编译 | Alyssa
Edit | Panken
Zhidong reported on May 7 that according to a report by the Wall Street Journal today, Apple is developing and designing chips for running artificial intelligence (AI) software for data center servers, a move that may give Apple an advantage in the increasingly fierce arms race in the field of AI.
Over the past decade, Apple has become an industry-leading chip designer by designing high-performance chips for iPhones, iPads, Apple Watches and Macs. According to people familiar with the matter, the internal code name of this server chip project is ACDC (Apple Chips in Data Center, Apple Data Center Chips).
Although the "ACDC project" has been quietly moving forward for several years, there is still uncertainty about when the new chips will be released and even if they will be available. Apple is expected to showcase a number of new AI products and make announcements at the Worldwide Developers Conference (WWDC) in June.
An Apple spokesperson declined to comment for the report.
Recently, Apple has worked closely with TSMC, the world's largest chipmaker, to design and start the production of a new generation of AI server chips, but there is still uncertainty about whether the chip has achieved substantial results.
However, for this rumor, a well-known Apple whistleblower, Bloomberg reporter Mark Gurman posted on social platforms that Apple launched a project to develop server chips in 2018, but that project was canceled, considering the cost, lack of differentiation, and the equipment nature of its large language model, Gurman does not believe that Apple will develop its own AI server chip.
1. Cook promised to increase AI investment, and Apple's AI chips are about to come out
According to the Wall Street Journal, citing people familiar with the matter, Apple's server chip is expected to focus on running AI models — that is, inference tasks — rather than training AI models.
Competitors such as Microsoft and Meta have poured billions of dollars into OpenAI's ChatGPT service and moved quickly to reorient their businesses around so-called generative AI, and Apple's server chips may have come just in time.
In the face of industry change, Apple's investors are unhappy with the company's lack of progress in the field of AI. Other tech companies have seen their stock prices climb with high-profile announcements in the AI space. Microsoft's stock price rose most significantly, thanks to its close early partnership with OpenAI. Earlier this year, Microsoft overtook Apple to become the world's most valuable company and continues to extend its lead.
Apple CEO Tim Cook tried to reassure investors that Apple is investing more in AI technology and heralded major AI-related announcements to come.
▲Apple CEO Tim Cook has repeatedly assured investors that the company is increasing its investment in AI technology. (Source: AFP)
"We strongly believe in the transformative power and potential of AI, and we believe we have the strengths to differentiate ourselves in this new era, including Apple's unique ability to integrate hardware, software, and services." Cook mentioned it on a recent quarterly earnings call.
2. Tech giants are racing to develop AI server chips, aiming to get rid of Nvidia's dependence
Behind the large language models represented by ChatGPT, generative AI systems often need to process massive amounts of data with the help of computer servers in data centers. Most large tech companies rely heavily on servers filled with Nvidia chips, and Nvidia is at the heart of the "AI gold rush". It is estimated that Nvidia has a market share of more than 80% in this field.
▲An Nvidia chip on display at a science and technology conference in March. (Source: Bloomberg)
In order to get rid of their dependence on Nvidia, most big tech companies have either developed their own AI server chips or are exploring this possibility. Google probably has the most mature in-house AI server-on-chip product, the Tensor Processing Unit (TPU), which has been in operation since 2013. OpenAI is also in discussions to fund an ambitious in-house chip project.
With its existing chip design prowess, Apple may be able to find a key advantage over other tech giants in the AI space. Among its competitors, Apple has the longest history and the most success in designing custom chips, and first began integrating its own processor chips into iPhones and iPads in 2010.
Since 2017, Apple has focused on integrating AI's processors into its mobile chips. By 2020, Apple will further promote its self-developed chip strategy, and its Mac product line will gradually replace Intel chips and use its own custom chips. This series of chip innovations has significantly improved the market competitiveness of Apple's products, making it a technological gap with its competitors.
It is reported that the high-end version of Apple's M-series chips, which are currently designed for consumer-level products, is expected to expand functions in the future to achieve AI tasks such as server-side inference.
Conclusion: After the end-side AI chip is blown up, who will be the next Apple's self-developed core?
Despite Apple's remarkable success in developing device processors and actively exploring the field of server chips, Apple has been challenged to develop cellular wireless chips that could replace mobile chip giant Qualcomm, and the launch of its internal cellular modem project has been repeatedly delayed.
Apple prefers that most AI applications run directly on the iPhone or Apple Watch, but there are still some processing processes that need to be completed by remote servers, and the role of Apple's own server chips is particularly critical at this time. By taking on more data center tasks through its own chips, Apple will not only be able to strengthen its data processing capabilities, but also better control its own "AI destiny".
Source: Wall Street Journal