Finally, Apple joins the "AI spending army"! But the "almost stingy" approach is highly favored by the market

According to the Zhitong Finance APP, after the quarterly financial report released by the consumer electronics giant Apple Inc. (AAPL.US) on Thursday in US Eastern Time, the most interesting topic for Wall Street analysts during the earnings conference call was undoubtedly a product called Apple Intelligence, an edge AI large model product that has not even been launched to the public, as well as Apple's capital expenditure plan on artificial intelligence development. However, from the earnings call and financial report disclosures, Apple's capital expenditure on AI can be described as extremely "stingy", with a scale far lower than tech giants like Microsoft (MSFT.US), Google (GOOGL.US), and Amazon (AMZN.US) who have spent a huge amount on purchasing NVIDIA AI GPUs. The growth rate may also be much slower, but this almost frugal expenditure may be the preferred approach in the current market.

The importance of artificial intelligence in Apple, which is centered around consumer electronics hardware such as smartphones and iPads, is increasing day by day. This has finally led Apple to embark on the long-awaited comprehensive layout of AI and various AI-related expenditure enhancement paths. However, compared to Microsoft, Apple, Google, Facebook, and Meta, the company has adopted a completely different "frugal spending model".

It is understood that compared to other tech giants, Apple's AI expenditure plan is very restrained. For example, Apple engineers choose moderately sized Google TPUs instead of expensive NVIDIA AI GPUs to train AI large models. In the future, they may also combine Apple's self-developed server AI chips to provide AI training/inference capabilities. Apple's outlook on AI expenditure does not seem aggressive, which is one of the reasons why analysts and investors are more optimistic about Apple after the earnings conference. After all, the market does not want to see the scale of AI expenditure growing larger and larger, to the extent that it significantly affects the profit margins of tech giants. This is reflected in Apple's stock price, which did not experience a significant drop like Microsoft, Google, and Amazon after announcing their earnings.

"Our comprehensive performance data shows that our spending on artificial intelligence and Apple Intelligence focused on the edge has been increasing year by year," said Apple CEO Tim Cook at the earnings conference on Thursday. Apple's third-quarter data as of June 29 showed that the company's total revenue increased by 5% year-on-year to $85.8 billion, exceeding the market's general expectation of $84.5 billion; earnings per share reached $1.40, surpassing the market's expectation of $1.35. Apple's Q3 revenue growth exceeded the expectations of many analysts, especially since this sales period is usually one of the slowest periods for Apple's revenue growth, as many consumers are waiting for the next generation iPhone to be launched in the fall.

Analysts generally believe that the upcoming iPhone edge AI system Apple Intelligence from Apple may stimulate a new round of iPhone upgrades and sales of other consumer electronics products. However, during the Q&A session with analysts in the conference call, CEO Cook and CFO Luca Maestri spent a significant amount of time avoiding questions related to the speed of Apple Intelligence's launch and whether the tech giant has seen a trend of revenue growth for this service The agreement between Apple and OpenAI to integrate ChatGPT into its software system still remains uncertain.

One key question that Cook is willing to partially answer is about the company's expenditure plan in the field of artificial intelligence servers, which is the most core infrastructure driving popular generative AI applications like ChatGPT. This has been a recurring question in the tech stock earnings season since 2023, as investors try to assess the specific progress of tech giants in artificial intelligence infrastructure development, and how much infrastructure is yet to be built, which is crucial for server AI chip leaders like NVIDIA (NVDA.US), AMD (AMD.US), Micron, and SK Hynix focusing on AI hardware.

Compared to tech giants, Apple's AI spending appears "very conservative"

Cook, who has always been reluctant to reveal too much information at earnings meetings, admitted during the meeting that costs related to artificial intelligence are rising. As described by CEO Cook, Apple's spending on overall artificial intelligence technology development and Apple Intelligence has been increasing year by year.

Apple's latest financial report for the third quarter of the 2024 fiscal year ending on June 29, revealed that its expenditure on real estate, factories, and infrastructure amounted to a high of $2.15 billion, representing an 8% increase from the previous quarter and about 3% increase from the same period last year. It is worth noting that Apple's aforementioned expenditure base was already high last year, mainly due to the expansion and upgrade of new product development, data centers, and other infrastructure, indicating that Apple's AI-related spending is increasing each quarter, albeit at a slower pace. However, some of this capital investment is not only targeted at artificial intelligence but also at other Apple businesses.

Compared to major tech peers such as Microsoft, Google, Amazon, and Meta, Apple's growth in AI-related capital expenditure appears very small. Apple's major tech peers are spending heavily on building and significantly expanding super-large-scale data centers centered around artificial intelligence, with core spending focused on investing heavily in purchasing NVIDIA AI GPUs.

For example, according to the latest statistics from FactSet, Microsoft reported a staggering capital expenditure of $13.87 billion in the quarter ending in June, a significant 55% year-on-year increase. Alphabet, Google's parent company, saw a remarkable 91% increase in expenditure, reaching $13.19 billion. Meta, the parent company of Facebook, saw a 31% increase in capital expenditure, reaching $8.3 billion in the quarter ending in June. While Apple's expenditure is relatively small compared to these tech giants, it is still significantly higher than Apple's.

However, the capital market undoubtedly favors a more conservative AI spending model like Meta and Apple, as shareholders and investors are more concerned about profit margins rather than the currently unclear long-term AI monetization prospects. Meta announced better-than-expected performance, with earnings showing that AI spending in the second quarter and the current quarter was lower than expected. There was no upward revision of the upper limit of the annual capital expenditure range, and the guidance for total expenditure for the year was not raised. The midpoint of the revenue guidance range for the third quarter exceeded market expectations, leading to a significant rebound of over 10% in the stock price after the earnings announcement, while Google and Microsoft both saw significant declines in stock prices after reporting earnings Meta CEO Mark Zuckerberg used game theory terms to explain the surge in AI spending by tech giants. He emphasized that the risk of missing out on the generative AI trend led by ChatGPT is much greater than the negative impact of spending too much on NVIDIA AI GPUs and other AI hardware, a view shared by Google's CEO - that the risk of "underinvestment" in AI is far greater than the risk of "overinvestment."

Zuckerberg even hopes to ensure that Apple does not control the next major human technological shift as completely as it controls smartphones - if it turns out to be artificial intelligence.

"In fact, I think all companies investing in AI are making rational decisions," Zuckerberg said in an interview last week. "Because the disadvantage of falling behind is that you will miss out on the most important technology in the next 10 to 15 years."

Apple is "playing a different game with AI spending"

Unlike Amazon, Google, and Microsoft, Apple is not involved in a new cloud computing business that leases AI cloud training/inference power to other tech companies. Meta also does not have this business, but the company is making massive investments in training Llama, the most powerful open-source AI language model ever launched by Meta, and using AI technology to provide core driving force for its massive ad recommendation engine and to expand ad coverage to a larger user base.

Apple, on the other hand, seems to be playing a different game with AI spending compared to tech giants, as evidenced by Apple's decision not to follow suit in purchasing NVIDIA AI GPUs on a large scale.

Apple revealed this week in a technical paper that the company has rented a relatively small and much cheaper Google TPU (Tensor Processing Unit) power system instead of deploying NVIDIA AI GPUs on a large scale like other tech giants to train large AI models. For the initial computing resources needed for Apple Intelligence and cloud AI inference power, Apple plans to use its in-house server AI chips later this year to support and cover large-scale AI inference power. In addition, Apple also rents cloud capacity from cloud service providers such as Amazon, Google, and Microsoft.

TPUs are specifically designed for deep learning and neural network inference and training optimization, especially excelling in operations like matrix multiplication, enabling more efficient execution of AI workloads. Google's TPU architecture is very simple, aimed at providing higher throughput for large-scale parallel processing tasks, so in specific AI tasks - such as AI large model inference tasks, TPUs often exhibit higher inference power performance. Compared to NVIDIA AI GPUs, TPUs are much cheaper, usually provided on the Google Cloud platform on a "pay-as-you-go" basis, allowing developers to flexibly configure hardware resources based on actual AI computing needs, avoiding large upfront capital expenditures According to the well-known Apple whistleblower Mark Gurman, Apple will deploy self-developed high-end AI chips in cloud computing servers to handle the most advanced artificial intelligence cloud computing tasks on Apple devices, while simpler AI-related functions will be processed directly on iPhone, iPad, Mac, and other end-side devices using AI large models. Apple is laying out its strategy in the generative AI field by combining "end-side + cloud computing power".

On Monday, Apple released the first test version of Apple Intelligence, a new set of artificial intelligence features that will significantly enhance Siri with the most powerful AI large models, enabling tasks such as generating and sending emails and images, as well as categorizing notifications. However, it is currently only available for developer testing.

With the support of Apple Intelligence, Apple iOS system users will be able to control all functions of a single application (App) using only voice commands, achieving comprehensive linkage and interaction to some extent. Siri's positioning may no longer be a clumsy formalized voice assistant, but rather a personalized "all-powerful AI companion" similar to the one in the movie "HER" who understands user preferences and habits very well.

As Apple builds and expands its AI infrastructure, the company may have a cost advantage over other tech giants due to its self-developed AI chips. Whether it is smartphones or consumer electronics products, or cloud server chips, Apple may not need to spend billions of dollars on third-party AI GPU hardware like NVIDIA and AMD AI GPUs.

As mentioned earlier, Apple adopts a "hybrid" approach for its AI data centers, combining end-side consumer electronics like iPhones with cloud server computing power. Some of the capital expenditures are transferred to AI partners such as OpenAI and converted into operating expenses for Apple.

During an earnings call with analysts, Cook stated, "In terms of AI-related capital expenditures, it's important to remember that we take a somewhat hybrid approach. We do things internally, and we also have some external partners, and capital expenditures will also appear in their respective businesses."

One of the partners publicly disclosed by Apple is the AI leader OpenAI, whose flagship product ChatGPT will be fully integrated into Apple Intelligence on iOS devices later this year. Under the Apple Intelligence technology framework, the globally popular AI chatbot ChatGPT will be seamlessly integrated with Apple's Siri voice assistant through an end-side approach.

However, Apple refused to discuss any details related to the OpenAI agreement on Thursday, calling it confidential. Cook, however, left open the possibility of collaborating with OpenAI to monetize opportunities.

Clearly, Apple's "frugal" scale of AI spending compared to tech giants like Microsoft and Google has been somewhat recognized by the market. After Apple's stock price rose more than 1% in after-hours trading on Thursday, pre-market trading on Friday briefly turned upwards, while the stock prices of other tech giants, except Meta, all fell significantly after announcing their financial reports After all, from the financial data of Q2 technology companies, the prospect of AI monetization is still elusive. Compared to the massive expenditure on AI infrastructure, AI revenue is not even worth mentioning. Therefore, the market hopes that tech giants can maintain strong profit margins while developing AI technology