The media reported that the startup OpenAI is currently exploring the possibility of developing its own chips for machine intelligence.
This company, which created the world’s most popular AI-based chatbot called ChatGPT, thought about the prospects of establishing its own production of microcircuits due to the difficult situation with the supply of expensive artificial intelligence chips, which are currently a scarce commodity. Also, as part of the search for a solution to the mentioned problem, the startup is exploring the possibility of acquiring a chip manufacturing firm.
The issue of the supply of microcircuits for OpenAI in a certain sense has an existential significance since without these products the company will not be able to carry out its activities. If maintaining the current capacities of artificial intelligence is an easier task, then the development of digital thinking and worldview systems that surpass existing developments in terms of capabilities is actually impossible without full access to chips in the required volumes.
Another option for a startup against the background of a shortage of microcircuits is to deepen cooperation with manufacturers of relevant products, for example, Nvidia. Also, as part of attempts to solve the problem, such tactics as diversification of supplies beyond the production capacities of existing partners can be applied. The media, referring to insiders who spoke to journalists on the rights of anonymity, report that the startup is considering the mentioned options in order to overcome the problem of chip shortage. The company has not yet made a final decision on the leash of tactics within the complex realities of the supply of microcircuits.
The shortage of graphics processing units (GPUs), which are necessary to launch artificial intelligence apps, according to experts, is a serious problem for OpenAI, not only in the obvious sense of a lack of components for developments but also in the aspect of the fact that Nvidia currently dominates the relevant market, brand’s share in global chip production is more than 80%. This position on the complexity of the startup’s position is based on the opinion that, provided there is a monopolist, solving the supply issue requires a lot of effort.
OpenAI CEO Sam Altman, according to media reports, identified the acquisition of more artificial intelligence chips as the company’s top priority.
If a startup decides to overcome the difficulties of scarcity by establishing its own production of microcircuits, it will join a kind of elite group of technology giants such as Google and Amazon, which develop their chips for specific needs. However, the implementation of such projects involves significant financial costs. The likelihood of assistance from Microsoft, which has invested heavily in OpenAI, can only be assumed. The media does not report that the startup is counting on external assistance as part of the potential implementation of its own chip production project. At the same time, participation in such an initiative could be useful for Microsoft, since the company is interested in the development and integration of AI technologies, for which microcircuits are needed. Own or at least joint production ensures independence, which is especially important in a situation of shortage.
OpenAI could buy the chip company, similar to Amazon’s acquisition of Annapurna Labs in 2015. The media reports that the startup conducted a comprehensive check of the potential acquisition target, but there is no specific data on this.
If OpenAI decides to develop its own chips, several years will pass before the start of full-fledged activity in this direction, during which the company will be in a state of dependence on commercial suppliers, including Nvidia and Advanced Micro Devices (AMD). Creating own processors on the results of practical experience in this area turned out to be a task solved with varying success.
The media note that the growth of the startup’s needs for chips is due to two factors, among which the mentioned shortage of advanced microcircuits necessary for software, and the significant costs associated with the launch of equipment to support machine intelligence systems.
The company spends up to $700,000 a day to maintain its basic infrastructure and servers. At the end of last year, the company recorded total losses of $540 million. According to preliminary estimates, ChatGPT-4 was trained on 10,000-25,000 Nvidia A100 chips.
As we have reported earlier, OpenAI to Hold Developer Conference.