Page 100 - ISC PROCEEDINGS 21.4

target of 30% (NSO, 2026), and TFP accounted for only approximately 20% of output-per-
worker growth in the analyzed period, the remainder driven by factor accumulation
rather than technological progress (CSIRO, 2025). Labor productivity stood at USD 9,182
per worker in 2024, growing at 5.88% annually but remaining significantly below regional
peers Thailand and Malaysia (Vietnam.vn, 2025). R&D expenditure at approximately
0.43% of GDP, far below the OECD's 2-4% average (World Bank, 2024a), further constrains
the endogenous innovation capacity required to sustain AI-led TFP growth.
The central research problem this paper addresses is the divergence between
Vietnam's high AI adoption intensity and its modest economy-wide TFP and productivity
gains, what this paper terms the "Vietnam AI productivity gap." This constitutes a critical
research gap: while existing studies examine either AI adoption in Southeast Asia or
Vietnam's macroeconomic trajectory, few integrate both dimensions through the lens of
productivity theory. Virtually none account for the transformative policy architecture
emerging between 2024 and 2026, notably Politburo Resolution No. 57-NQ/TW
(December 22, 2024), the AI Law (Law No. 134/2025/QH15, effective March 1, 2026), and
the National Assembly's 2026 socio-economic plan targeting GDP growth of at least 10%
and labor productivity growth of 8.5% (NAV, 2025). Closing this research gap is essential
for understanding how structural constraints and institutional reform interact to
determine whether AI-driven TFP growth is achievable in a lower-middle-income
economy.
This paper pursues three research objectives: (1) to theoretically situate Vietnam's
AI-TFP nexus within GPT theory, the AI productivity paradox, and TFP accounting
frameworks; (2) to analyze the empirical landscape of AI deployment in Vietnam during
2021-2025, identifying the key structural constraints, skills deficits, data fragmentation,
and institutional gaps, that explain the productivity gap; and (3) to develop a policy
reform framework, grounded in Vietnam's specific institutional context and latest
legislative architecture, that enables AI's transition from experimentation to scaled,
economy-wide TFP contribution.
The paper makes three contributions. Theoretically, it advances the application of
GPT theory and TFP decomposition to a lower-middle-income economy in rapid digital
transition, generating new analytical insights into the conditions under which developing-
country "AI adoption" translates into "AI productivity" and TFP growth. Empirically, it
provides the most current and comprehensive synthesis of Vietnam's AI-economy
landscape through 2025-2026, integrating data from the OECD Economic Surveys: Viet
Nam 2025, the e-Conomy SEA 2025 report, and the World Bank's 2025 Vietnam Economic
Update. In terms of policy, it delivers a structured, actionable reform framework
anchored in Vietnam's emerging legal and strategic instruments, including the National AI
Development Fund, BIM mandates, and the CBDC-AI productivity nexus, not previously
analyzed in academic literature.
2. Theoretical framework and literature review
2.1. Theoretical foundations of AI as a general purpose technology
The dominant theoretical lens for analyzing AI's relationship to productivity is the
General Purpose Technology (GPT) framework, developed by Bresnahan and Trajtenberg
(1995) and subsequently applied to ICTs by Helpman (1998) and Brynjolfsson and McAfee
(2014). A GPT is characterized by pervasiveness across sectors, continuous improvement,
and capacity to spawn complementary innovations, criteria that AI satisfies fully: it has
penetrated agriculture, healthcare, manufacturing, finance, and public administration;

99