news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Author: Huang Yu

Source: Hard AI

Nvidia is once again at the center of the AI ​​war.

GPU leader Nvidia had a roller coaster ride this week:It fell 7% on Tuesday, then surged 13% on Wednesday, and fell nearly 7% on Thursday. The volatility of its stock price even exceeded that of Bitcoin. Data shows that Nvidia's 30-day option implied volatility soared to 71%, while Bitcoin's DVOL index (an indicator of 30-day implied volatility) fell to 49%.

Behind the sharp fluctuations in Nvidia's stock price is the fact that the shadow of costs is looming over the technology giant and the market's concerns about the returns on huge AI investments are growing.

The bearish side represented by Wall Street investment banks and hedge funds, believing that technology giants have invested heavily in AI but have not been able to generate sufficient returns, and that the current application scenarios of AI are limited, so such a huge investment is very unwise and dangerous.

And many parties represented by technology giants such as Mag 7, believes that it is necessary to increase capital expenditure in the AI ​​field now, otherwise we will miss the coming AI era.

Wall Street, where money is paramount, places more emphasis on input/output ratio

On the one hand, Wall Street is gradually losing patience.It is believed that the capital expenditure of technology giants in the field of AI is so high, but it has not brought corresponding returns and more efficient applications.In the past two years, only two phenomenal AI products have emerged: ChatGPT and Github Copilot.

As early as the end of June, a report by Goldman Sachs pointed out that the development of AI technology may not be as rapid as expected, and its cost-effectiveness may not be as attractive as imagined.

It is predicted that in the next decade, AI will only increase US productivity by 0.5%, and its cumulative contribution to GDP growth will be only 0.9%.

Barclays also pointed out in its August report that technology giants are throwing money at AI out of "FOMO" and are afraid of missing out on AI development opportunities.

Although AI technology is still in its infancy, large companies' capital expenditure on AI has shown an irrational boom, with FOMO sentiment dominating. As this sentiment subsides, some large companies will gradually cut their AI investments next year.
Analysts expect AI toCapital expenditures will total $167 billion from 2023 to 2026, and this figure is based on optimistic expectations for demand for AI products.
In stark contrast, however, AI is expected toThe incremental revenue from cloud services was only $20 billion.

According to the mediaFridayReport,Elliott Management, a top Wall Street hedge fund, is even more radical, pointing out that large technology giants, especially Nvidia, are in a bubble and that the artificial intelligence technology that has driven their stock prices soaring is over-hyped.

Elliott believes that current artificial intelligence cannot really work efficiently and consumes a lot of energy:

So far, AI has failed to deliver the big productivity gains it promised, with few practical uses beyond summarizing meeting minutes, generating reports and helping with computer coding.
Artificial intelligence is actually software and has so far not delivered value commensurate with the hype.

Jim Tierney, head of strategy at AllianceBernstein, said: "Tech giants are increasing their bets on AI spending and creating an atmosphere of 'trust us'. However, investors are still unclear about the corresponding business models and returns. However, investors will not be reassured just by looking at the total spending and return rates."

Silicon Valley's tech giants believe AI will change the future and continue to invest in it

On the other hand, there are technology giants who withstand the pressure of doubt, continue to be optimistic about the future of AI, and choose to continue investing money.

Including Microsoft, Alphabet, Amazon and Meta, in their latest quarterly financial reports, they revealed that capital expenditures in the first six months of 2024 increased significantly - the cumulative total reached US$106 billion, and they will further increase investment in the next 18 months.

Google parent Alphabet's capital expenditure surged 90% to $25 billion in the first half of 2024.
Microsoft's capital expenditures increased 78% to $33 billion in the fourth quarter alone.
Amazon's real estate and equipment investments, including spending on its e-commerce and logistics network, surged 27% to $32.5 billion in the first half of this year, and it expects total capital spending to grow significantly in 2024.

Even as shares of Google, Microsoft and Amazon sold off immediately after their earnings reports, executives at the tech giants insisted on expanding their investments in AI.

Google CEO Sundar Pichai said it will take time for artificial intelligence products to mature and become more useful.

AI is expensive, but the risk of underinvestment is even greaterGoogle may have invested too much in AI infrastructure, primarily by buying GPUs from Nvidia. Even if the AI ​​boom slows, the data centers and computer chips the company has purchased could be put to other uses. To us, the risk of underinvesting is far greater than the risk of overinvesting.

Amazon Chief Financial Officer Brian Olsavsky said the spending will be primarily for new cloud computing infrastructure, with generative artificial intelligence now a key business for Amazon.

Meta CEO Zuckerberg also said last week that in order to ensure Meta remains at the forefront of AI, the company has spent billions of dollars to purchase Nvidia's GPUs to develop and train advanced AI models. However, he also admitted that the hype about AI may lead to excessive investment.

It is better to take the risk of increasing investment in advance than to enter the market late, because the consequence of lagging behind is that you will be at a disadvantage in the most important technology in the next 10 to 15 years. Meta's capital expenditure is expected to reach US$40 billion this year.

Zuckerberg estimated that the computing power required to train the next generation of large-scale language models will be "nearly 10 times" that of the previous version, while also admitting that it will take "years before Meta's AI chatbot can make money on its own."

Analysts at Dell'Oro Group predict that as much as $1 trillion could be invested in infrastructure such as AI data centers over the next five years.

What is the key to this “bull-bear battle” around AI?

No matter how Wall Street dislikes the tech giants' continued all-in AI approach, the key point of this tug-of-war is still focused on: how long can the tech giants' momentum of spending money last? And when will AI technology achieve greater breakthroughs in the future?

First of all, the technology giants can continue to invest money because their confidence comes from the relatively stable profits of their main businesses.

For example, Meta's share of the digital advertising market, its core business, continued to grow in the second quarter. The growth of this business mainly came from Facebook and Instagram. The advertising revenue from these two businesses increased by 22% year-on-year, causing Meta's net profit in the second quarter to increase by 73% year-on-year to US$13.47 billion.

Zuckerberg also emphasized in the earnings call thatHow AI helps advertising business grow:

AI improves recommendations, helps people find better content, and makes the advertising experience more effective, and these products are already available at scale.

In contrast, rival Google's second-quarter advertising revenue grew only 11% to $64.6 billion, but its core business remains healthy enough to bear the burden of increased spending.

Although technology giants are well-funded and willing to invest in AI, Barclays analysts predict thatAfter 2025 or later, the big players will pull back and cut (AI) capital spending plans。

Because with the advancement of technology and recent breakthroughs in smaller basic models, a large number of AI products and search functions will be able to move from the cloud to local PCs or mobile phones in the next few years. The cost of AI reasoning may be greatly reduced, and such huge capital expenditures will not be required.

As for when AI technology will achieve greater breakthroughs in the future, it seems more like an unknown.After all, after investing a lot of resources in AI, we have only produced two phenomenal AI products, ChatGPT and Github Copilot, and there is still a long way to go in the future.

Barclays predicts thatBy 2026, AI capital expenditures (approximately $167 billion) will be enough to support more than 12,000 ChatGPT-scale products, of which approximately $70 billion will be invested in training basic models, leaving approximately $95 billion for inference.

By then, many new AI-based services will emerge, and these services will promote the positive development of the market and industry.

But what is certain is that Nvidia is still the biggest winner in the AI ​​competition among technology giants, as Amazon, Google, Microsoft and Meta's AI capital expenditures have basically flowed into building data centers and purchasing Nvidia's GPU products, and this momentum may continue until next year.

Although Nvidia will not release its latest quarterly financial report until the end of August, analysts generally expect that Nvidia's data center revenue in the new quarter will reach nearly US$25 billion, equivalent to the data center revenue for the whole year of 2023.

It will soon be time to verify the quality of "AI faith".