ChatgPT-5 is a failed product. And it was supposed to be [OPINIA]
Technological start-ups are governed by their own laws. The very popular model of these companies is first generating noise, then obtaining users, and in turn – an attempt to transform these users into paying customers. Investors finance the first two stages, when popularity grows and the recipient base is growing. Later, however, you need to send a simple message to users: It's time to start spending money.
In previous years, this process usually lasted several or even several years. But OpenAI is a special organization because it has a breakthrough product. He obtained its first million chatgpt users within five days, and in two months he had 100 million people using generative artificial intelligence.
Read also at Business Insider
The deadly pace has its disadvantages
If your product is gaining such enormous popularity in such a short time, of course you have a reason to celebrate. It is a clear signal from the market that you have found the need for recipients who They “vote” for your product not yet with wallets, but with their time and commitment.
Such a rapid growth of users is also a real challenge from the cost.
In the case of OPENAI, we are talking about colossal calculation bills or electricity itself, but also the salaries of top specialists developing AI. If the number of users is growing at an express pace, costs are rising just as quickly. For OpenAI, a million dollars are important as for the average average company in Poland about 10,000 PLN – it's always a amount that can be used, but it can end quite quickly. Tens of billions of dollars have already been invested in the company, but Investors can already feel considerable stress, seeing how quickly chatgpt “devours” money.
Today, approx. 700 million active users are already using CHATGPT. In April, the company reported that it has about 20 million paid subscribers, of which about 80 percent. These are owners of the cheapest plus subscription ($ 20/mc).
It cannot be said that Opeli does not earn because he earns. The thing is that he earns … too slow.
With such a express rate of users' growth and – thus – an equally express rate of expenditure growth, OpenNAI needs to generate more and more revenues very quickly. According to the data, which is referred to by Bloomberg, the company It does not assume the generation of positive cash flow before 2029. Although users' growth is amazing, the costs of chips, data centers, expenditure on research and development, or the salaries themselves are simply colossal. Revenues are growing (from around $ 3.7 billion in 2024 to around $ 10 billion in June this year), but not enough.
Read also: AI Champions League. Deepmind pays as for Neymar, and that's just the beginning
They tried, but it didn't work out
OPENAI-I assume that under the pressure of investors-it premiered ChatgPT-5 as this new, best AI tool, explaining that now you will no longer have to choose AI models from the interface. Artificial intelligence was supposed to decide what would be best for our queryproviding more intuitive service, but equally good and even better results.
It did not happen. ChatgPT-5 after the premiere provided relatively poor results, and users accused that it offers what the previous version, but to a limited extent. After the wave of complaints, OpenAI was restored by the older GPT-4O model for Plus subscribers ($ 20/mc)and also added “fast” options (fast answers) “thinking” (thoughts longer to better answer) to chatgPT-5. The head of ChatgPT Nick Turley later admitted that “the company underestimated users' attachment to the 4o model and revises the way of extinguishing models.”
In practice It was rather that ChatgPT-5 freshly offered quite limited functionality after the premiere. Why? It was an easy to guess a way to encourage users plus to buy a subscription Pro (for $ 200/mc), in which they would get a slightly better artificial intelligence than before the premiere of “Friday”, but also much more expensive.
Altman himself, the head of Opeli, also added after the premiere that the company committed “technical slip -ups”. It turned out that the mechanism that automatically selecting the working mode in ChatgPT-5 chose “dumb” models more often than it should. The only question is whether it was actually a slip -up or a cool calculation and Naive counting on the fact that users are sufficient for worse AI models, and OpenAI will be able to significantly reduce costs? The answer limits themselves were also too low and OpenAi increased them, because users complained that their favorite tool communicates too quickly, that access to advanced AI and GPT will switch to simpler (cheaper) models, unless they buy a more expensive subscription.
Some of the best engineers in the world work at Opennai. Will such a team, introducing not the first product, but already with the designation “5”, would they really offend too low response limits and a “stupid” mechanism of choosing AI models? It's hard for me to believe it.
See also: And without the internet. Openai provides two offline models
Now ads are being considered
Openai stands people who care about AI's development and user satisfaction. I am able to imagine that they were reluctant to introduce such a “underpayed” version of ChatgPT-5 and decisions on restrictions to reduce costs and clearly encourage users to buy more expensive subscriptions were taken under the pressure of investors. Those on the other hand, too You can understand because they have already spent many billion dollars and are starting to be afraidthat the return on investment will not come or will come much later than they assumed.
Everyone, however, probably agrees that the client is the most important and the client expects quality. Openai, therefore, restored (partly) the selection of models and loosened restrictions (higher limits for Thinking mode).
The problem of too high, too fast -growing costs remains unsolved. So what remains to Gigant AI? We should expect further actions around “soft” moving demanding users towards the more expensive Pro ($ 200/mc) subscription, where there will be a full selection of models and the highest response limits.
Financial pressure is gigantic. Opeli entered in 2025 with a powerful increase in revenues and valuation, but the share of paying customers remains minority, and the costs of infrastructure (from computing power to energy) explode throughout the industry. In this situation, it is natural “More for more” strategy – more expensive functions, higher limits and greater control – but in more expensive thresholds. For plus and free users, there will be restrictions, models rotation and default GPT-5.
And is One more option, which OpenAi did not even want to think about before, and today he is seriously considering it. These are ads. Nick Turley in a recent interview with The Verge admitted that “advertising is one of the options considered” and this form of earning is not excluded, although if we are to see advertising in chatgPT, they will have to be developed “with great caution and good taste”.
For now, Altman himself, commenting on the introduction of ads in chatgpt at last year's discussion at Harvard Business School, said that for him ads in AI are “in a sense extremely disturbing” and it would be treated as a “last resort”.
If users are not eager to go from $ 20 to $ 200/mc, this last resort may come even in 2026.
Author: Grzegorz Kubera, Business Insider Polska journalist
