I was able toOpen AI(Open AI) has become the talk of the town in various sectors, whether related to technology or not, thanks to its introduction of generative artificial intelligence technology capable of interacting directly with users and answering their questions.
Artificial intelligence was not the exclusive domain of OpenAI, nor was generative AI the exclusive domain of the company. Despite the emergence of many competing models to the OpenAI model,Shot at Bay(ChatGPT) created by the company, the latter’s star has risen in a way that is difficult to compete with, becoming the first ambassador for artificial intelligence technologies.
OpenAI’s success was clearly evident in the investments it was able to attract, specifically from Microsoft, as well as attracting millions of users on a daily basis, to the point that the company needed to implement waiting list systems to organize the process of users’ access to the technology, which enhanced the success of the technology and the wide public demand for it.
But this popular success does not always mean a commercial success of equal size, and this may be the case that OpenAI has reached after reports emerged indicating that the company will go bankrupt in the coming months due to the high costs of the technology.
Cost exceeds $700,000 per day
Last year, a report appeared on the Technext platform talking about the daily costs of running ChatGPT services, and according to this report, the daily cost of running the service reaches $700,000.
While the total cost of running OpenAI’s AI technology may seem very unreasonable, it reflects the reality of this sector, which currently requires deep pockets capable of spending on it, as demonstrated by Nebuly’s report on the costs of running OpenAI’s services.
The report was able to distribute the costs of operating the service across 8 different sectors, starting with the cost of research and development, which costs 4% of the total daily costs, then 21% for the process of generating images from text commands and integrating them into the programming interfaces in an easy way, then 2% for each of the cost of message storage spaces, testing and quality assurance, in addition to maintenance and updates. The company spends about 1% of the total daily costs on development operations and performance monitoring, in addition to 3% for integrating updates and compatibility with other platforms and applications.
But the biggest spending hole in OpenAI’s API is the cost of text commands and continuous text generation. Text generation costs 65% of the platform’s total daily spending, meaning that generating text and answering questions of any kind costs the company approximately $450,000 per day.
Although the report is clear and shows a lot of details, it ignores a significant part of the costs of operating artificial intelligence services as fixed costs that are only injected for the first time, which are the costs of servers and cards to operate the technology. Nvidia’s success and its arrival at the top of the trillion-dollar club indicates the costs of these cards and the profits the company makes from them.
Last June, Bloomberg published an interview with Brian Venturo, CEO of CoreWeave, an Nvidia partner in the AI card sector that specializes in building AI data centers. During the interview, Venturo talked about the cost of building AI data centers, and pointed out that these centers typically require 32,000 AI cards from Nvidia’s leading cards.
Venturo did not address the actual cost of building the centers in terms of buildings, infrastructure, and the pieces that complete the picture, but he was satisfied with referring to the number of cards, and through this number we can set an initial cost for building artificial intelligence servers, as Nvidia’s artificial intelligence cards of the “H-100” type cost between approximately $30,000 and $40,000, and when building a center consisting of 32,000 cards, the cost of one card starts at $30,000, so the final value exceeds $960 million in cards alone, without taking into account the rest of the cost, which logically exceeds $1 billion.
Continuous need for funding
Reports of the costs of building AI centers as well as running OpenAI services on a daily basis justify the company’s need for ongoing funding. In recent years, the company has raised more than $11 billion through a group of 7 funding rounds since its founding in 2015, according to a report by Tracxn, a startup tracking site.
Much of this funding is due to Microsoft’s massive support for the company, which has invested billions of dollars over the years so that OpenAI could launch ChatGPT services for the first time and make them available to everyone.
Despite this significant support, the costs of operating OpenAI operations are still so high that Microsoft is unable to provide ongoing support to the company and fears for the ultimate value of its investments in this emerging technology.
Losses up to $5 billion
A report from Munch, a website that tracks corporate news around the world, about OpenAI’s costs appeared, and the report indicated that the company may lose more than $5 billion by the end of fiscal year 2024 due to high operating costs.
The report takes into account the expected costs of expanding OpenAI’s operations through 2024, as the company is likely to need more than $7 billion to train new talent to use and develop its technologies, in addition to $1.5 billion to transfer workers between its branches and data centers. Despite this, the loss of $5 billion negatively affects the company’s final value, especially since the total funding it has received is only $11 billion.
It is worth noting that the company constantly needs to develop its technologies and work on innovating new technologies to keep pace with the rapid development in this sector. Although OpenAI was one of the first providers of artificial intelligence services to the public, it is not the largest provider of them and there are many fiercer and larger competitors than it, so the company constantly needs to develop its services to keep pace with this challenge.
Will OpenAI survive?
When looking at the news of the company’s loss and operating costs, a direct question comes to mind about the paid services provided by the company, as it offers paid services, whether from “Chat GPT” or other services that it provides to users and companies alike.
But these paid services may not generate enough profits to cover the company’s spending. The Munch report stated that OpenAI’s paid services for individuals and companies generate only $3 billion annually for the company, which is not enough to compensate for the loss the company is facing.
OpenAI is trying to find other innovative solutions to address this loss, starting from contracting with Apple to provide its services in upcoming iPhones to trying to develop its own separate search engine, but will these attempts succeed in addressing the crisis or will OpenAI, which was the reason for the spread of artificial intelligence, disappear?
