The 'Missing Step' in the AI Industry: What Really Stands Between Hype and Profitability

Introduction: A Sobering Look Behind the Frenzy

Since 2024, global AI industry funding has repeatedly hit new highs. Tech giants are racing to pour tens of billions of dollars into computing infrastructure, and startup valuations have skyrocketed. Yet amid this seemingly unstoppable technological revolution, an uncomfortable truth is gradually surfacing — the vast majority of AI companies still haven't found a clear path to profitability.

As MIT Technology Review's AI newsletter "The Algorithm" recently pointed out, the AI industry faces a classic "Underpants Gnomes" dilemma: Step one is collecting technology (building large models), step three is profit, but that critical step two in between — how to convert technology into sustainable business value — remains perpetually unclear. This iconic metaphor from the animated series South Park now precisely describes the state of the entire AI industry.

The Core Problem: The Chasm Between Hype and Profit

The current business logic of the AI industry can be distilled into a troubling formula: massive investment + vague business plan = some future return. This logic may still hold water in capital markets, but it's becoming increasingly untenable in the real business world.

Take the large language model (LLM) space as an example. OpenAI's annualized revenue in 2024 has surpassed several billion dollars, but its operating costs are equally staggering, with losses continuing to widen. Even ChatGPT, regarded as the industry benchmark, faces severe challenges in paid user conversion and retention rates. Not to mention the wave of small and mid-sized AI companies that followed suit — most of them haven't even found their first batch of customers willing to pay on an ongoing basis.

The root of the problem is that most current AI products remain at the "impressive demo" stage rather than the "indispensable productivity tool" stage. Users may pay once for the novelty factor, but if an AI tool can't tangibly solve specific pain points and deliver quantifiable efficiency gains, renewal becomes an almost insurmountable hurdle.

In London, flyers distributed by participants in anti-AI marches, though fiercely worded, also reflected public skepticism about AI's value from another angle: Who does this technology actually serve? Does the value it creates justify the resources it consumes?

Deep Analysis: Three Structural Barriers

Barrier One: Inverted Cost Structures

While the marginal cost of AI inference continues to decline, it remains too high for most use cases. The cost of an enterprise AI assistant processing a single complex query can consume a significant portion of the service's subscription fee. This means the more users engage, the more the provider loses — the complete opposite of the traditional SaaS logic of "scale equals profit."

Barrier Two: The Value Measurement Challenge

The efficiency gains AI delivers are often implicit and diffuse, making them difficult to reflect directly on a company's financial statements. A marketing manager uses AI to help draft copy and saves two hours, but how do you quantify the value of those two hours? Enterprise procurement decision-makers need to see clear ROI, and most current AI products cannot provide convincing data.

Barrier Three: Absence of Moats

Homogeneous competition in the current AI application layer is extremely intense. Applications built on the same underlying models have very limited room for differentiation in functionality and experience. This triggers frequent price wars, further compressing already razor-thin profit margins. AI applications without technological or data moats find it very difficult to establish long-term competitive advantages.

Who Is Breaking Through First?

Despite the overall challenging landscape, some AI companies are beginning to demonstrate the outlines of sustainable business models.

Vertical domain specialists are carving out a viable path. For example, AI companies focused on specific scenarios such as medical imaging analysis, legal document review, and code generation can offer highly customized solutions because they deeply understand industry pain points, resulting in notably higher customer willingness to pay and retention rates.

Infrastructure-layer "pickaxe sellers" are the most certain beneficiaries of the current AI wave. Nvidia's financial results have amply demonstrated this. Cloud computing providers, data annotation platforms, model deployment tool providers, and other infrastructure players can generate consistent revenue regardless of how the AI application layer reshuffles.

Additionally, some traditional enterprises have achieved strong results by embedding AI capabilities into existing products and workflows rather than building standalone AI products. Microsoft integrating Copilot across the Office suite and Adobe incorporating generative AI into Creative Suite exemplify an "AI-enhanced" rather than "AI-replaced" strategy that makes it easier for users to perceive value and be willing to pay for it.

Outlook: A Return to Rationality and Value Reconstruction

History repeatedly shows that every major technological transformation goes through a cycle from frenzy to disillusionment to rational prosperity. After the dot-com bubble burst, the truly valuable companies — Google, Amazon — actually rose from the rubble. The AI industry will most likely follow a similar trajectory.

Over the next 12 to 18 months, the AI industry may undergo a significant shakeout. Companies that cannot demonstrate business value will face funding difficulties, while those that find the "missing step" will stand out. This "missing step" is not some magical business secret but rather a series of solid fundamentals: deeply understanding customer needs, meticulous cost control, and building data flywheels and network effects.

For the industry as a whole, the mental shift from "what technology can do" to "what users need" may be the key to bridging this "missing step." AI should not be a solution in search of a problem but rather a powerful tool that solves real problems. When the industry truly completes this cognitive transformation, AI's business value can move from hype to reality, from concept to profit.

As the oft-quoted yet still profound saying goes: We tend to overestimate the short-term impact of technology and underestimate its long-term impact. The key lies in who can stay clear-headed amid the short-term noise and prepare for long-term value creation.