AI hype vs. real business
A popular Spanish video questions whether OpenAI‑era visibility actually converts to sustainable business — it frames the problem as a question of cost structure, durable revenue, and whether customers will pay for outcomes rather than experiments. That pushback matters because investors and buyers are starting to demand unit economics and workflow embedment, not just headline model capabilities. (youtube.com)
The AI boom is entering its awkward phase. Not the phase where the models stop improving. The phase where everyone has to explain how the money works. That is the force behind a widely shared Spanish video now pushing back on the OpenAI-era story line. The question is not whether the tools are impressive. They are. The question is whether that visibility turns into a business that can survive once the novelty wears off. That means looking past demos and asking three blunt things: what it costs to serve each customer, whether revenue repeats, and whether buyers are paying for real outcomes or just funding one more experiment. This is not a fringe complaint anymore. Even OpenAI is now talking in those terms. In January, CFO Sarah Friar said the company’s model is to make revenue “scale with the value intelligence delivers,” through subscriptions, workplace plans, API usage, commerce, and eventually ads. She described the shift from curiosity to daily workflow as the core business transition. That framing matters because it quietly admits the old benchmark is over. The product is no longer judged by whether it can amaze people for five minutes. It is judged by whether it becomes infrastructure inside work. (openai.com) The problem is that infrastructure is expensive. Traditional software gets cheaper to deliver as it scales. Generative AI often does not. Each useful answer can trigger fresh inference cost, plus spending on chips, cloud capacity, retrieval systems, monitoring, and human review. That is why the industry’s financial debate has turned so quickly from model quality to margin structure. Sequoia’s now-famous “$600B question” asked where the revenue would come from to justify the scale of AI infrastructure spending. Goldman Sachs made the same point more bluntly, arguing that the broader ecosystem was on track for enormous capital expenditure long before there was matching evidence of economic payoff. (sequoiacap.com) OpenAI’s own numbers show both sides of that tension. The company said in January that annualized revenue had surpassed $20 billion in 2025, up from $6 billion in 2024. That is extraordinary growth by any standard. It also does not settle the underlying question. Fast revenue growth proves demand exists. It does not prove the margins are durable, or that customers will keep paying once procurement teams start measuring cost against hard business results instead of strategic fear of missing out. (openai.com) And that is exactly where big companies are getting tougher. Deloitte’s 2026 enterprise AI survey found that organizations are moving from experimentation toward scaling, but only 25% had moved 40% or more of their AI pilots into production. Just 30% were redesigning key processes around AI. A striking 37% said they were still using AI only at a surface level, with little or no change to the underlying business process. In other words, many companies bought access to AI before they figured out where it actually belongs. (deloitte.com) That gap between access and embedment is the whole story. Deloitte’s earlier generative AI work found that most companies expected less than 30% of their experiments to reach full scale within six months, even as 78% planned to increase investment. Nearly all organizations with scaled deployments reported measurable returns, but the returns showed up where AI was tied to a real function like IT or cybersecurity, not where it floated as a general-purpose assistant in search of a job. The winners are not the firms with the flashiest model screenshots. They are the ones that can attach AI to a workflow, remove labor or time, and charge for that result in a way the customer can defend in a budget meeting. (deloitte.com) That is why the Spanish critique lands now. It is arriving at the exact moment the market has stopped rewarding AI for being visible and started asking whether it is indispensable. The test is no longer whether a company can get attention by saying “we use OpenAI.” The test is whether a buyer will renew after six months because some messy internal process now breaks without it. Only a quarter of surveyed companies say AI is already having a transformative effect on their business, and that small number is more revealing than the hype ever was. (deloitte.com)
