Measuring
Artificial Intelligence Investment in Canada: A Functional Approach within the
System of National Accounts
Manir
Hossain
March 2026
Abstract
As
artificial intelligence (AI) adoption accelerates across the Canadian economy,
accurately quantifying AI-related investment has emerged as an important
macroeconomic measurement challenge. This paper develops a functional approach
to measuring AI investment within the Canadian System of National Accounts
(CSNA).
National
accounting frameworks—including the System of National Accounts 2008 (SNA 2008)
and ongoing updates toward SNA 2025—classify assets by economic function rather
than underlying technology. As a result, AI is not recognized as a distinct
asset class. Instead, AI-related investment is embodied within existing
categories of gross fixed capital formation.
The paper
argues that AI capital is distributed across intellectual property products
(IPP), including software, research and development (R&D), and databases,
supported by complementary investment in machinery and equipment (M&E) and
non-residential structures. Approximating AI investment therefore requires a
functional (use-based) interpretation of existing asset categories, which
aligns investment with its role in AI production. This approach preserves the
integrity of the national accounts while providing a practical framework for
analyzing AI’s contribution to capital formation and economic growth in Canada.
1.
Introduction
Artificial
intelligence (AI) has emerged as a transformative technology with the potential
to reshape economic activity through its productivity-enhancing role in
production processes, business operations, and service delivery. In Canada, AI
adoption is steadily rising; Statistics Canada’s Survey of Business Conditions
indicates that 14.5% of businesses used AI in 2025Q3, up from 6.1% in 2024Q2. This
enterprise growth is mirrored by widespread individual adoption; by the second
half of 2025, 35.0% of the Canadian working-age population regularly used
generative AI tools, placing Canada among the top 15 nations globally for AI
diffusion (Microsoft, 2026). This upward trend highlights the growing need for
researchers and policymakers to quantify the scale and composition of
AI-related investment. However, measuring AI investment presents conceptual and
practical challenges for national accounting systems, which classify assets
based on their economic function rather than the technologies they embody.
This paper
examines how AI investment can be measured within the existing framework of the
Canadian System of National Accounts (CSNA), consistent with the international
System of National Accounts 2008 (SNA 2008) and emerging guidance associated
with ongoing updates toward SNA 2025. The analysis focuses on identifying the
asset categories through which AI-related capital formation is recorded, with
particular emphasis on intellectual property products (IPP), while also
considering machinery and equipment (M&E) and non-residential structures.
2.
Conceptual Treatment of AI in the SNA Framework
Artificial
intelligence is not explicitly identified as a distinct asset class in the SNA
2008 framework. Instead, AI investment must be defined functionally as capital
formation undertaken to develop, deploy, or operate AI systems. Under this
approach, AI-related capital is embodied within existing asset categories when
acquired for AI production purposes.
Within the
current framework, AI-related capital formation is best understood as being
distributed across multiple components of intellectual property products (IPP),
including software, research and development (R&D), and data assets. While
many AI systems are implemented through software, their economic value also
reflects underlying research activity and the accumulation of structured data
used in training and deployment.
This
principle of relying on existing asset classifications continues to guide
ongoing updates to the SNA. Rather than introducing new asset categories for
specific technologies, international statistical efforts emphasize improving
the measurement of digital and intangible assets within the existing framework.
AI has been identified as a priority area within this broader digitalization
agenda, although statistical definitions and measurement approaches remain
under development.
A key
implication of this framework is that AI investment cannot currently be
directly observed in standard national accounts statistics. Instead, it must be
approximated using supplementary indicators, or allocation techniques applied
to existing asset categories. This introduces both conceptual and empirical
challenges in distinguishing AI-specific investment from broader digital and
intangible capital formation.
3.
Measuring AI Investment in the Canadian Context
3.1
Intellectual Property Products (IPP)
Within the
CSNA, IPP comprises computer software, research and development (R&D), and
databases. Statistics Canada capitalizes both purchased and own-account
software, including internally developed and externally acquired systems used
in production. Accordingly, AI systems—such as machine learning applications
and predictive analytics tools—are recorded as software investment when they
meet capitalization criteria.
R&D is
also treated as a fixed asset, consistent with SNA 2008. Although AI-specific
R&D is not separately identified, aggregate R&D investment includes
expenditures related to algorithm development, experimental model design, and
applied AI research.
Databases
represent another important component. Investments in data engineering,
curation, and maintenance that support AI systems are capitalized when they
meet asset boundary conditions such as ownership, control, and measurable
economic value. However, not all AI-relevant data assets are fully captured,
particularly those generated through informal or platform-based activities.
Taken
together, software, R&D, and databases constitute the core of AI-related
capital formation within the national accounts.
3.2
Machinery and Equipment (M&E)
AI-related
capital within machinery and equipment investment is concentrated in
information and communication technology (ICT) equipment, particularly
high-performance computing systems used for model training and deployment.
While communications equipment may support distributed AI processes, most
non-ICT equipment plays a limited direct role in AI production.
A key
measurement challenge is that ICT equipment is typically multi-purpose, making
it difficult to isolate the portion attributable specifically to AI activities.
3.3
Non-Residential Structures
Non-residential
construction investment includes facilities such as data centres that support
digital infrastructure. A subset of these facilities is designed or upgraded to
accommodate AI workloads, particularly those requiring intensive computational
capacity.
However,
most data centres support a wide range of digital services beyond AI.
Distinguishing AI-specific infrastructure from general-purpose digital
infrastructure requires additional assumptions and supplementary data.
3.4
Measurement Challenges
Beyond
classification issues, several measurement challenges complicate the estimation
of AI investment within the CSNA:
- Identification: AI-related
expenditures are not separately reported in standard statistical sources.
- Price measurement: Rapid
improvements in AI systems complicate quality adjustment and the
construction of appropriate deflators.
- Boundary issues: AI services
delivered through cloud platforms may be recorded as intermediate
consumption rather than capital formation.
- Data valuation: The economic
value of data used in AI systems remains difficult to quantify and is only
partially captured.
These
challenges imply that current estimates of AI investment are approximate and
likely incomplete.
4.
Conclusion
As
artificial intelligence continues to reshape the Canadian economy, accurately
measuring AI-related investment has become increasingly important. Under SNA
2008 and consistent with ongoing international statistical guidance, AI is not
defined as a distinct asset category within the national accounts. Instead, it
is embedded within existing components of gross fixed capital formation,
particularly within intellectual property products.
A functional
(use-based) interpretation of existing asset categories, which attributes
investment according to its role in AI production, allows analysts to
approximate AI investment while maintaining consistency with established
accounting principles. However, significant conceptual and empirical challenges
remain, suggesting that current estimates should be interpreted as partial and
evolving. Continued methodological development will be essential for improving
the measurement of AI in official statistics.
References
Corrado, C.,
Haskel, J., Iommi, M., & Jona-Lasinio, C. (2022). Measuring data as an
asset: Framework, methods and preliminary estimates (OECD Economics
Department Working Papers No. 1731). OECD Publishing.
Intersecretariat
Working Group on National Accounts. (2023). Guidance note on the treatment
of artificial intelligence and digital assets in the national accounts
(Draft/background paper). United Nations Statistics Division.
Microsoft.
(2026). Global AI adoption in 2025: A widening digital divide. https://www.microsoft.com/en-us/research/wp-content/uploads/2026/01/Microsoft-AI-Diffusion-Report-2025-H2.pdf
Organisation
for Economic Co-operation and Development. (2019). Guidelines on measuring
the digital economy. OECD Publishing.
Organisation
for Economic Co-operation and Development. (2023). Measuring the digital
transformation: A roadmap for the future. OECD Publishing.
Statistics
Canada. (2022). Annual survey of research and development in Canadian
industry (RDCI).
Statistics
Canada. (2023a). Capital and repair expenditures, non-residential tangible
assets.
Statistics
Canada. (2023b). Investment by asset type and industry. Canadian Economic
Accounts.
Statistics
Canada. (2023c). Digital infrastructure and the Canadian economy.
United
Nations, European Commission, International Monetary Fund, OECD, & World
Bank. (2009). System of National Accounts 2008.
United
Nations Statistics Division. (2023). Guidance on digitalization and emerging
technologies.
