- OpenAI is reportedly developing its first custom AI chip with Broadcom
- The chip could be produced as early as 2026
- This move could help reduce the cost of running OpenAI-powered apps
OpenAI is one step closer to developing its first AI chip, according to a new report – as the number of developers creating apps on its platform rises alongside cloud computing costs.
It was first reported that the ChatGPT maker was in talks with several chip designers, including Broadcom, in July. Now Reuters claims that a new hardware strategy has seen OpenAI choose Broadcom as its custom silicon partner, with the chip potentially landing in 2026.
Before then, it appears that OpenAI will add AMD chips to its Microsoft Azure system, in addition to existing ones from Nvidia. The AI giant’s plans to create a ‘foundry’ – a network of chip factories – have been scaled back, according to Reuters.
The reason for these reported steps is to help reduce the rising costs of AI-powered applications. OpenAI’s new chip will apparently not be used to train generative AI models (which is the domain of Nvidia chips), but will instead run the AI software and respond to user requests.
At today’s DevDay London event (which followed the October 1 San Francisco version), OpenAI announced some improved tools it’s using to win over developers. The largest, Real-time API, is essentially an advanced voice mode for app developers, and this API now has five new voices with improved range and expressiveness.
Currently, three million developers from around the world use OpenAI’s application programming interface (API), but the problem is that many of its features are still too expensive to use widely.
OpenAI says it has reduced the price of API tokens (in other words, how much it costs developers to use its models) by 99% since the launch of GPT-3 in June 2020, but there is still a long way to go – and this habit An AI chip could be an important step in making AI-powered apps cost-effective and truly mainstream.
OpenAI-powered apps are coming
The skyrocketing costs of cloud AI processing still pose a handbrake on apps that build OpenAI’s tools into their offerings, but some startups have already taken the plunge.
The popular online video editor Veed connects to various OpenAI models and offers features such as automated transcriptions and the ability to extract the best sound bites from long videos. Called an AI-powered notepad Muesli also uses GPT-4 and GPT-4o to transcribe meetings and send you follow-up tasks, without needing a meeting bot to join your call.
Outside of consumer apps, a startup called Tortus uses GPT-4o and OpenAI’s voting models to help doctors. The tools can listen to doctor-patient chats and automate much of the administration, such as updating health records, while also apparently improving the accuracy of diagnosis.
Beyond the potential privacy and hallucination issues of AI models, developers clearly want to harness the power of OpenAI’s tools – and there’s no doubt that low-latency voice mode has huge potential for customer service.
But while you can expect to soon be talking to one of OpenAI’s voice models when you call a store or customer service center, those AI operating costs could slow the rate of adoption – and that’s why OpenAI is seemingly keen on developing its own AI chip to develop. sooner rather than later.