Oslo, Norway – A recent comprehensive analysis by Nordea analysts Kirsti Sunde Midttun and Ole Håkon Eek-Nielsen has cast a critical eye on the long-term profitability of artificial intelligence (AI) business models, citing structural pressures stemming from exorbitant inference costs, the rapid depreciation of cutting-edge models, and escalating competition from free and open-source alternatives. Their report, which comes amidst a significant buildout of AI infrastructure driving a meaningful share of US economic growth, directly challenges the sustainability of current industry practices and highlights a growing skepticism among investors, leading to a notable rotation out of technology stocks and into more cyclical, defensive, and value-oriented sectors.

The Nordea research posits that despite the seemingly relentless growth and transformative potential of AI, the underlying economics present substantial hurdles to achieving widespread, sustained profitability for many of the leading model developers. "With the AI buildout now driving a meaningful share of US growth, we examine the sustainability of the underlying business models and whether the recent market scepticism is warranted," the analysts stated, concluding with a more cautious outlook on the industry’s financial prospects than often portrayed in mainstream narratives.

The Triple Threat to AI Profitability

Midttun and Eek-Nielsen pinpoint three primary structural challenges that collectively undermine the durability of AI margins:

1. High Inference Costs: The operational expenses associated with running AI models, particularly large language models (LLMs), at scale are astronomical. Known as inference costs, these involve the computational power (primarily high-performance Graphics Processing Units, or GPUs), energy consumption, and cooling required to process user queries and generate responses. Unlike the one-time cost of training a model, inference costs are ongoing and scale directly with usage. "The net effect is that inference costs remain the central economic challenge for AI developers, and a key reason why the leading model companies are, for now, not profitable," the report emphasizes. Each interaction with a sophisticated AI model consumes significant resources, making it challenging to achieve positive unit economics, especially for consumer-facing applications that might generate millions of queries daily.

2. Rapid Model Depreciation: The pace of innovation in the AI sector is unprecedented, leading to an unusually short useful life for even the most advanced "frontier models." Nordea analysts describe these models as "infrastructure with an unusually short useful life: the value must be extracted before the technology is obsolete." A model considered state-of-the-art today can be surpassed by a newer, more efficient, or more capable iteration within a matter of months. This rapid obsolescence necessitates continuous, massive reinvestment in research and development, as well as in the retraining and deployment of new models, creating an ongoing capital expenditure cycle that pressures profitability. The constant need to upgrade hardware and software to keep pace with evolving capabilities further exacerbates this issue.

3. Growing Competition from Free and Open Alternatives: The proliferation of powerful open-source AI models, exemplified by Meta’s Llama series and offerings from companies like Mistral AI, is exerting significant downward pressure on pricing across the market. These readily available, often highly capable models provide developers and businesses with viable alternatives to proprietary, paid solutions. "Publishing capable models free of charge suppresses willingness to pay across the market and undercuts the business models of developers who charge for access," the analysts note. This phenomenon, while beneficial for democratizing AI access and fostering innovation, fundamentally challenges the monetization strategies of companies that invest heavily in developing proprietary models and rely on subscription or usage-based fees.

The AI Hype Cycle and Investor Re-evaluation

The Nordea report’s findings arrive after a period of unparalleled investor enthusiasm for AI, particularly following the public launch of OpenAI’s ChatGPT in November 2022. This event catalyzed a global interest in generative AI, leading to a surge in investment, a re-rating of technology stocks, and a significant portion of the market’s attention directed towards companies perceived as leaders in the AI race. Chipmakers like Nvidia, cloud providers such as Microsoft and Amazon Web Services (AWS), and software giants like Google saw their valuations soar, driven by expectations of a new era of productivity and innovation.

Throughout much of 2023, the narrative around AI was overwhelmingly positive, with venture capital pouring into AI startups and established tech companies aggressively repositioning themselves as AI-first entities. However, as the summer of 2023 progressed, Nordea observed a palpable shift in investor sentiment. "Over the summer, we have also seen some scepticism towards AI-related equities," the report highlights. This skepticism is not a rejection of AI’s technological prowess but rather a more pragmatic assessment of the economic realities and the path to profitability. Investors began to demand clearer business models, tangible returns on investment, and sustainable growth strategies beyond the initial hype.

This re-evaluation has manifested in a "notable rotation out of tech stocks and into cyclical, defensive, and value-oriented sectors." This shift reflects a broader market trend where investors, facing persistent inflation, rising interest rates, and geopolitical uncertainties, are prioritizing companies with stable earnings, strong balance sheets, and proven profitability over high-growth, often unprofitable, technology ventures. The AI sector, while still commanding significant interest, is no longer immune to these fundamental economic considerations.

Background Context: The AI Buildout and its Costs

The "AI buildout" referenced by Nordea refers to the massive global investment in the infrastructure necessary to develop, train, and deploy AI models. This includes:

• Advanced Semiconductors: The demand for high-performance GPUs, primarily from Nvidia, has skyrocketed. These chips are the backbone of AI training and inference. Companies like Google and Amazon are also developing their custom AI chips (TPUs and Inferentia/Trainium respectively) to control costs and optimize performance.
• Data Centers: Hyperscale data centers are being expanded and built at an unprecedented rate to house the computational power and store the vast datasets required for AI. These facilities consume enormous amounts of electricity and require sophisticated cooling systems. Estimates suggest that AI-related power consumption could dramatically increase global electricity demand in the coming years.
• Talent Acquisition: The competition for AI researchers, engineers, and data scientists is fierce, driving up salaries and operational costs for companies.
• Data Acquisition and Curation: Training large models requires access to vast, diverse, and high-quality datasets, which can be expensive to acquire, license, and curate.

The scale of these investments is staggering. Major tech companies have pledged tens, if not hundreds, of billions of dollars towards AI initiatives. For instance, Microsoft’s multi-billion dollar investment in OpenAI, Google’s continuous R&D into its Gemini models and TPUs, and Amazon’s commitment to AI services through AWS all underscore the immense capital outlay involved. While these investments are driving innovation and capability, they also create a formidable barrier to entry and place immense pressure on eventual monetization.

Industry Reactions and Future Outlook (Inferred)

While Nordea’s report offers a critical perspective, it is important to note that major AI developers are acutely aware of these challenges and are actively working on solutions.

• Optimization and Efficiency: Companies like OpenAI, Google DeepMind, and Anthropic are heavily invested in making their models more efficient, reducing the computational resources required for both training and inference. This includes research into smaller, more specialized models, novel algorithmic approaches, and custom hardware designs. The development of techniques like quantization and pruning aims to reduce model size and accelerate inference without significant performance degradation.
• Diversified Monetization Strategies: Beyond direct API access, AI companies are exploring various revenue streams. This includes enterprise solutions tailored to specific industry needs (e.g., healthcare, finance, legal), partnerships with software vendors to embed AI capabilities, and developing niche applications where the value proposition is clearer and customers are willing to pay a premium. The focus is shifting from generic AI to highly specialized and integrated AI solutions that solve specific business problems.
• Hardware-Software Co-design: The tight integration of hardware and software is seen as crucial for future cost optimization. Companies designing their own AI chips, like Google with its TPUs, aim to create highly efficient ecosystems that can deliver superior performance at lower operational costs compared to general-purpose GPUs.
• Ethical AI and Trust: As AI becomes more pervasive, the emphasis on explainability, fairness, and safety is also growing. Building trustworthy AI solutions can differentiate offerings and potentially command higher prices, though it adds another layer of complexity and cost.

From an investor’s perspective, the Nordea report serves as a timely reminder that technological innovation, while exciting, must eventually translate into sustainable economic value. The market’s shift reflects a maturation of the AI narrative, moving from unbridled optimism to a more discerning evaluation of business fundamentals. This does not imply an "AI winter" – the underlying technology continues to advance rapidly – but rather a recalibration of expectations regarding the timeline and mechanisms for profitability.

Broader Implications for the Tech Sector and Economy

The implications of Nordea’s findings extend beyond just AI model developers:

• Sector Consolidation: The high costs and intense competition could lead to consolidation in the AI industry. Smaller startups with less capital might struggle to compete with well-funded tech giants, leading to acquisitions or failures.
• Focus on Value Chains: The value in AI might increasingly shift towards specific parts of the value chain. For instance, companies providing the foundational data, specialized hardware (like Nvidia), or unique integration services might capture more sustainable profits than those merely offering generic model access.
• Enterprise vs. Consumer AI: The report suggests that enterprise applications, where AI can deliver clear, measurable ROI (e.g., automating customer service, optimizing supply chains, accelerating drug discovery), might be the more immediate and viable path to profitability compared to broad consumer-facing models.
• Macroeconomic Impact: If AI’s profitability challenges persist, it could temper overall economic growth projections that heavily rely on AI-driven productivity gains. It could also influence capital allocation decisions across the technology sector, favoring proven business models over speculative growth plays.
• Regulatory Scrutiny: As AI systems become more powerful and pervasive, regulatory scrutiny regarding data privacy, algorithmic bias, and market concentration is also increasing. Compliance costs will add another layer of expense for AI developers.

In summary, the Nordea analysis by Kirsti Sunde Midttun and Ole Håkon Eek-Nielsen offers a sobering, yet essential, counter-narrative to the prevailing AI optimism. While the transformative potential of AI remains undeniable, the path to sustained profitability is fraught with significant structural challenges. The combination of high inference costs, rapid model obsolescence, and fierce competition, particularly from free alternatives, is forcing a critical re-evaluation by investors and industry players alike. The market is increasingly demanding not just groundbreaking technology, but also clear, sustainable business models that can withstand these economic headwinds and deliver tangible returns.