Why software underperforming amid the AI craze?

Table of Contents

Software still plays the key role

I have repeatedly emphasized the importance of software in my books and this blog. This is impossible to change in the development of human science and technology, even under the current wave of artificial intelligence.

It is obvious that artificial intelligence requires software to be realized, to be implemented in the daily life scenes of the public, and to bring benefits to everyone. But why do software stocks generally perform poorly amid the artificial intelligence craze?

Why haven’t software caught up AI tide?

Software stocks take a hit

I published an article a few days ago, “Semiconductor replaced software and become investors’ darling“. The main purpose of this post is to remind investors that software has been replaced by semiconductors and has become the favorite of U.S. stock investors, both in terms of market value and growth.

Software stocks have been falling behind semiconductor stocks for the first time, and at least in the short to medium term (no one can say for sure in the long term), it is unlikely to reverse.

In my post of “Artificial intelligence benefits industries“, although software vendors were also listed, the so-called degree of benefit is still very different.

Why haven’t software stocks caught up with the rising tide of artificial intelligence? This is the focus of this article.

Software and semiconductor are worlds apart

According to data from Business Insider, the growth performance of semiconductor and hardware device stocks has exceeded that of software stocks by 30% since 2024. MongoDB, Workday, Adobe, Salesforce.com, Snowflake and other well-known software stocks have seen their stock prices crash repeatedly during this artificial intelligence trend.

But why?

The main reasons are as follows:

How much does generative AI cost?

It costs a lot of money. For detailed figures and examples, please see my post of “How Much Does Generative AI Cost?“

Software development cycle is too long

Even if there is the ability to build the required data centers and infrastructure, it will take more manpower, material resources, and time to fully integrate generative artificial intelligence into the company’s software and systems. After switching to the actual application level, it will take some time to see significant revenue returns. This journey is actually very long, and there is also the risk of failure.

Hardware is quite different

Because if you want to enter generative artificial intelligence, you need to spend huge sums of money to purchase a large number of GPU chips, purchase servers starting in the thousands or tens of thousands, purchase network and communication infrastructure related to the data center, and establish daily operations of the data center, can we start the development of LLM.

Note: UBS estimated in 2023 that earlier versions of ChatGPT would require approximately 10,000 GPUs. Musk estimates that an upgraded version of ChatGPT would require three to five times that number of GPUs.

Please note that before these links, every company involved in making profits is a hardware device manufacturer, a semiconductor manufacturer, a network communication provider, an infrastructure provider, and a data center operator. In addition, most of these types of manufacturers are one time, a huge number of profiteers. This also fully explains why in this wave of artificial intelligence craze, it is they who are making money and profits, and the stocks are rising sharply, not the software vendors who spend money.

As long as the hardware devices, chips, servers, and network communication gears are designed, basically the factory can work overtime day and night, and the machines will automatically produce as much as the manufacturer needs. However, software requires a large number of software engineers to use their brains, write codes, hold meetings and coordinate, integrate, debug, and test back and forth. Sufficient programs usually cannot be launched in years.

Life is not created equal, and manufacturers are engaged in different fields. This is also true. Factory robots don’t need to rest, won’t get sick, and don’t have emotional distress; but people will get tired and fall asleep, so working overtime at night is not necessarily useful to software engineers.

Barriers to entry

How expensive are AI chips?

For Nvidia, which accounts for 90% of the market, the current price of the latest Blackwell architecture GPU artificial intelligence chip is between US$30,000 and US$40,000. The price of the previous generation H100 GPU chip was as high as US$45,000. The H100 The average selling price of the H800 is about 2 to 2.5 times that of the earlier generation of GPU chips A100 and A800.

The price of H100 is more than four times that of AMD’s MI300 artificial intelligence chip product series, which has the second largest market share, and both are in short supply.

Please note that I mentioned in the previous article that the purchase amount of large and significant leading manufacturers is based on the number of servers starting in the thousands or tens of thousands. You can calculate how high the cost is.

Only tech behemoths are capable

Because of this, large technology companies such as Microsoft, Google, Amazon, Meta, Oracle, Tesla, Alibaba, Tencent, Zidong, Baidu, etc. have also made significant capital expenditures on artificial intelligence and cloud infrastructure in the past two years. increase. Please note that these companies are all very large companies, not ordinary large companies. This shows that the entry barrier for generative artificial intelligence is very high.

Big tech spending soars

According to figures released by various listed companies──Microsoft’s capital expenditures increased by 79% in the first quarter of 2024, reaching a record of US$14 billion; In July 2024, it was even reported that Microsoft would double its capital expenditures for artificial intelligence from US$32B in fiscal year 2023 to US$63B in fiscal year 2025.

Alphabet’s Google’s capital expenditures also increased by 90%, reaching a record of US$120 billion. One hundred million U.S. dollars. Meta is also ramping up capital spending, and while its quarterly capex fell to $6.7 billion from $7 billion a year ago, its full-year capex forecast was raised to $35 billion to $40 billion from $30 billion to $37 billion.

These capital expenditures are used to upgrade and build data centers, develop and purchase GPUs and specialized chips, and build infrastructure to support the expansion of generative artificial intelligence applications.

In order to prevent such large expenditures from affecting profits, these two companies also carried out large-scale layoffs to free up budgets and spend money wisely.

Conclusion

Strictly speaking, among the current large software companies, only Microsoft actually generates considerable revenue in artificial intelligence. This is why in my post of “Top five lucrative artificial lucrative intelligence listed companies“, only Microsoft is a software company.

The result is that this stock has the highest market capitalization, and the company’s stock is easy to rise but difficult to fall.

For the following reasons of information systems, software stocks have no reason to rise no matter how you look at them:

The software industry must spend a lot of money on expensive data centers, servers, and unattainable chips; capital expenditures increase significantly, which of course affects profits. The financial report cannot be good-looking!
Wall Street and the media are pursuing artificial intelligence and believe that it is profitable to invest money in stocks that are currently benefiting from artificial intelligence (this is indeed true at present. After all, it is a fact that the chip stocks listed in this article that are currently benefiting from artificial intelligence are making big money. ), the masses are blind.
The most serious investors are in love with the new and hate the old, blindly touting the chip stocks listed in this article that are currently benefiting from artificial intelligence.
Due to the impact of inflation and poor economic climate, the software industry in Silicon Valley has laid off 100,000 people so far this year alone. Most companies are cautious in investing in information systems, which has affected the revenue of the software industry.

Disclaimer

The content of this site is the author’s personal opinions and is for reference only. I am not responsible for the correctness, opinions, and immediacy of the content and information of the article. Readers must make their own judgments.
I shall not be liable for any damages or other legal liabilities for the direct or indirect losses caused by the readers’ direct or indirect reliance on and reference to the information on this site, or all the responsibilities arising therefrom, as a result of any investment behavior.