Powering Intelligence: Secured Lending in the AI Infrastructure Economy

Artificial intelligence (“AI”) is rapidly moving beyond experimentation and into broad enterprise adoption at remarkable speeds. Once confined largely to niche technology companies, research institutions and startups, AI now sits at the centre of many traditional industries, reshaping sectors such as financial services, healthcare, manufacturing, logistics, energy and professional services, while creating ripple effects across global markets. Statistics Canada reported that AI adoption among Canadian businesses increased from 6.1 per cent in the second quarter of 2024 to 12.2 per cent a year later. By Q2 2026, adoption rates had reached 19.2 per cent.^[1]

This acceleration is reshaping demand across the AI value chain. Beyond software, talent and data, companies now require the physical infrastructure capable of supporting AI at scale, making AI Data Centres (as defined below) one of the most capital-intensive asset classes in the market.^[2] Canada is emerging as a serious contender in the race to build the infrastructure behind AI. Meta’s recently announced C$13-billion data centre project in Alberta, its first in Canada, points to the country’s growing appeal as technology companies search for the energy, land and operating conditions needed to support AI at scale.^[3] For lenders, sponsors and investors, however, the opportunity comes with a more complex set of legal and commercial considerations.

The Physical Backbone of AI

Large language models, including tools such as Claude, ChatGPT, Gemini and Grok, rely on substantial computing capacity to process large volumes of data, identify patterns, generate content and perform complex tasks that previously required human analysis.^[4] These functions are supported by data centres: large facilities which house the servers, storage systems, networking equipment, cooling systems and power infrastructure required to train and deploy AI models (“AI Data Centres”).^[5]

Unlike traditional data centres, which generally support a broad range of business and information technology functions, AI Data Centres represent a more specialized form of this infrastructure, built to handle the far greater processing demands associated with AI applications, including model training, inference, computer vision and other data-intensive operations.^[6] Meeting these demands requires advanced graphics processing units, dense server configurations, large-scale storage, specialized cooling systems and reliable access to significant electricity supply.^[7]

The New Financing Landscape for AI Infrastructure

The rapid expansion of AI infrastructure is reshaping financing needs across the sector. As capital requirements rise, hyperscalers (companies that operate cloud-computing infrastructure at enormous scale) are increasingly turning to external funding sources such as corporate debt, private credit and securitization markets. For secured lenders, however, the most notable development may be the growing use of asset-based financing. AI development requires significant investment in tangible, identifiable assets, including AI Data Centres, graphics processing units, servers, cooling systems and electrical infrastructure. Together with the contractual cash flows they generate, these assets can form a strong collateral base for financing structures tailored to the needs of AI infrastructure.

Rather than relying primarily on a technology company’s overall creditworthiness or projected profitability, lenders can assess the underlying assets based on their value, useful life, location and revenue-generating potential (long-term leases or committed demand for computing capacity can create predictable revenue streams). Based on the nature of the assets, the applicable security package and the overall transaction structure, an AI infrastructure financing may draw on elements of asset-based lending, asset and equipment financing, commercial real estate financing and project financing.

Recent transactions suggest that financings of specific AI Data Centre assets (e.g., asset-backed securities (“ABS”) and single-asset single-borrower financings) is emerging as an important tool of the AI infrastructure boom, as demand for AI continues to fuel investment in digital infrastructure. In July 2025, eStruxture Data Centers secured C$1.35 billion to support its Canadian expansion, comprising C$750 million raised through ABS, described as the first rated Canadian data centre securitization backed exclusively by Canadian assets, together with a revolving facility of up to C$600 million.^[8] That same month, CPP Investments committed C$225 million for a 50 per cent interest in a construction loan supporting a hyperscale data centre expansion in Cambridge, Ontario.^[9]

For lenders, the practical concern is not simply the availability of collateral but whether an effective security package can be established, perfected and maintained. Such a package may include security over real property, equipment, accounts, insurance proceeds, material contracts and ownership interests in the entity holding the project. Where assets are held through a special purpose vehicle, lenders must also consider the obligations of the project entity, any sponsor support and the allocation of rights among creditors.

Why Traditional Financial Covenants May Not Be Enough

The specialized nature of AI hardware and infrastructure creates additional considerations for lenders. Because these assets are often highly specialized, they can present unique challenges in valuation, enforcement and resale, potentially affecting collateral preservation and recovery prospects. While lenders typically mitigate these risks through financial covenants, the effectiveness of such protections may be constrained in more complex financing arrangements or in rapidly evolving technological and market environments.

Traditionally, such covenants provide lenders with early warning signals about a borrower’s financial health over the life of the loan; however, because they rely on predefined, backward-looking metrics, they may not fully capture emerging risks in real time.^[10] By establishing objective benchmarks related to debt, cash flow, asset value and liquidity, these provisions can flag potential concerns, support ongoing monitoring of borrower performance and enable lenders to engage proactively before a default occurs, yet they may still fall short where risks are structural, project-specific or not readily reflected in financial ratios.^[11]

Financing related to AI Data Centres, and AI assets generally, may raise risks that are not fully captured by financial metrics alone. AI businesses tend to have heavy upfront research and development expenditures, intangible assets (like algorithms and data) and ambitious scaling plans, all under fast-evolving regulatory uncertainty.^[12] Unlike manufacturers with physical collateral and steady cash flows, an AI startup’s balance sheet might consist largely of code repositories and patent applications – assets that are “mysterious and largely invisible” to conventional lenders.^[13] As a result, lenders may need to consider risks that extend beyond traditional financial covenant testing.

Several risks are of particular significance to lenders:

• Asset Valuation Risk: Traditional covenant testing often assumes that a borrower’s asset base can be valued with relative consistency and monitored over time using established methodologies. In the AI Data Centre context, however, asset value may depend not only on land, buildings and physical equipment, but also on specialized hardware, software integration, proprietary operating systems and other assets whose economic value may change quickly as technology advances. This can make conventional net worth and loan-to-value analysis less reliable as a measure of long-term credit strength.
• Revenue Stability Risk: Many financial covenants are built on the assumption that borrower revenues and earnings will be sufficiently stable to support meaningful leverage and coverage testing. AI Data Centres do not always align with that framework. Their income streams can fluctuate due to shifting demand for compute resources, pricing pressures, rapid technological change and the evolving requirements of enterprise clients.
• Technology Obsolescence Risk: Unlike more conventional infrastructure assets, AI Data Centres operate in an environment where technical relevance can deteriorate rapidly. Processing equipment, cooling systems, networking architecture and power configurations may require material reinvestment on compressed timelines in order to remain commercially competitive. A facility may therefore continue to satisfy traditional financial covenants even as its underlying infrastructure becomes less efficient, less marketable or less capable of supporting current generation AI workloads.
• Limited Performance History Risk: Many AI infrastructure projects are being developed in a market that is expanding quickly but remains relatively young. In a number of cases, there is limited long-term operating history available to support covenant calibration, stress testing or benchmarking against comparable assets. This can create difficulty for lenders seeking to set covenant thresholds that are neither overly restrictive nor insufficiently protective, particularly where future performance depends on assumptions about demand growth, customer uptake and technological change.

Given these limitations, traditional covenant packages may not always provide lenders with a complete picture of borrower risk. Financing arrangements for AI Data Centres are therefore often extensively negotiated and tailored to the specific characteristics of a project, reflecting both the unique features of the underlying assets and the continued development of market practice.^[14] As the sector matures, lenders and borrowers may increasingly supplement conventional covenant frameworks with additional reporting and monitoring tools designed to provide greater visibility into project performance and credit risk.

Future View

With demand for AI infrastructure continuing to grow, Canada’s energy availability and favourable operating environment makes it a compelling market for AI Data Centre investment.^[15] The growing pipeline of announced and planned projects further signals strong momentum, rising market confidence and an expanding role for Canada in the global AI Data Centre landscape.^[16] Despite these advantages, however, AI infrastructure generates significant legal and commercial risks that traditional financing frameworks, as presently applied, may be ill-equipped to fully manage.

At the same time, the pace of technological development in AI is being matched by increasing regulatory scrutiny, with policymakers and industry participants focusing more closely on issues such as transparency, data governance and market oversight. The ultimate shape of the AI regulatory landscape remains uncertain, making regulatory risk an increasingly relevant consideration in the evaluation and financing of long-term AI infrastructure projects.

In this environment, early engagement with Canadian legal counsel is critical to ensure that financing structures are appropriately tailored to the realities of this rapidly evolving asset class. As investment in AI infrastructure continues to accelerate, careful legal and risk allocation planning will be essential to unlocking its full potential while preserving long-term value.

The Financial Services Group at Aird & Berlis LLP advises lenders, borrowers and investors on financing structures for AI infrastructure and data centre projects. If you have questions about financing AI infrastructure or related matters, please contact the authors or a member of the group.

[1] Table 2, Analysis on Artificial Intelligence Use by Businesses in Canada, Second Quarter of 2026 | Government of Canada