What technology stack does PNC run its AI on?
PNC is building its AI on infrastructure it owns. The 2026 proxy says head of technology Deborah Guild advanced a national data-center strategy, with a new Eastern facility under construction and a Midwestern hosting site secured; on October 15, 2026 Demchak said GPU chips are already in PNC's data centers and their cost shows up as rising depreciation. Underneath sit a data lake with 'one source of truth', digitized policies and procedures, and HR systems that recognize agents. Models are mixed by job: in September 2026 CFO Rob Reilly described a harness that picks the cheapest adequate model, frontier models reached through hyperscalers, open-weight models hosted in-house and proprietary small language models; head of data and automation Ned Carroll has said he is 'very bullish on small language models'. On top, Microsoft Copilot and AI-enabled search run for employees, about 115 orchestrated AI agents built the July 2026 PNC Mobile app on a patented data-streaming microservices architecture, and the patented data migration factory moved FirstBank's customers in June 2026. Technology spend is about $3.5 billion a year.[1][2][3][4][5][6][7][8][9]
Data as of Oct 16, 2026 · PNC profile updated Oct 13, 2026 · 22 sources cited on this page · 140 banks profiled on this site · Markdown
What are the layers of PNC's AI and technology stack?
As PNC executives and filings describe them. PNC has not published an architecture diagram, a capital figure for the data-center build or the names of its hyperscalers.
| Layer | What PNC has said | Source | Date |
|---|---|---|---|
| Data centers | National data-center strategy: a new Eastern facility under construction and a new Midwestern hosting site secured | 2026 proxy (DEF 14A) | March 11, 2026[1] |
| GPU compute | 'We will have our own GPU compute'; buying Nvidia chips; by October 2026 GPU chips are in PNC data centers, adding depreciation | Demchak | June 9 and October 15, 2026[10][2] |
| Data foundation | Data lake and data configuration 'all in one place, one source of truth'; digitized policies, procedures and papers | Demchak, Q3 2026 call | October 15, 2026[2] |
| Identity and access for agents | HR systems that recognize agents as well as employees; a risk framework on access control, already built | Demchak, Q3 2026 call | October 15, 2026[2] |
| Model layer | A harness chooses the cheapest adequate model; frontier models through hyperscalers; open-weight models and proprietary small language models in PNC data centers | Reilly, Barclays conference | September 14, 2026[3] |
| Small language models | Carroll: 'very bullish on small language models' because the work is language interpretation; PNC's own vocabulary is context for fine-tuning; the test is 'price, efficacy and latency' | Ned Carroll | June 2, 2025[4] |
| Model selection | 'I don't need a model to answer advanced calculus when I want to understand a policy or procedure around a check return' | Ned Carroll | June 29, 2026[11] |
| Employee AI | Multiple generative-AI platforms including Microsoft Copilot; AI-enabled search for customer-facing teams | 2026 proxy | March 11, 2026[1] |
| Software delivery | Agentic development system; about 115 orchestrated agents built and tested the new mobile app with human review; Reilly called it '100%' agentic coding | Banking Dive; Reilly | February and July 2026[5][12] |
| App architecture | Patented data-streaming microservices architecture behind PNC Mobile, designed for embedded generative-AI assistance; about 8 million app users | PNC release; Banking Dive | July 7-8, 2026[6][5] |
| Data migration | Data migration factory, granted US patents 12,321,323 and 12,517,875; used for the BBVA and FirstBank conversions | Google Patents; Q2 2026 call | 2025-2026[13][14][7] |
| Embedded treasury | PINACLE Connect integrated with Oracle Fusion Cloud ERP for corporate and commercial clients | CIO Dive | July 29, 2025[15] |
Does PNC build or buy its AI technology?
Mostly build, by stated policy. Asked in February 2026 which vendors PNC uses, Reilly answered 'everybody, and a lot in-house', adding that the bank is 'inclined to, you know, use our own cooking for the most part'. In October 2026 Demchak said success needs infrastructure that lets PNC work 'without vendor support, without compute costs coming from third parties and with your own personnel instead of vendors'.[12][2][16]
The reason given is cost. Demchak told the Morgan Stanley conference in June 2026 that AI productivity 'can be taken away by the cost of tokens'; American Banker reported a spokesperson saying the aim is to avoid vendor lock-in and reliance on usage-based pricing while still mixing third-party and internal models by use case.[10][11]
What comes next, per Demchak on the Q3 2026 call, is 'the reengineering of the company using AI harnesses', which he expects to speed development and lower the cost of technology development and maintenance, followed by product delivery that aids revenue.[2]
What does the PNC profile record on this topic?
The rows below are the entries of the PNC AI strategy profile that bear on this page, as of Oct 13, 2026; the profile holds the full record of 26 dated moves, 14 use cases and 37 sources.
| Use case | What the record shows | Status | Source |
|---|---|---|---|
| Agentic software development | The system that built the new PNC Mobile app. | In production | [17] |
| Microsoft Copilot, firm-wide | Deployed across the organisation as one of several generative-AI platforms; the 2026 proxy credits it to the enterprise AI strategy. | In production | [1] |
| AI-enabled search for customer-facing teams | Named in the 2026 proxy alongside the Copilot rollout. | In production | [1] |
| Open-weight models in PNC data centres | A harness picks the cheapest adequate model; frontier models are reached through hyperscalers, open-weight models run in-house. | In production | [3] |
| Date | Move | Detail | Source |
|---|---|---|---|
| Sep 14, 2026 | Barclays conference: agents at the centre of development, 5–10x | Wiedman: using agents as the centre of software development, rather than as assistants, shows 5 to 10x productivity, work of ten weeks done in a week or less; Reilly: own data centres, own GPUs, proprietary small language models, open-weight models in-house, frontier models through hyperscalers. | [3] |
| Jul 15, 2026 | Q2 2026 call: the patented data factory ran the FirstBank integration | Demchak: the data factory, patented in its first form for the BBVA deal, performed even better inside the FirstBank integration; the continuous-improvement programme is on track for $350 million of 2026 savings; AI is affecting hiring 'at the margin'. | [7] |
| Jul 8, 2026 | About 115 AI agents built the new app; a PNC digital assistant is planned | Banking Dive: the agents handled development and testing with human review of their output; about 8 million customers use the app, mobile active users up 8% year over year; the release says the platform is designed for embedded generative-AI assistance and that frontier models were used to harden the development process. | [5][6] |
| Jul 7, 2026 | New mobile app built with agentic development | PNC's agentic development system and patented data-streaming microservices architecture. | [17][18] |
| Jun 9, 2026 | 'We are building our own AI factory' | Demchak at the Morgan Stanley US Financials Conference: own GPUs, own models, less reliance on external tokens. | [10][19] |
| May 18, 2026 | Tim Ferriter hired from JPMorgan Chase | Head of retail product, digital and growth, reporting to Overstrom; at Chase he ran the mobile app, online banking and the consumer-facing AI strategy. | [20] |
| Mar 11, 2026 | Proxy: Microsoft Copilot firm-wide, AI-enabled search, two new data-centre sites | The 2026 DEF 14A credits the enterprise AI strategy with multiple generative-AI platforms including Microsoft Copilot and AI-enabled search for customer-facing teams; a new Eastern data-centre facility is under construction and a Midwestern hosting site secured; pre-provision revenue rose to $9.3 billion 'supported by continued technology modernization, including AI-driven automation'. | [1] |
| Feb 10, 2026 | BofA conference: the 'Big Five', $1.5 billion of AI-addressable spend | CFO Rob Reilly names software, retail operations, AML compliance, the client care centre and commercial loan processing (Midland); the new mobile app was '100%' agentic coding; on vendors, 'everybody, and a lot in-house'. | [12] |
| Dec 2, 2025 | Mobile app being built by orchestrated AI agents, 40–50% complete | Head of retail banking Alex Overstrom: a series of orchestrated AI agents is building the new app faster; launch targeted ahead of the June 2026 FirstBank conversion. Earlier, the online-banking rebuild let PNC change in hours what took six months. | [21] |
| Nov 7, 2025 | Investor deck: new mobile app 'developed using agentic AI' | A conference presentation filed with the SEC previews the 2026 mobile banking launch, built with agentic AI. | [22] |
Does PNC run its own AI data centers?
Yes, by its own account. The 2026 proxy describes a national data-center strategy with a new Eastern facility under construction and a Midwestern hosting site secured, and Demchak said on October 15, 2026 that GPU chips are already installed in PNC's data centers, with the cost showing up as rising depreciation.
Which language models does PNC use?
A mix chosen per task. CFO Rob Reilly said in September 2026 that a harness picks the cheapest adequate model: frontier models through hyperscalers, open-weight models hosted in PNC data centers, and proprietary small language models. PNC has not named the hyperscalers or the open-weight model families it runs.
How was the new PNC mobile app built?
With PNC's agentic development system: about 115 orchestrated AI agents handled development and testing with human review of their output, per Banking Dive in July 2026. PNC's release says the app runs on a patented data-streaming microservices architecture designed for rapid updates and embedded generative-AI assistance.
What data foundation does PNC say AI needs?
Demchak said on the Q3 2026 call that the data lake and data configuration are 'all in one place, one source of truth', with policies digitized and HR systems that recognize agents. Ned Carroll has said 'AI, absent well-organized, well-curated data, is nothing more than interesting math on a whiteboard'.
- The PNC Financial Services Group, Inc. 2026 Proxy Statement (DEF 14A)
- PNC (PNC) Q3 2026 Earnings Call Transcript
- PNC at Barclays conference: growth, AI and branch push (transcript)
- Creating The Right Foundation and Architecture for AI (Digital Banking 2025 session)
- PNC's new mobile banking app built on personalization
- PNC's New Mobile App Delivers an Intuitive, Personalized Experience to Help Clients Seamlessly Manage Their Day-to-Day Financial Lives
- PNC (PNC) Q2 2026 Earnings Call Transcript
- PNC Completes FirstBank Customer Conversion
- 'We're coming' to 'fight you,' PNC's Demchak warns banks
- Why PNC is building its own 'AI factory'
- How banks are looking to tame their growing AI bills
- PNC Financial at Bank of America Conference: Strategic Growth in 2026 (transcript)
- US 12,321,323 B1: Data migration factory (assignee PNC Financial Services Group; filed 2024-02-20)
- US 12,517,875 B2: Data migration factory (PNC Financial Services Group; issued 2026-01-06)
- PNC embeds commercial finance platform in Oracle Fusion ERP
- PNC Quietly Building AI Agents That Don't Rely on Tech Vendors
- PNC's New Mobile App Delivers an Intuitive, Personalized Experience to Help Clients Seamlessly Manage Their Day-to-Day Financial Lives
- PNC Bank rolls out new mobile banking app
- PNC's solution to rising AI costs: build its own AI
- PNC Bank Announces Tim Ferriter as Head of Retail Product, Digital and Growth
- PNC wants human-level quality in its digital experience
- PNC investor conference presentation, November 2025 (Form 8-K exhibit)
When PNC 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