AI is already useful, but the economics of it still don't add up

AI can be beneficial to you personally while remaining a questionable investment for the market. This is the main conflict: there are already tangible results, but they fall significantly short of the promises and do not always cover the cost of the entire infrastructure.

At the level of individual AI agents, losses and inefficiencies are evident. At the market level, there are large debts which, as the article argues, are not yet backed by corresponding productivity. Investors continue to fund Anthropic, OpenAI, and other companies, while those firms are flooding the market and already eyeing IPOs.

For the average user, the picture looks different. ChatGPT or Claude can indeed reduce cognitive load and speed up certain tasks. But there’s a huge gap between “it’s become more convenient” and “productivity has increased tenfold”: in practice, the result is more often about a 2x improvement, not the advertised 10x.

That’s precisely why the question isn’t simply whether the bubble will burst. Even if the investment frenzy ends, useful applications of AI aren’t going anywhere. However, companies whose business models rely on cheap access to expensive computing power may face a very different reality.

The limitations here are significant: the article does not provide specific financial metrics, debt figures, or the results of independent performance benchmarks. Therefore, it is impossible to say with certainty whether the bubble has already burst and which companies will survive the market cooling. For now, we can only note the gap between the personal benefits of AI, losses at the agent level, and the scale of investors’ expectations.

Have you already achieved a tenfold speedup from AI, or would it be more honest to describe the result as “roughly twice as fast” for now?

Source: All Articles / Artificial Intelligence / Habr