What technology stack does TD Bank run, and how is AI built on it?
TD runs on Microsoft Azure, adds Google Cloud for trading workloads, and builds its main models in-house at Layer 6. TD chose Azure as its 'cloud foundation' in April 2019 and in January 2022 accelerated an enterprise data estate on Azure under a multi-year Databricks agreement; a Databricks blog written in TD's voice describes a single lakehouse on Azure Databricks run by a Data as a Service group of over 2,000 colleagues, and the June 2025 Prism release says a large-scale migration of data records into the cloud platform was completed 'last year'. Google Cloud joined in April 2024, starting with Google Kubernetes Engine for TD Securities Automated Trading. On that base, TD AI Prism, a predictive foundation model built internally, runs 'on the Bank's secure cloud platform ensuring customer data does not leave the Bank'. Generative AI comes from partners: TD Securities' assistant uses OpenAI's enterprise GPT as 'a natural extension of TD's cloud partnership with Microsoft and Azure', and Cohere's models are being explored with Layer 6 under a commitment of up to $25 million over three years (September 2026). GitHub Copilot has reached all TD engineers and developers, including 1,800 in the U.S. Technology and equipment expense was $2,872 million in fiscal 2025, in Canadian dollars, and the U.S. bank's 2026 priority is to deliver 'a scalable, cloud-native, modular environment'. TD does not disclose its cloud share or AI spending.[1][2][3][4][5][6][7][8][9][10]
Data as of Oct 31, 2026 · TD profile updated Oct 28, 2026 · 22 sources cited on this page · 141 banks profiled on this site · Markdown
Which clouds, data platforms and models make up TD's stack?
Group-level choices that the U.S. bank shares unless marked U.S. 'Since' is the first dated mention.
| Layer | What TD uses | Since | Who said it |
|---|---|---|---|
| Public cloud | Microsoft Azure as 'the cloud foundation' for technology and design teams; by 2022 Microsoft Canada described Azure as TD's primary cloud platform across all lines of business | April 23, 2019 | TD and Microsoft[1][2] |
| Second cloud | Google Cloud under a multi-year relationship; Google Kubernetes Engine already supported TD Securities Automated Trading (TDSAT), based in Chicago | April 3, 2024 | TD and Google Cloud[5] |
| Data platform | An enterprise data estate on Azure with a multi-year Databricks agreement and Delta Lake; Azure Synapse and Azure Machine Learning for data scientists and analysts | January 26, 2022 | Jeff Martin, SVP Corporate Platforms[2] |
| Data as a Service | A single lakehouse on Azure Databricks run by a Data as a Service organization 'over 2,000 colleagues strong'; described as 'a new, secure, stable, and trusted AI platform' | June 21, 2023 | Databricks blog, written in TD's voice[3] |
| Cloud data migration | A large-scale migration of data records into the Bank's cloud platform completed 'last year', per the June 11, 2025 release; Vlad Shpilsky said it 'helped fuel AI-innovation' | June 11, 2025 | TD[4] |
| In-house models | TD AI Prism, a fully internally built predictive foundation model grown from Layer 6 research, processing 100 times more data variables and replacing many individual models | June 11, 2025 | TD[4] |
| Generative AI models | OpenAI's enterprise GPT for the TD Securities assistant via the Microsoft partnership; Cohere's secure enterprise models explored with Layer 6, up to $25 million over three years | October 9, 2025; September 22, 2026 | Dan Bosman, TD Securities; TD[6][7] |
| Coding tools | GitHub Copilot rolled out to all TD engineers and developers, 76% of them active users; 1,800 U.S. developers enabled | 2026 | TD Stories; Leo Salom[8][9] |
How much does TD spend on technology, and how many people build its AI?
Expense lines are from the fiscal 2025 annual report (Canadian dollars, group); the U.S. priority is from the U.S. Retail section of the same report.
| Measure | Value | As of |
|---|---|---|
| Technology and equipment expense | $2,872 million, against $2,656 million in fiscal 2024 | Fiscal 2025[10] |
| Equipment, data processing and licenses | $2,572 million, against $2,379 million in fiscal 2024 | Fiscal 2025[10] |
| AI and data builders | 2,500 data scientists, engineers, data analysts and experts building proprietary platforms and applications; the 2025 investor day counted more than 2,500 AI-related roles | September 29, 2025; October 9, 2025[6][11] |
| Engineers using AI | Over 7,000 engineers using AI for software development; the most active saw a 29% increase in throughput | May 28, 2026[12] |
| Pull request throughput | Up 80% from 2025 to 2026, aided by GitHub Copilot | 2026 (to July 31)[8] |
| U.S. technology priority | 'Transform data and technology architecture to deliver a scalable, cloud-native, modular environment' and 'Deploy AI to automate operational processes' | Priorities for 2026[10] |
Does TD build or buy its AI technology?
TD builds the predictive layer and buys the generative one. Layer 6 built TD AI Prism on TD's own data inside its cloud platform, while generative assistants run on OpenAI's enterprise GPT through Microsoft; Bosman said 'With Microsoft's partnership with OpenAI, there's an enterprise-grade platform'. The September 2026 Cohere deal adds a second model source, with Cohere experts working alongside Layer 6 at the MaRS Discovery District.[4][6][7]
The data platform came first. The Databricks blog, written in TD's voice, says the lakehouse on Azure was meant to simplify the data platform architecture, speed time to market and lower build and run costs, with self-healing operations now standard; Shpilsky credits the cloud migration with making data easier to access at scale for AI.[3][4]
Not disclosed: the share of workloads in the cloud, GPU or on-premises AI infrastructure, model-provider spending, or technology spending for the U.S. bank on its own.[10]
What does the TD profile record on this topic?
The rows below are the entries of the TD AI strategy profile that bear on this page, as of Oct 28, 2026; the profile holds the full record of 39 dated moves, 18 use cases and 55 sources.
| Use case | What the record shows | Status | Source |
|---|---|---|---|
| TD AI Prism personalisation | Predictive foundation model for customer needs and marketing. | In production | [4][13] |
| Knowledge-management assistants | Branch, wealth, securities and insurance virtual assistants built by Layer 6. | In production | [14][15][16] |
| AI-assisted software development | 1,800 US developers on GitHub Copilot with 30-40% technology-capacity gains; 7,000-plus engineers group-wide, the most active achieving a 29% increase in throughput. | In production | [9][12] |
| TD Securities AI Virtual Assistant | Equity research and market-data assistant on OpenAI enterprise GPT, built with Layer 6; turns research queries into summaries, tables and SQL-backed charts for sales and trading. | In production | [6][17] |
| ML in transaction monitoring and the new KYC platform | ML models went into transaction monitoring in 2025 with more planned; the centralised KYC platform went live to business users in February 2026 and an improved customer risk rating model followed. | In production | [18][19][12] |
| Cohere collaboration with Layer 6 | Up to C$25 million over three years; a dedicated Cohere team works beside Layer 6 at MaRS on knowledge management, security and responsible adoption. | Announced | [7] |
| Date | Move | Detail | Source |
|---|---|---|---|
| Sep 22, 2026 | Layer 6 and Cohere: up to C$25 million over three years | Strategic collaboration on enterprise AI, with a focus on knowledge management, security and responsible adoption. | [7] |
| Sep 17, 2026 | US bank details AI in AML, engineering and the call centre | Three AI models in the AML program, an investigator desktop in development, 1,800 developers on GitHub Copilot, and a vendor voice AI agent targeted at about 50% of incoming calls. | [9] |
| Sep 3, 2026 | Shpilsky reappointed to lead Global Technology and Solutions | Senior leadership appointment continuing his mandate over the group's technology and digital organisation, which has 'deployed new innovations and platforms and launched new AI-enabled applications'. | [20] |
| May 28, 2026 | Q2 fiscal 2026 call: US AML runs on AI-enhanced transaction monitoring | Salom: new monitoring system with embedded ML and AI, new KYC platform and improved customer risk rating model; group: over 40,000 colleagues using Copilot (the call did not name the product's vendor), over 7,000 engineers using AI, mortgage pre-adjudication cut from about 15 hours to 3 minutes. | [12] |
| Feb 26, 2026 | Q1 fiscal 2026 call: US ML models in transaction monitoring, new KYC platform live | Salom: the new KYC platform went live to business users that month and more ML models follow in coming quarters; Chun: first scaling of agentic AI in real-estate lending, C$1 billion AI target split C$500 million revenue and C$500 million expense, with upside beyond it. | [18][21] |
| Oct 20, 2025 | TD Wealth Virtual Assistant | A generative knowledge-management system built by Layer 6 for wealth colleagues. | [14] |
| Oct 9, 2025 | TD Securities assistant runs on OpenAI enterprise GPT | CIO Dan Bosman: chosen as a natural extension of TD's Microsoft and Azure partnership; built with Layer 6; TD has about 2,500 people building AI platforms and applications. | [6][17] |
| Jun 11, 2025 | TD AI Prism launched | A predictive foundation model to anticipate customer needs, built entirely in-house. | [4] |
| Apr 22, 2025 | Layer 6 opens a New York office | To bring gen-AI expertise and delivery in-house for the US bank, per Leo Salom. | [22] |
| Jan 1, 2018 | TD acquires Layer 6 | The Toronto AI research firm becomes the group's AI R&D centre. | [22] |
Which cloud does TD Bank use?
Mainly Microsoft Azure, chosen as TD's cloud foundation in April 2019 and described by Microsoft Canada in 2022 as its primary cloud across all lines of business. Google Cloud was added under a multi-year relationship in April 2024, starting with TD Securities' automated trading business in Chicago.
Does TD build its own AI models?
Its predictive models, yes. Layer 6, the AI lab TD acquired in 2018, built TD AI Prism, a predictive foundation model launched in June 2025 that runs on TD's secure cloud platform so customer data does not leave the bank. For generative AI TD uses partners' models, notably OpenAI's enterprise GPT and, from September 2026, Cohere's.
What data platform does TD use?
A lakehouse on Azure Databricks and Delta Lake. TD signed a multi-year Databricks agreement in January 2022 as part of moving its enterprise data estate to Azure, and its Data as a Service organization of more than 2,000 colleagues runs the platform. TD said in June 2025 that a large-scale migration of data records into the cloud had been completed.
How much does TD spend on technology?
TD Bank Group reported technology and equipment expense of $2,872 million in fiscal 2025, against $2,656 million in fiscal 2024, in Canadian dollars. It does not break out AI spending or the U.S. bank's technology budget.
Does TD's U.S. bank have its own technology stack?
It shares the group's clouds, data platform and Layer 6 models, which has an office in New York. U.S.-specific items on the record are 1,800 developers with GitHub Copilot and a 2026 U.S. Retail priority to deliver a scalable, cloud-native, modular data and technology architecture.
- TD and Microsoft Enter into a Strategic Relationship to Power the Future of Digital Customer Experiences
- Data platform modernization delivering richer insights and powering dynamic digital experiences for TD customers
- How Databricks' Lakehouse is helping to power a new era for TD Bank Group's Data Transformation
- TD Announces Launch of Groundbreaking Predictive Foundation Model
- TD and Google Cloud Enter into a Strategic Relationship to Power Innovative Banking Experiences
- 'Give me five trade ideas': TD's traders embrace ChatGPT
- TD announces $25 million strategic relationship with Cohere to support AI adoption
- How some Canadians feel about using AI at work (TD Stories; undated page with TD figures to July 31, 2026, retrieved 2026-10-31)
- TD Bank Targets 100 U.S. Branches as AML Remediation Continues
- TD Bank Group 2025 Annual Report: Management's Discussion and Analysis (Form 40-F, Exhibit 99.2)
- TD Investor Day 2025: full presentation
- TD (TD) Q2 2026 Earnings Call Transcript
- TD Bank Builds AI Model to Predict What Customers Need and Not What They Don't Want
- New TD Wealth Virtual Assistant Set to Enhance the Colleague and Client Experience
- Virtual AI Assistant To Help Power TD Securities
- TD Bank Group Reports First Quarter 2026 Results (report to shareholders)
- TD Securities taps Layer 6 and OpenAI to deliver real-time equity insights to sales and trading teams
- TD Bank Group Q1 2026 Earnings Conference Call transcript
- TD has completed 'majority' of US AML remediation
- TD Bank Group announces senior leadership appointments to further accelerate strategy
- TD Bank Group First Quarter 2026 Report to Shareholders (Form 6-K, Exhibit 99.1)
- TD Announces Opening of New Layer 6 Office in New York City
When TD 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 PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning