news

AI Future Guide - Class Representative Series: The fastest and most comprehensive interpretation of major events in the AI ​​field. This article digs deep into the artificial intelligence bills disclosed in the financial reports of the seven technology giants and discusses whether the generative AI field has reached the critical point of bubble bursting.

Tencent Technology Author Guo Xiaojing

Edited by Su Yang

The share prices of Nvidia, Meta, Tesla, Amazon, Google's parent company Alphabet, Microsoft and Apple rose by 239%, 194%, 102%, 81%, 59%, 57% and 48% respectively in 2023, earning them the name "Magnificent Seven" in the market. During the same period, the S&P 500 index only rose by 24% overall.

On July 24, 2024, it seemed that this vigorous rise had come to an end. The S&P 500 index fell by more than 2%, and the Nasdaq Composite Index fell by more than 3.6%. Both indexes recorded their largest single-day declines since the end of 2022. The panic moment of the U.S. stock market coincided with the release of the quarterly reports of Google's parent company Alphabet and Tesla.

In the following trading days, Microsoft, Meta, and Apple also released their financial reports, and the stock market fluctuated violently. Although the reasons for the fluctuations may be diverse, such as the expectation of interest rate cuts, the release of employment data, and the criticized impossible triangle - the US dollar index, US bond yields, and US technology stocks have always maintained three highs that seem to violate financial common sense.

However, it is undeniable that one of the main threads "involving" market sentiment is still a huge disagreement - whether the huge AI investment by the leading American technology companies is an "investment in the future" or a "bill" that shareholders have to pay for.

The biggest driving force behind this wave of growth is the market'sGenerative AIIrene Tunkel, chief U.S. equity strategist at BCA Research (a global economic analysis company), commented that the main reason for the outstanding stock price performance of the "Big Seven" in 2023, except for Nvidia, is the multiple expansion. This shows that investors have high expectations for the future earnings growth of these companies.

As earnings season results come in, those lofty growth expectations are starting to swing wildly.

The most special one, Nvidia, which is indeed supported by explosive performance, has seen its market value fall by more than 22% compared to June, evaporating 5.2 trillion yuan, and has fallen by 20% in 17 consecutive trading days recently. In addition to the negative information about the mass production of Blackwell chips, the capital market has also begun to worry that if technology giants cannot prove that AI can bring enough growth to their business, they will not be able to persist in their investment in the field of AI, and Nvidia's performance will not continue to exceed expectations again and again.

Expectations are always the biggest driving force behind stock prices and are far more important than historical performance.

The capital market seems to be split into two equally powerful forces. On the left, there is a strong vision of artificial intelligence changing the world and a steady stream of capital investment; on the right, there is a deep suspicion of the input-output ratio of artificial intelligence and the huge bubble that artificial intelligence is creating:

1. Will the giants’ huge investments in generative artificial intelligence make short-term financial reports look ugly?

2. Can such a huge investment really bring growth? When can it be realized?

3. If generative AI is a distant future, is the bubble getting bigger and bigger?

4. Why are the giants so determined to bet big on generative artificial intelligence?

After digging deep into the AI ​​bills of tech giants, we realize that the future may be far more complicated than we imagined.

1. Technology giants tighten their belts to make way for AI investment

Based on the just-released quarterly reports, this article sorts out and lists the capital expenditures of major technology companies and extracts the description of their investment in artificial intelligence:

1、 Microsoft: Capital expenditures in the last quarter were $13.9 billion, mainly used for artificial intelligence, up 55% from the same period last year; including financial leasing, total capital expenditures were $19 billion, up from $10.7 billion in the same period last year.

2、Google's parent company Alphabet: Capital expenditures will reach or exceed $12 billion per quarter in the second half of the year, and total annual spending may exceed $49 billion, 84% higher than the average annual spending over the past five years.

3、 Meta: Capital expenditures for the quarter were $8.47 billion, up nearly 33.4% from the same period last year; the minimum capital expenditure forecast for 2024 was raised from $35 billion to at least $37 billion, but the maximum spending forecast of $40 billion was maintained unchanged.

4、 Amazon: Capital expenditure will accelerate in the second half of 2024, higher than the US$30.5 billion in H1, mainly used for AWS infrastructure construction.

5、apple: Apple Chief Financial Officer Luca Maestri did not give a specific figure for capital expenditure in the second quarter of fiscal 2024 during the earnings call.

When asked by analysts whether Apple's gradual shift of focus toward artificial intelligence and generative AI would affect the company's capital expenditure rhythm, its Chief Financial Officer Luca Maestri said that Apple has been working hard to promote innovation in all its businesses and fields for many years. In the past five years alone, Apple's research and development expenditures in related fields have exceeded US$100 billion.

6、 Tesla: No specific data on AI has been released. Musk only revealed in a conference call that capital expenditures in 2024 may reach $10 billion. In January 2024, Tesla announced an additional investment of $500 million to purchase about 10,000 H100 GPUs from Nvidia; its CEO Musk also posted on social media X that Tesla may spend $3 billion to $4 billion this year (2024) to purchase Nvidia chip hardware.

Judging from these figures, each giant invests more than 10 billion US dollars in AI every year. It is only an estimate how much these giants will increase their capital expenditures for artificial intelligence by the end of 2024. Recently, Barclays analysts pointed out in a report,Capital expenditures in the AI ​​sector are expected to reach a cumulative $167 billion from 2023 to 2026, a figure based on optimistic expectations for demand for AI products.However, according to the capital expenditures announced by the giants listed above, this figure is not groundless. However, in sharp contrast,It is estimated that by 2026, the incremental revenue from cloud services will be only US$20 billion.

Using only the incremental revenue from cloud services as a comparison may not objectively explain the problem. However, it can indirectly reflect that with such huge investments, the giants may still not be able to answer the ROI question of AI-related investments well, at least until 2026.

However, this does not prevent the giants from cutting budgets in other areas, or even laying off employees, and continuing to invest firmly in the field of artificial intelligence. There is no result in the short term, and as a listed company, it also faces huge pressure from the capital market. Why do the giants do this?

“We are clearly in the early innings of a very transformative space,” said Sundar Pichai, CEO of Google parent Alphabet. “The risk for us to underinvest is far greater than the risk of overinvesting,” he added, not to mention that tech rivals Microsoft, Amazon and Meta Platforms are all investing record capital expenditures in the same space.

Meta CEO Zuckerberg also expressed a stronger view, "At this point, I would rather take the risk of building capabilities before I need them than when it's too late because launching new reasoning projects takes a long time to prepare."

“This is a high-risk business,” said Brian Olsavsky, Amazon’s chief financial officer. “This is a revolutionary shift in many industries. We think we cancloud computingWe are taking advantage of our existing position in the field and engaging with it in a very high-end way.”

CEOs see the risks of huge investments, but they still invest firmly.It seems that this is not a "big gamble", but a ticket to the "Noah's Ark" that must be bought.

2. Can AI really bring incremental growth to technology giants?

Why is this? What incremental benefits can generative AI bring? Currently, we have not seen any clear new business models emerge. The stories told by technology giants are mainlyCloud services, advertising,Autopilot, and edge intelligence。

The business and payment system of cloud services themselves are complex. The main logic of the increase comes from an assumption that generative AI will make more companies want to use generative AI, and generative AI requires a lot of computing and storage resources. Whether it is deploying private models or using existing large models, cloud service providers cannot be avoided, which will bring incremental new customers to cloud services.

Microsoft, Amazon, and Google are known as the “three clouds” of the United States. Microsoft can be said to be a technology giant that has taken the lead in this wave of generative AI. It has invested in startups that have ignited this wave of generative AI.OpenAI, has invested $13 billion in OpenAI. Microsoft is OpenAI's exclusive cloud provider and uses OpenAI's models in commercial customers and consumer products.

OpenAI's big model is recognized as the most powerful closed-source big model at present. But even with the strong combination, the incremental growth brought to Microsoft's cloud business is limited. According to the latest financial report, artificial intelligence services contributed 8 percentage points to Azure's revenue growth this quarter, while the contribution in the previous quarter was 7 percentage points. Growth seems to be getting slower and slower.

In Q2 2024, Amazon AWS's revenue was $26.281 billion, up 18.7% year-on-year, slightly better than expected, but net profit growth slowed slightly, and AWS's operating profit margin fell by 0.6% month-on-month. Although the difference is not much from expectations, the market reaction is still negative. The slowdown in growth, poor performance of other businesses (such as e-commerce), and the need to invest heavily in the field of AI have made investors worried.

Google's parent company Alphabet's financial report performed well in all aspects, with total revenue of $84.7 billion, a year-on-year increase of 14%; net profit of $23.6 billion, both higher than analysts' expectations. Although the cloud business is not the bulk of Google's revenue, its quarterly revenue exceeded $10 billion for the first time, reaching $10.347 billion, a year-on-year increase of 29%.

However, the subsequent stock price performance showed the investors' entanglement. After the market, it first rose by about 2%, and then fell by 2.18%. As of the close of July 25, it fell by 2.99% to $169.16 per share. The main reason is that Alphabet's capital expenditure is higher than market expectations, almost twice that of the same period last year, and this huge investment will continue. The increase brought by AI cannot offset the panic of continued huge capital investment.

Judging from the three clouds in the United States,AI has indeed brought incremental growth to the cloud business, but the growth is lower than market expectations.

In the advertising business, the number of active users is the fundamental of the market. AI helps improve the accuracy of recommendations and the creativity of advertisements, so that target users can see and be willing to actively click to open them. This is the most ideal result that advertisers hope to achieve, and these two points are indeed what generative AI is good at.

Meta's advertising business grew 22%, and its family of apps (Facebook, Instagram, WhatsApp and Messenger) had 3.27 billion daily active users in June, up 7% from the same period last year. Facebook's monthly active users exceeded the 3 billion mark for the first time in history, reaching 3.03 billion, up 3.4% year-on-year. Zuckerberg is also confident that MetaAI will become the world's most widely used artificial intelligence assistant by the end of this year.

In this wave of generative AI, Meta's Llama series of models has successfully become the leader of the global large language model open source camp, and its position is becoming more and more stable. Meta has just released the Llama 3.1 series.But according to Zuckerberg, the training resources required for Llama 4 are ten times that of Llama 3.

Google's advertising business is also growing well. In the second quarter of 2024, Google's advertising revenue was $64.616 billion, an increase of 11% year-on-year. Among them, Google search and other advertising revenue increased by 14% year-on-year; YouTube advertising revenue increased by 13% year-on-year. The performance of Google's large model Gemini is also ranked among the top among all models.

However, even the giants themselves may not be able to fully explain whether this growth is truly driven by AI.

Tesla and Apple are relatively unique.

Tesla is a representative of the radicals. Founder Musk has made comprehensive arrangements in the most cutting-edge fields including autonomous driving, brain-computer interface, aerospace, etc., each of which requires a large amount of capital investment. However, he is always able to convince the world to believe his story, obtain financing, and create a part of the cash flow in these "money-burning" fields to support the continued exploration of the grand narrative of the future.

Tesla is a huge part of the whole picture. In addition to looking at delivery volume like analyzing other car companies, investors will also pay more attention to Musk's narrative about the future. Robotaxi and the humanoid robot Optimus are Tesla's "AI story".FSD(Full-self driving) technology is one of the core capabilities of Tesla's autonomous driving, Robotaxi, and humanoid robots, and the supercomputer Dojo is the brain that supports all of this.

Tesla has been increasing its investment in this entire AI narrative, and Musk has also posted on social media X that Tesla may spend $3 billion to $4 billion to purchase Nvidia chip hardware this year (2024).

Investors are more tolerant of Tesla's increased investment. According to a report by Goldman Sachs, 68% of investors view AI as the main driver of Tesla's stock price in the next year, while only 33% prefer electric vehicles.

Compared with Tesla, Apple seems to have always been somewhat "conservative" in its attitude towards generative AI. There has been no "sharp increase" in large capital investment, and the CFO's external statements have all been that the investment has been continuous for "the past five years." At the WWDC in June this year, the release of Apple Intelligence made the outside world focus on the new iPhone to be released this fall. Can this phone with AI (Apple Intelligence) "break through the innovation of squeezing toothpaste" and create a "new paradigm of AI mobile phones" to stimulate the sales of smartphones, which are under great pressure.

For generative AI, the seven giants have their own designs, which can be summarized as mentioned at the beginning of this paragraph: cloud services, advertising, smart terminals. Even Musk, who has the biggest imagination, has not created a new way to play. Generative AI is more like a more powerful brain, an infrastructure that the giants have to compete to invest in. So the biggest doubt from the outside world is that, apart from writing crappy articles and drawing weird pictures, I really don’t understand what else generative AI can do.

Barclays' research report calls this arms race-style investment "FOMO (fear of missing out)". However, imagine that if all competitors have upgraded their business foundation to "computers" and you are still using "abacuses", it may not be a matter of missing out, but of disappearing directly. For example, cloud customers will directly choose cloud service providers that can provide generative AI capabilities.

What the giants are competing to lay out is actually the infrastructure of the future, rather than disruptive innovative applications that directly face users. This includes the underlying computing power infrastructure, as well as a powerful self-developed basic large model.All of these require large capital expenditures.

3. Is the bubble of generative AI gathering?

Since it is infrastructure, the time difference between investment and return is bound to be long. This wave of technology stock rise driven by generative AI reflects everyone's excitement about the birth of new technologies. However, no giant can yet say what the incremental value of generative AI is and how much incremental value it brings.

In addition to their competitive relationship, the seven giants have formed a network of relationships in the field of generative AI. Except for Apple, the other giants are all purchasing chips from Nvidia; Apple's first large model for external cooperation is OpenAI's large model, and Microsoft is OpenAI's largest investor and exclusive cloud service provider; Apple has not disclosed whether it has used Nvidia's computing power, but it has clearly stated in the external documents of Apple Intelligence that it has used Google's TPU.

Beyond the giants, there is the ambitious and even more fanatical Silicon Valley.

For OpenAI, by the end of 2023: annualized revenue is $1.6 billion. In June 2024: The Information estimates that annualized revenue in 2024 will reach $3.5 billion to $4.5 billion. Although it is expected to still suffer a substantial loss of around $4 billion, the revenue scale is indeed growing at a maximum rate of three times. Its valuation has also tripled, with a valuation of approximately $29 billion in April 2023. In February 2024: the valuation reaches $86 billion.

At this stage, compared with profitability, capital values ​​OpenAI's ability to capture territories and seize the market more.From this perspective, it seems that OpenAI's valuation growth is relatively rational.

According to data from The Information, the total amount of funds raised by generative AI companies in the second quarter reached US$12.2 billion, continuing to break historical records, and the number of companies receiving financing also continued to set a new high, reaching 55.

Such a hot atmosphere, soaring valuations and unclear business models have caused institutions including Goldman Sachs to begin to question whether the bubble of generative AI has been inflated.

"I think the natural comparison that comes to mind for many investors is the telecom bubble of the late 1990s and early 2000s," said Michael Hodel, an analyst at Morningstar. "Most of the companies involved in that expansion went bankrupt. This expansion seems similar in some ways...but the main difference is that the companies involved in most of this expansion have highly profitable existing businesses and solid balance sheets."

If the definition of a bubble here is "a continuous increase in pure P/E ratio or valuation without profit fundamentals", it seems that a bubble has not yet formed. The seven giants, including Nvidia, have solid ability to create profits. And the investment of the giants is not very aggressive compared to the cash flow they generate.

However, if the macro fundamentals deteriorate, the earnings expectations of the giants are lowered, and capital expenditures cannot be reduced quickly, and the chips, servers, and data centers purchased with previous capital investment cannot be immediately converted into profits, then the price-earnings ratio will naturally soar, and the so-called bubble will expand.

The definition of a bubble is dynamic. The factors that drive the stock market up and down are very complex. The collapse of expected consistency or the deterioration of macro fundamentals can cause drastic fluctuations in the stock market. After a continuous rise, any one event can become a fuse for a downward trough.

The capital expenditure decisions of CEOs of US stock giants may be based on the core competitiveness of the company in the next ten to twenty years; Silicon Valley investors may look at the probability of new giants growing in the next thirty years; and the rise and fall of the stock market is the result of short-term complex market factors and emotions. These three things are related, but to a greater extent they are three independent things.

However, there are several key issues that the core participants of this generative AI revolution may need to seriously consider together:

• Generative AI consumes a lot of resources. How can we make it more efficient?

• The cost of computing power is huge, how to make it cheaper and cheaper?

• Is there any comparisonTransformerWill better architectures, or even new technologies that subvert the current deep learning routes, make artificial intelligence as efficient as the human brain?

When these problems are gradually solved, we will see the arrival of the artificial intelligence singularity, and looking back at the current investment, perhaps everything is worth it.