer | LLM Suite: proprietary abstraction layer through which third-party models (OpenAI, Anthropic and others) are swapped in and out | American Banker | May 22, 2025[3] |
| Applications on the model layer | 'We've built over 200 applications on top of it' | Lori Beer | October 1, 2026[1] |
| Model choice | Match the right model to the right task, balancing security, resiliency and cost-effectiveness | Beer at Innovation Week | June 15, 2026[7] |
| Software delivery | 'Most code is now assisted or written with agentic coding capabilities. We're rebuilding the factory of how we create software'; coding portion of the lifecycle down 10% to 20% | Lori Beer | October 1, 2026[1][4] |
| Mainframe modernization | Chase card explains legacy code with generative AI, generates modern-language code in a sandbox and accepts it only if the same tests pass | Roman Eisenberg via CIO Dive | October 17, 2025[5] |
| Security engineering | AI Threat Modeling Co-Pilot: 20% efficiency in threat modeling and an average of nine additional novel threats found per model | JPMorganChase | October 3, 2025[6] |
| Guardrail testing | Fence framework tests LLM use cases for hallucination, topic drift and prompt injection with synthetic data | JPMorganChase | April 2, 2026[8] |
| Quantum-AI research compute | Dedicated environment at OQC's London Quantum-AI Data Centre with AMD AI and classical compute | JPMorganChase | June 3, 2026[9] |
Does JPMorgan Chase build or buy its AI technology?
It builds the platform and rents the models. LLM Suite is the bank's own layer, with models from OpenAI, Anthropic and others swapped in; the Hong Kong change in June 2026, when Claude was removed from the approved list for staff there, shows the switch is used for jurisdictional reasons too. Beer says many of the bank's current agents are built by her own engineering teams.[3][10][1]
On infrastructure it is multicloud by design: Dimon's 2024 letter described several cloud partners to avoid 'lock-in' with one vendor, alongside private-cloud data centers. Beer expects AI costs to come down over the next year, 'which allows this to be democratized, like most technologies'; Barnum said in October 2026 that token costs are part of why AI gains are not yet visible.[2][1][11]
What does the JPMorgan profile record on this topic?
The rows below are the entries of the JPMorgan AI strategy profile that bear on this page, as of Oct 13, 2026; the profile holds the full record of 40 dated moves, 13 use cases and 67 sources.
| Use case | What the record shows | Status | Source |
|---|---|---|---|
| LLM Suite and agentic workflows | Firm-wide gen-AI platform; agents in production for software testing, private-banking research screening and payments; long-running agents planned for 2026. | In production | [12][13][14] |
| Coding assistants and AI-assisted mainframe refactoring | A bank-built coding assistant lifted engineer productivity 10% to 20%; Chase card documents and rewrites mainframe code with generative AI inside a sandbox, accepting output only when the original tests pass. | In production | [4][5] |
| Date | Move | Detail | Source |
|---|---|---|---|
| Jul 2, 2026 | AI chief Heitsenrether to retire | The chief data and analytics officer, who led the LLM Suite build-out, leaves at year-end; her title and portfolio go to CTO Scot Baldry. | [15] |
| Jun 5, 2026 | Lockbox robots and an LLM assistant | J.P. Morgan Payments says mail-handling robots built with Ripcord handle about 4,000 envelope and document types in a process that covered some 480 million checks and documents a year, now largely automated at over 99.9% accuracy, and staff use an LLM assistant for real-time status. | [16] |
| Apr 2, 2026 | Fence guardrail framework | A data-driven framework that tests LLM use cases for hallucination, topic drift and prompt injection using synthetic data. | [8] |
| Feb 24, 2026 | Company update: 150,000 weekly LLM Suite users, use cases doubled | Dimon says 150,000 people use the platform every week and think they save about four hours a week; Barnum says AI use cases in production doubled and about $600 million of identified efficiencies are partly AI-related; Mary Erdoes says one controls-review AI tool spread from 200 to 3,000 staff. | [17][18] |
| Feb 23, 2026 | Company update deck: AI solutions in production doubled in 2025 | The 8-K presentation guides 2026 technology expense to about $19.8 billion, up 10%, and says revenue is the largest contributor to AI benefits, with generative AI the fastest-growing segment. | [19] |
| Oct 17, 2025 | Chase card uses generative AI to refactor mainframe code | Roman Eisenberg, head of technology for Chase card and connected commerce, describes explaining legacy code with generative AI, generating modern-language code in a sandbox and accepting it only if the same tests pass. | [5] |
| Jun 3, 2025 | LLM Suite named American Banker Innovation of the Year | The bank's post says the platform went from zero to 200,000 onboarded users in eight months after its summer 2024 release; Derek Waldron calls it the firm's planned 'AI hub for employees'. | [20] |
| May 19, 2025 | Investor Day: 200,000 LLM Suite users, 175 AI use cases in CIB | The investor day transcript reports over 200,000 employees with LLM Suite access, more than 175 AI use cases in production in the commercial and investment bank, a 40% fall in KYC unit cost since 2022, and a plan to cut consumer operations headcount about 10% over five years while the business grows more than 25%. | [21] |
| Mar 13, 2025 | Beer: bank-built coding assistant lifts engineer output 10-20% | Global CIO Lori Beer tells Reuters tens of thousands of software engineers gain 10% to 20% from the bank's own coding assistant, that the bank has about 450 potential AI use cases and that Dimon expects 1,000 by 2026. | [4] |
| Aug 9, 2024 | LLM Suite reaches 60,000 employees | CNBC reports the opt-in rollout; by mid-2025 American Banker describes it as available to almost all employees. | [3] |
| Jan 1, 2023 | Employees get OpenAI models through an early LLM Suite | The groundwork for the firm-wide platform: a corporate ChatGPT for drafting emails and summarising documents. | [22] |
Which cloud providers does JPMorgan Chase use?
Several. Dimon's April 2024 letter named AWS and Google Cloud among multiple cloud partners, chosen to avoid lock-in, and said $2 billion was going into four private-cloud data centers. AWS said the bank had nearly 1,000 applications on its cloud by end-2024, including chase.com.
What is LLM Suite technically?
A proprietary abstraction layer and portal through which large language models from OpenAI, Anthropic and others are made available to employees under the bank's controls, with models swapped in and out. Global CIO Lori Beer said in October 2026 that more than 200 applications have been built on top of it.
How is JPMorgan using AI to write software?
Beer says most code is now assisted or written with agentic coding capabilities, and coding assistants have cut the coding portion of the development lifecycle by 10% to 20%. Chase card uses generative AI to refactor mainframe code in a sandbox, accepting output only when the original tests pass.
How big is JPMorgan Chase's technology operation?
American Banker reported in October 2026 that Lori Beer oversees more than 65,000 technology employees, over 7,000 applications and nearly an exabyte of data on a $19.8 billion budget, supporting about $12 trillion of digital payments a day.
- The Most Powerful Women in Banking, #5, Lori Beer, JPMorganChase
- JPMorgan's migration to AWS cloud jumps 10X in 2 years
- How JPMorganChase democratized employee access to gen AI
- JPMorgan engineers' efficiency jumps as much as 20% from using coding assistant
- Chase enlists generative AI to drive modernization
- Revolutionizing Threat Modeling with AI: The Threat Modeling Co-Pilot
- Innovation for business impact: Key takeaways from JPMorganChase's 11th Innovation Week
- Strengthening LLM guardrails with synthetic data generation
- OQC, JPMorganChase and AMD commence research collaboration to develop new quantum-AI platform
- JPMorgan Chase cuts off Anthropic access for its Hong Kong staff
- JPMorgan CEO Dimon warns customers of 'liability transfer' with AI
- JPMorgan Chase plans to deploy more powerful AI agents this year
- Translating technology from cost to competitive edge (Gill Haus)
- JPMorgan Payments' agentic commerce play
- JPMorgan AI chief Teresa Heitsenrether retiring in 2026
- JPMorgan Chase taps AI to process checks
- 2026 Company Update Transcript - Chairman and CEO Q&A
- 2026 Company Update Transcript (full event)
- 2026 Company Update presentation slides (Form 8-K, Exhibit 99.2)
- LLM Suite named 2025 'Innovation of the Year' by American Banker
- 2025 Investor Day Transcript - Full Transcript
- Here's JPMorgan Chase's blueprint to become the world's first fully AI-powered megabank
When JPMorgan moves on AI, you'll read it here first.
when one of the banks we cover moves on AI, it is in the next morning's brief · six sourced stories · 7 am ET · free
plus every tracker, bank and agent page update, the morning after · leave any morning