The AI Bubble: Deflating Expectations

Many are bracing for a sudden downturn in the AI sector, reminiscent of the dot-com crash. However, this analogy is misleading. The issue lies not with the capabilities of AI models themselves, but rather with the financial frameworks supporting them, which are slowly deflating. The narrative surrounding revenue growth can be interpreted differently by 2026. During a period of generous subsidies for subscriptions, the question of profitability was often overlooked. As companies shift to a token payment model, the reality of returns is becoming clearer, revealing a stark contrast between attractive metrics and actual benefits. For instance, Uber reportedly spent its entire annual token budget in just one quarter, raising concerns about the sustainability of its model. Similarly, Meta, after encouraging token usage, has imposed limits on it. Recent reports, including one from KPMG, have been retracted due to inaccuracies, highlighting the fragility of the claims surrounding AI productivity. This situation underscores a broader collapse in funding models, as even major players like AWS admit that AI code may not enhance efficiency as previously claimed. The latest findings from NBER indicate that while the volume of code has increased, the number of functional applications has not.