news

7moon15On the 2nd, Morgan Stanley released its latest research report.discussedMicrosoftof“AIMonetization”Question.thinkThe market is optimistic about Microsoft“AIMonetization”Concerns about the company's stock price,existMicrosoft's stock price has lagged behind its peers over the past three monthsscience and technologyStocks and the market, reflecting Microsoft's“AIMonetization”The medium-term outlook is underestimated.

Morgan Stanley's report predicts that Microsoft's total capital expenditure will almost double from $32 billion in fiscal year 23 to $63 billion in fiscal year 25. However, AI revenue will also increase from $5.8-9.6 billion in fiscal year 24 to $46.5-77.4 billion in fiscal year 27.

Therefore, Morgan Stanley is confident that core IT spending will drive growth in commercial returns from Microsoft's AI business.

But some brokerages are not so confident about the monetization prospects of AI. Recently, the team of Goldman Sachs strategist Ryan Hammond reported that Internet giants including Amazon, Meta, Microsoft and Google have spent about $357 billion on capital expenditures and research and development in the past year, and a "large part" of these expenditures have been used for artificial intelligence. But these super-large companies will eventually be required to prove that "their investments can generate revenue and profits. If there is no sign of profitability, it may lead to a depreciation of valuations."

According to technology media The Information, although Microsoft has a huge customer base with its Office 365 software and claims that 60% of Fortune 500 companies are paying for the intelligent assistant Copilot service, this market advantage has not yet been reflected in the company's performance report. "In fact, from the fourth quarter of 2023 to the first quarter of 2024, the growth rate of corporate sales of Office applications slowed by 2 percentage points. Even optimistic analysts believe that Microsoft will only make about $10 billion from AI this year."

Similar to The Information's view, a well-known business magazine concluded some time ago that AI technology has produced almost no economic benefits so far. The article pointed out that the five major technology giants, Google's parent company Alphabet, Amazon, Apple, Meta and Microsoft, are expected to invest about $400 billion in AI capital expenditures in 2024, which has led investors to have optimistic expectations for the future earnings of these companies, and the market value of these five giants has increased by $2 trillion. According to estimates, these technology giants still have a long way to go to achieve large-scale revenue in the field of AI.

Recently, Bill Gates complained in a podcast:“It is unprecedented that so much capital has poured into a new field. From the perspective of market value and valuation, the entire AI market has entered a frenzy, the extent of which dwarfs the frenzy during the Internet and automobile eras in history.”

So, this waveAIIs there a bubble in the wave?

01

In the past two years, Wall Street has been hyping up the concept of the "Magnificent Seven", which deeply bundles together the seven hottest and largest technology stocks in the current basic market - Apple, Microsoft, Alphabet, Amazon, Nvidia, Tesla, and Meta.

The performance of the "Big Seven" has also become a barometer of the overall trend of U.S. stocks. In the S&P 500 index, the weight of the "Big Seven" has reached 27.9%.

Especially this year, the capital market's confidence in AI has pushed the US stock market to hit new highs. The stock prices of the "Big Seven" have soared by 40% at least, and have even doubled at most, pushing the S&P 500 index up by more than 15%. Some analysts believe that the market's enthusiasm for AI may lead the S&P 500 index to a peak of 7,000 points in 2025.

However, the overly strong growth curve also reminded some market veterans of the "Internet bubble" at the beginning of the millennium. Amid the cheers, Goldman Sachs became the company that stepped forward to pour cold water.

“The current hyperscalers will eventually be asked to demonstrate that their investments will generate revenue and profits.”

AI is undoubtedly a "money-burning war". The high R&D and training costs of large models are so high that even the wealthy tech giants have to grit their teeth before they can afford it. Meta executives said that Meta has spent $30 billion on GPUs, which is more than the US Apollo moon landing program. Of course, these two figures are sixty years apart, and inflation must be taken into account.

At the second quarter earnings conference, Meta raised its full-year capital expenditure for 2024 to between $35 billion and $40 billion, higher than the previous range of $30 billion to $37 billion.

According to statistics from the Goldman Sachs team, Amazon, Meta, Microsoft, and Google invested a total of US$357 billion in capital expenditures and research and development in the past year, of which a "large portion" was spent on AI research and development, accounting for nearly a quarter of the total capital expenditures and research and development expenditures of the S&P 500 index.

Companies are afraid of being left behind in the AI ​​trend and are willing to do their best to try their best to take a chance on the future. It is estimated that Silicon Valley will spend $1 trillion on AI capital expenditures in the next few years. However, there are almost no substantial results to prove that the investment is worthwhile. Even Microsoft, which is at the forefront of the industry, will only earn at most $10 billion from AI this year, far less than the investment.

Goldman Sachs warns that the AI ​​industry should be wary of "overinvestment". Goldman Sachs has constructed a stock market index that tracks companies that are expected to have the greatest potential for profit through improved productivity through AI. However, since the end of 2022, the stock prices of these companies have not outperformed the S&P 500, indicating that investors do not see the prospect of additional profits.

Judging from the current performance, only NVIDIA has really made real money from the AI ​​explosion. GPUs are in short supply, and monetization does not require any pre-sets. It is all real.Stock prices change every day. Goldman Sachs predicts that after Nvidia, companies that provide AI infrastructure will widely absorb this round of hot money, including but not limited to the semiconductor industry, data centers, and cloud service providers. At the same time, the controversial field of AI security will also see the emergence of hot companies.

When the infrastructure is fully laid out, AI can truly reach the general public, and IT service companies that generate incremental value through AI will become the next round of beneficiaries. When AI is widely used in all walks of life and productivity is significantly improved, companies with greater profit growth potential will emerge in the sub-sectors.

Currently, many companies have entered the AI ​​infrastructure market, trying to catch the overwhelming wealth. The AI ​​infrastructure market has become very crowded, and the "Big Seven" are both the first party and the second party in the business, hoping that their input-output ratio will exceed that of their peers and get a bigger share of the pie.

02

How much change can AI bring to human life? This may be a question that everyone is thinking about.

Unlike Ryan Hammond's team, economists at Goldman Sachs also expect generative AI to increase productivity by 9% and GDP by 6.1%.

But in reality, the actual penetration rate of artificial intelligence is indeed not as good as the technology industry has advertised. Recently, a report from the U.S. Census Bureau showed that only 5% of American companies are using AI, and it is expected to rise to 6.6% in the third quarter.

On the negative side, AI has not yet improved productivity.

At this stage, the intelligence level presented by AI cannot replace real people in most scenarios. Imagine that at a train station, an AI information counter and a staff member are in front of passengers at the same time. Most passengers will still choose to ask the staff to solve their problems. The staff can judge the passenger's problem and the passenger's overall status, while AI can only answer the question itself.

The present is the most important thing. The sensory recognition and comprehension abilities of a normal adult are beyond the reach of AI. People can make real-time judgments and actions based on the situation in front of them. In contrast, the responses given by AI are based on historical data, which is preset and lacks the ability to react instantly to ongoing events.

To some extent, in order to promote AI, technology giants have deliberately ignored the complexity of human daily communication behaviors. According to US media reports, companies such as Walmart and McDonald's that have tried to introduce AI to assist in improving service quality have expressed dissatisfaction with AI's performance, and McDonald's has also terminated its "AI ordering" cooperation with IBM.

For individual users, AI services are still at the entertainment stage, and users' willingness to purchase is not high.

After ChatGPT came out, optimistic people believed that AI would liberate humans from tedious tasks, comprehensively improve the productivity of physical and mental labor, and allow people to do more of what they want to do in a limited time.

However, as AI application scenarios become more specific, AI has received less and less public support. Practitioners from all walks of life do not think that AI is a tool for self-achievement, but rather a monster that will "steal" their jobs.

Many scholars have come forward to question whether the $1 trillion cost of AI investment can be realized in terms of productivity.

At present, generative AI competes with intellectual workers in terms of efficiency and is unlikely to play a role in production processes involving spatial scope or physical labor. Some scholars believe that only 23% of production tasks can be automated through the AI ​​economy in the next 10 years, saving an average of about 27% in labor costs. Daron Acemoglu, a professor at MIT, estimates that generative AI will only increase economic productivity by about 0.5% and GDP by about 1%.

If the benefits of AI are limited to improving efficiency and not opening up new production activities, then multiple expansions cannot be achieved and it becomes internal consumption. In other words, using very expensive technology to replace more cost-effective manpower is neither in line with business logic nor in line with the needs of government public management.

Goldman Sachs estimates that AI will put 7% of the workforce at risk of complete unemployment. Once the cost of AI's popularization is the intensification of conflicts between people, there will be strong intervention.

But looking at the bright side, the impact of artificial intelligence on the economy is still far from coming, and there is still a lot of room for improvement. In fact, no one really knows how it will reshape the economy and the job market.

03

What we can be sure of is that the arms race in artificial intelligence will not be stopped anytime soon.

According to statistics, on average, large-scale enterprises can convert 31% of their capital and R&D expenditures into revenue within three years. Based on this calculation, with an investment of $1 trillion, the AI ​​field needs to generate $310 billion in revenue to keep up with the past.

As the company with the most stable business model in the AI ​​boom, Nvidia's data center business will have revenue of US$47.5 billion in 2023, but its full-year revenue is not enough to cover the costs incurred by Amazon.

At present, the delivery cycle of most of Nvidia's chip products has been shortened from one year during the hottest period to two months. The number of chips hoarded by technology giants is basically enough to support the operation of large models. Since the beginning of this year, GPU prices have shown a significant correction. Once technology giants significantly reduce their purchases, Nvidia's AI business model will have to be re-laid out.

Rapidly rising, expectations were full, and then quickly disappointed. From upstream to downstream, everyone seems to be stuck in the quagmire of high AI costs and no profit in sight, looking for a way out.

Achieving technological breakthroughs in AI software and hardware is the current focus.

The high energy cost of large model training has caused many technology companies to focus on the field of nuclear power generation and enter the energy market. In addition, some companies are exploring GPU resource sharing models to reduce resource idleness, reduce the construction cost of data centers, and lower the entry threshold for small and medium-sized enterprises. Since the beginning of this year, many large models at home and abroad have announced price cuts, and AI large models are setting off a "price war".

However, the monetization explorations mentioned above are all related to the generative characteristics of AI or creative output. They are more like using a banner full of technology to support the facade, and the final monetization model returns to the existing traditional route - trading products or renting services.

The lack of a reliable business model has made the capital market skeptical of generative AI applications. So far, there has not been a single public offering in the AI ​​field, and even OpenAI has no plans to go public in the short term.The market’s enthusiasm for the “Big Seven” is based on AI’s potential to empower the existing business models of technology leaders, rather than AI itself.

OpenAI is reportedly secretly developing a new model codenamed "Strawberry", with the goal of enabling AI to conduct research autonomously with the help of a computer usage agent (CUA) and take action based on the research results. "We hope that the artificial intelligence model can see and understand the world like us."

Although major companies have invested heavily in the field of artificial intelligence, the proportion is still not comparable to the dot-com bubble era of the millennium. According to calculations by Ryan Hammond's team, at the height of the dot-com bubble, technology, media and telecommunications companies spent more than 100% of their operating cash flow on capital expenditures and research and development. Today, this figure is 72%.

From the perspective of game theory, in this round of artificial intelligence arms race, no matter which giant it is, will easily withdraw. Instead, they will continue to increase their investment and find a truly sustainable business model.

After all, human development certainly requires scientists to explore new hot spots, but the most important thing for market behavior is to find universal points that can be widely integrated with daily behavior.

A good business model should be based on existing demand, then products, and then products will generate new demand and continue to expand. In the AI ​​era, it seems that products are ahead of demand. Companies are frantically increasing the computing power efficiency of large models, but most users do not want AI to replace their brains.

From looking up at the stars to keeping your feet on the ground,AIThe industry is taking this critical step. No one wants to see“Internet Bubble”It happens again.