The race to join the production of Generative AI tools It has only just begun. In this spiral of competitiveness, OpenAI has decided to go one step further and announce significant improvements to its AI models, which will significantly reduce their cost and present more efficient possibilities.
At the conference held in San Francisco, Sam Altman (CEO of the entity), has announced that its objective is to cross the limits of consumers and become a reference supplier for a successful development platform. With the unconditional support of MicrosoftOpenAI has achieved revolutionize the Generative AI sector in the production of images, text and software code.
The importance of reducing costs
It is one of the biggest concerns for OpenAI partners, whose goal is build businesses based on the development and sale of AI softwarebut reducing the large implicit operating expenses.
After the launch of ChatGPT 3.5, it was publicly concluded that the cost of operating the service was excessive. In fact, Generative AI cost $700,000 in daily expenses, which translated into monthly figures would be about 20 million dollars. To these figures we would have to add the costs involved in Dall-E and other tools that consume computing resources.
The company intended reach 1,000 million euros in turnover during 2024, although it is expected that it could achieve them during the current course, although taking into account the significant operational and personnel expenses. In 2022 they registered losses of 540 million dollars.
OpenAI’s economic problems are solved thanks to the merger with Microsoftalthough the startup’s profit margins would be reduced as it would be forced to pay 75% of your profits to Microsoft until the investment was recovered. All this until reaching a figure of 92,000 million dollars.
OpenAI seeks to optimize its costs while the rest of the technological giants such as Google and amazon They continue to aim for greater investment in Generative AI. In fact, Amazon has invested $1.25 billion in Anthropic, which with Claude has become the direct rival of ChatGPT. And just as OpenAI has been integrated into Microsoft’s Azure services, Anthropic is available in AWS.
Each ChatGPT query costs about 4 cents on averageso if queries increase by a tenth of Google searches, OpenAI would need an investment of 48.1 billion dollars in GPUs and 16 billion dollars in chips per year. The vast majority of operating expenses are associated with the support infrastructure necessary to operate.
Own chip production
It is the most effective formula that OpenAI plans to achieve streamline your processes and significantly reduce costs, and even more so given the shortage of these components. In this way, you could optimize the operation of your software and gain an increase in energy efficiency.
By producing its own chips, OpenAI would eliminate the technological dependence on Nvidia and would achieve greater performance in all its operations. With a unique, natural and efficient design It would reduce the costs of servers in its data centers and improve the competitiveness of its software.
For the development of the first chips, engineers listed in the market and relationships with component suppliers, taking into account the logistics and the capital that must be invested. In addition, OpenAI would avoid geopolitical tensions between China and Taiwan with the US.
It must be taken into account that the Taiwanese company TSMC It has a delay of about 18 months in the production of AI processors, since its plants are the only ones that manufacture Nvidia’s H100 and A100, which are the ones used by ChatGPT. To the diversify chip supplier portfolioit will be possible to solve the supply problems and the high costs associated with the operation of the hardware to promote its products.
Remember, when Amazon created Inferentia for Alexia, it managed to reduce the cost of the assistant by 30% and latency by 25%.
New capabilities
By incorporating innovative vision capabilitiesOpenAI software will be able analyze images and describe themallowing developers to create applications with new uses in different disciplines ranging from entertainment to the medical sector.
There is also the possibility of that GPT-4 adjusts to the new functionalities. This would encourage companies to use OpenAI’s innovative technology to create chatbots and autonomous agents powered by AI to perform tasks automatically and without human intervention.
