Researchers at HSE University have determined that the global generative artificial intelligence (GenAI) market is evolving more rapidly than it is generating revenue. Their findings, published in the journal Foresight and STI Governance, raise questions about the sustainability of current investment patterns in AI technologies.
In recent years, GenAI has attracted substantial financial backing, with companies investing billions of dollars into hardware such as chips, servers, and data-center infrastructure, anticipating quick economic returns from advanced large language models. However, according to the study led by Yaroslav Kuzminov, Academic Supervisor at HSE University, and Ekaterina Kruchinskaia, an Associate Professor in the Faculty of Social Sciences, there appears to be a significant disconnect between these hefty investments and the actual revenue generated from AI solutions.
The research employed the Data Envelopment Analysis (DEA) method, a framework used to evaluate the efficiency of complex economic systems based on multiple input and output factors. In this context, ‘inputs’ constituted revenues from AI hardware manufacturers such as AMD, Intel, and NVIDIA, while ‘outputs’ represented revenues from companies developing AI applications, including OpenAI, Google DeepMind, Amazon, and Apple.
The study spans the years 2016 to 2024, treating individual years as the units of analysis rather than companies. This approach aimed to provide a comprehensive overview of the efficiency of AI development over time, rather than focusing on specific entities. To ensure the robustness of the findings, the researchers conducted their calculations in both absolute terms and adjusted for global GDP, allowing for a relative efficiency assessment of the generative AI market across different years.
The analysis revealed that the development of the GenAI market is nonlinear. Efficiency surged during the initial commercialization phase from 2016 to 2021, but began to decline in 2021, despite increased investment. After a brief recovery in 2023, efficiency metrics reverted to levels seen in 2022. As Kruchinskaia explained, “From a purely methodological perspective, the results suggest that the AI solutions market is developing according to a catch-up model: revenues from software products do not yet compensate for the massive investment in hardware infrastructure.”
The researchers indicated that while demand for chips and computing power is amplified by the growth of large language models, the commercial returns from these models remain limited, failing to offset the substantial costs associated with hardware technologies and ongoing investments. This development model appears to bolster hardware manufacturers’ positions but yields restricted economic returns, as investments in computing power are becoming an end in themselves.
Challenges facing the market for AI solutions include high hardware costs, a shortage of qualified personnel, and the technological limitations of existing models. Additionally, concerns persist about the sector’s capability to generate adequate revenue relative to the scale of investment required. “AI is indeed transforming not only the economy and companies’ business models but also everyday social life. This influence is spreading more slowly than it may appear and is less productive than many would like,” Kuzminov stated.
The researchers caution against potential market bubbles, a phenomenon not unfamiliar in the global economy. They assert that these risks are tangible, emphasizing the need for more practical discussions around the application and efficiency of AI technologies. “Without improving the efficiency of applied solutions, expanding their adoption, and pursuing more balanced investment planning, further positive progress will be difficult,” Kuzminov added.
This study not only serves as a critical resource for the academic community but also offers valuable insights for businesses and investors navigating the evolving landscape of artificial intelligence. The insights gained could inform future investment strategies and technology policies aimed at fostering a more sustainable and effective generative AI market.
See also
Google Launches Lyria 3 Music Generation Model with 30-Second Song Limit in Gemini App
NVIDIA Researchers Reveal Uniform-State Diffusion Surpasses Masked Models in Reasoning Tasks
Tavus Launches Phoenix-4 Model with Real-Time Emotional Intelligence and Sub-600ms Latency
Top 10 AI Image Generators of 2026: Tools for Stunning Visuals Unveiled
Google Launches Lyria 3 in Gemini App for AI Music Creation from Text and Images
