Apple is working on a chip for data centers. What is known so far

Apple is working on Artificial Intelligence (AI) processors for data centers. According to an article in the Wall Street Journal, Apple’s project is called ACDC – a likely acronym for Apple Chips for Data Centers), and involves the development of chips for servers to run AI models. It appears to have been in the works for several years but has no clear timeline, according to people familiar with the matter.

The WSJ article comments that Apple has been working with Taiwan Semiconductor Manufacturing on the design and production of this chip, but it is not known to what extent these efforts have produced any positive results.

Another speculation is that this new Apple server chip will likely focus on AI inference and not on training AI models, an area already dominated by NVidia processors.

AI training relies on learning algorithms that take in large volumes of data to produce a model. It usually involves millions of parameters and requires great computational capacity. AI inference uses these trained algorithms to make predictions or classifications from new sets of data. The objective is to apply the knowledge learned in the training phase in the real world and help in making decisions or predictions. In general, inference does not require the same level of infrastructure and computational resources required in training.

Apple in the AI race

Apple’s initiatives associated with Generative AI have still been shy compared to those of its peers in the technological universe, such as OpenAI, Google, Microsoft and Meta. However, experts believe that the company will soon be able to catch up to its competitors, mainly with annual investments of around US$1 billion in Generative AI solutions, according to Bloomberg.

“Apple is a technology company with good resources and capable of competing in any area in which it intends to invest. It already has a good group of AI scientists, so the main question is how long it will take and how it will fit into its current business model – which is quite different from its competitors in Generative AI,” said Michael Kearns, professor at the UPenn School of Engineering, in an interview with the AI Business website.

For example, Apple incorporates AI features into its products, such as digital assistants, photo enhancement, automatic correction and many others, but it has not yet launched any product genuinely for Generative AI, unlike its competitors.

According to Bloomberg, Apple already has its own language modeling framework called Ajax and is working internally on a chatbot known as Apple GPT. It appears that, so far, these AI innovations have not been integrated into any of its products, but possibilities are being explored with several members of the Apple ecosystem, such as Siri, Messages and Apple Music.

Apple’s strategy for AI is expected to be presented at the company’s next Worldwide Developers Conference (WWDC) to be held in June. Tim Cook, Apple’s CEO, mentioned last February that the company is “investing significantly” in AI and hinted at a possible announcement “later this year”, causing many analysts to believe it will happen at WWDC.

Because building Generative AI models is expensive and complex, even for a company like Apple, we can envision a scenario where a few large companies create them, leaving it up to other developers to tweak those models for specific uses, Kearns explained. Will we see from Apple, then, a rich ecosystem of developers creating applications based on these few Generative AI models, in the Apple Store fashion?

Apple could lead in the AI space by improving the user experience, not just focusing on technology, as it has done in the past with cell phones and music consumption (whether it be pop or piano music), as Iliya Rybchin, partner at Elixirr Consulting, noted in an interview.

Make your bets…