Context

• The artificial intelligence industry has entered a critical phase. After years of rapid expansion and heavy capital deployment, the central issue has shifted from technical capability to long-term profitability.
• Massive spending on compute power, data centres and foundational systems has proven that AI can function at scale, but not that it can consistently generate profits.
• The emerging evidence shows that durable value is being created not at the level of infrastructure, but through practical AI applications that solve real business problems.

The Limits of Infrastructure-Led Growth and The Rise of AI Applications and Real Demand

• The Limits of Infrastructure-Led Growth
  • The AI sector has been shaped by unprecedented investment in infrastructure, reaching hundreds of billions of dollars annually.
  • Despite this scale, foundational AI providers continue to struggle financially. Operating margins remain thin due to high inference costs and intense competition among model providers, which suppresses pricing power.
  • Even firms reporting strong revenue growth often remain unprofitable, relying on external funding to offset operational losses.
  • A further weakness of this model is the prevalence of circular spending. Much reported revenue originates within the AI ecosystem itself, particularly through discounted compute arrangements between large technology firms.
  • This dynamic inflates revenue figures while masking limited external demand, raising concerns about the sustainability of infrastructure-driven growth.
• The Rise of AI Applications and Real Demand
  • In contrast, AI applications show clear signs of genuine market traction.
  • Corporate spending on applied AI tools has grown rapidly, reflecting widespread adoption rather than experimentation.
  • These tools are increasingly embedded in daily operations across industries, driving efficiency and cost savings.
  • The commercial success of application-focused companies is evident in their recurring revenue, with multiple products reaching substantial annual income levels within a short time frame.
  • This performance demonstrates that customers are willing to pay for AI systems that deliver concrete outcomes, validating application-led business models and highlighting the limits of purely technological differentiation.

Investment Shifts and Market Validation

• Market behaviour among investors further confirms this shift. Capital is increasingly flowing toward AI firms with proven products, stable customers and clear paths to profitability.
• Strategic acquisitions now focus on application providers rather than infrastructure assets, reflecting confidence in businesses that offer immediate operational value.
• High-profile purchases of young but revenue-generating AI companies illustrate this trend.
• These deals reward speed to market, usability and customer relevance, reinforcing the idea that successful AI strategies prioritise execution over scale alone.

Departmental AI and the Concentration of Value

• The strongest concentration of AI value is found in departmental AI tools, particularly those designed for coding.
• These applications represent the largest share of departmental AI spending and enjoy exceptionally high daily usage rates among developers.
• Their success is driven by clear gains in productivity, making their value immediately measurable.
• Large technology firms have responded by acquiring application-focused startups that enhance employee efficiency and automate routine tasks.
• These transactions underline the growing consensus that AI’s economic contribution is maximised when tools are tightly aligned with specific job functions.

Applications as the Driver of Model Success

• Shifts within the enterprise AI market further support the primacy of applications.
• Changes in market share among leading models have been driven less by technical superiority and more by dominance in high-value use cases such as software development.
• This demonstrates that applications generate demand for underlying models, reversing the assumption that better models naturally lead to better businesses.
• As AI systems mature, the greatest returns accrue to companies offering integrated solutions rather than standalone model access.
• Deep integration into organisational workflows creates switching costs and long-term customer dependence, strengthening profitability over time.

Policy and Regulatory Challenges

• The expansion of AI applications raises complex policy questions.
• As large AI providers move downstream into applications, competition risks intensify, potentially disadvantaging smaller innovators.
• At the same time, AI solutions tailored to specific verticals increase exposure to legal issues around data use, privacy and accountability.
• Effective regulation must balance oversight with flexibility. Overly restrictive rules could suppress experimentation, while weak enforcement may allow dominant firms to stifle competition.
• The goal should be to preserve market openness while protecting users and maintaining trust.

Conclusion

• The evolution of AI mirrors earlier technological revolutions.
• Just as the Internet derived its value from services built on top of connectivity, AI will be monetised through applications that convert computational power into business results.
• Infrastructure enables progress, but applications deliver innovation and lasting economic impact.
• As capital markets and policymakers refocus on fundamentals, the future of AI is increasingly defined by usefulness, integration and real-world outcomes rather than scale alone.