Chat gpt It has been exactly two years since the start of the productive artificial intelligence revolution, and these days investors have started to turn their attention from the benefits of this magnificent technology to its very high costs. Every day people question why Meta, Amazon, Tesla spend so much money on this business. In fact, this is an inevitable part of the excitement cycle experienced with every revolutionary new technology, first very high excitement, then questioning. Major technology companies will spend more than $200 billion in capital expenditure on artificial intelligence this year. In this case, a little questioning seems normal. So where will they spend it? Of course, the expenses to build data centers, buy the chips necessary to train artificial intelligence models, and provide energy for the centers to operate are really scary. But it is also possible that artificial intelligence will surprise investors with the benefits it provides in the coming months and years. When asked, many S&P 500 company executives say that artificial intelligence changes business processes and information access, and increases employee productivity. They describe the incredible capital spending wave of technology companies as necessary and rightful.
On the other hand, there are those who compare the parabolic rise in stock values of companies such as Nvidia and Palantir to the period before the dotcom crash in 2000. Let's remember what it was like at that time, it was believed that the internet would change everything and that all trade would be done over the internet. Money was flowing into companies related to fiber optic cables in particular. At that time, the value of a company selling pet food became equal to 20 years of the food consumption of all pets in the world. Such a madness occurred, then the bubble burst in 2000 and the crash occurred. There are many who compare this period to that period, but I think this is a completely wrong analogy. The construction of artificial intelligence data centers does not resemble the construction of fiber optic infrastructure of this century. Artificial intelligence data centers are more like highways or electric transmission lines.
Ultimately, they will create a much more productive economy. In this context, they will have revolutionary effects on the economy. I have stated for 2 years that investors should not stay away from opportunities in the field of artificial intelligence. I have talked about dozens of companies on this subject. The incredible rises in almost all AI-related stocks over the last two years have proven my claim right, and now I claim that we are just at the beginning of this process. We still have a long way to go. The AI revolution hasn’t even started yet, and it’s not just about Nvidia’s chips anymore, there are other areas that will grow in AI. Energy is one of them; data centers that power cloud computing platforms today typically use around 50 MW of power. AI data centers will need 10 times that much.
Giordano Albertazzi, CEO of Vertiv Holdings, which provides power and cooling infrastructure equipment for data centers, says that his company’s customers are increasingly building data centers with capacities of 500 MW or more. We are moving from 50 MW to 500 MW. An industry executive said that he has seen plans for at least 10 different data center projects exceeding 500 MW, and that they are planned to be built in the next 3 years. The technology underlying AI continues to benefit from the laws of scaling. This means that AI models continue to evolve in parallel with how much computing power and data is used to train each successive model. In other words, the more powerful the data center, the better the AI, the better the learning models.
Mark Russinovich, Microsoft Azure chief technology officer, says that there seems to be no end to scaling so far. Although there are different statements on this subject. Some say that AI can no longer grow with just scaling, that it may need new models, and that large language models are slowly coming to an end. But Azure chief technology officer, who is from the industry, argues the opposite, and of course the costs of creating new models are also increasing. According to Bernstein's estimates, it took $300 million to train Open AI's current gpt 4 model. It seems that several billion dollars will be needed for the next version. The next one will require more than $25 billion. Because AI uses increasingly large language models. It is becoming able to perform more diverse and complex calculations. This increases the need for a data center behind the scenes.
There are so many different places that Elon Musk's newly founded XAI, as you know, this cloud AI company that works on Twitter, has already started using more than 100 thousand Nvidia Hopper graphics processes. The bill for these GPUs alone is several billion dollars. Musk won't stop there. Musk says he plans to double the capacity of his AI startup in the coming months, reaching 200,000 GPUs. Oracle says it will use 131,072 Nvidia GPUs in the data center, which will be available in the first half of 2025 and will be used to solve AI workloads for its customers. This will be the largest GPU cluster ever, even bigger than Elon Musk's XAI. Meta's unpleasant CEO Mark Zuckerberg also announced last month that his company trained Llama 4 on more than 100,000 Nvidia GPUs. Meta's current Llama 3 model is trained on only 16,000 Nvidia GPUs. Here, we see an increase of almost 10 times in the number of GPUs required.
If 10 times more GPUs are needed to train each new AI model, all the current estimates regarding AI chips, data centers, and Nvidia's growth may seem ridiculous. Bank of America Global Research estimates that data center investment spending could increase by 14% annually, from $215 billion in 2023 to $311 billion in 2026. Nvidia's CEO Jensen claims that they have a $1 trillion market ahead of them and says that current CPUs will also turn into GPUs. This is another dimension of the business, frankly. I think these estimates will increase even more in the coming days. Because so far, we have only been talking about data center founders, whom we call Hyper Scalers. There are larger institutions, such as Google, Meta, Amazon.
I mean large institutions that are not technology companies, like banks, fast-moving consumer goods companies, big retailers. They will also need AI and maybe build their own data centers, and that's not even taken into account. Companies want to make their employees more productive by processing the information stored in their databases faster and easier, and this productivity, efficiency increase issue is not as interesting as chatbots like chat gpt, but it is no less disruptive and revolutionary, and I think there are actually huge opportunities for change in this area. Palantir's extraordinary success this year, which has provided large productivity increases to large organizations through AI, is proof of what I'm saying.
Arthur Lewis, president of Dell Technology's infrastructure solutions group, says that if you think about all the data in the world, most of it is just sitting in archives and is useless. Companies now want to use this data to feed AI models and understand the full value of this data. A technical researcher working at Google Cloud says that an increasing number of his customers are using AI to synthesize and analyze large amounts of complex data using a conversational interface. Yes, some customers also come and get these services from companies like Google.
Hewlett Packard Enterprise CEO Antonio Neri says that customers choose a broad language model and fine-tune the models with their own data according to their unique and specific needs. So they go and buy the chat gpt language model, for example. But they feed it with their own data and develop their own unique AI. According to the HP CEO, the first issue is that AI is a tool for increasing business efficiency. This is where he sees the real growth opportunities. Jon Lin, vice president of data center operator Equinix, says that drawing cats is not the main thing that excites me in AI, but basic material discovery, drug discovery, new measures to improve health, these are the areas where AI has the most potential.
Another example is that pharmaceutical giant Bristol Myers Squibb bought an Nvidia AI data center a year ago and the company says it is already reaping the rewards. Greg Meyers, Bristol's digital and technology manager, says AI has accelerated our capabilities. Finding new drugs to combat really difficult diseases is a scientific problem as well as a computational problem. Here he is talking about the processing capacity of computers. Meyers says that AI allows Bristol's drug discovery scientists to try more permutations, do more experiments, do more research.
AI also allows the company to leverage data from previous clinical trials to better design future clinical trials, the executive says. As a result of AI work, we are on track to reduce our clinical trial cycle time by almost 2 years, he says. This is truly a revolution in the pharmaceutical industry. Using AI to do more simulations and develop drugs faster is the dream of the pharmaceutical industry. It would also be an important tool in defeating diseases in the world.
Megacap tech executives are also praising the growth rates of new AI startups, calling them unprecedented growth. For example, Amazon CEO Andy Jassy recently told investors that the company’s generative AI business is on track to generate billions of dollars in revenue per year. The AI business is growing more than twice year over year, with a growth rate that exceeds the early stages of the Amazon Web Services cloud business. Remember, Amazon Web Services was the company’s fastest-growing area. The growth rate of AI startups is now double that. Jassy said at Amazon’s latest earnings meeting that generative AI is truly extraordinary. It’s perhaps a once-in-a-lifetime opportunity. It’s really hard not to get excited.
Microsoft CEO Satya Nadella has similarly praised the rapid growth of its cloud business. Microsoft’s CEO said in its earnings meeting last month that AI-driven businesses are approaching $10 billion in revenue, making it the fastest business in Microsoft’s history to reach $10 billion in revenue. Meanwhile, these companies aren’t just using AI to serve customers; they’re also improving their own efficiency. Meta’s Zuckerberg told investors last month that AI-driven streaming and video recommendations have led to an 8% increase in time spent on Facebook and a 6% increase on Instagram this year. In other words, they’re presenting users with more relevant, more personalized content. This way, they’re getting them to spend more time in front of their screens.
Of course, this may not be a good thing for humanity, but it’s a very good thing for Meta. Because it gets the chance to show more ads. Zuckerberg says that more than 1 million advertisers use the company's AI tools to create ads, and that the initial results have increased by 7% in revenue. These are huge numbers, so Meta is already a very big company, a 7% increase in ad revenue is amazing. Sundar Pichai, CEO of Alphabet or Google, which seems to be a little behind in terms of AI but has actually achieved important things on the Gemini side, also makes statements on this subject. The CEO says that more than a quarter of new codes at Google are now created first by AI. It is then checked and reviewed by humans. Pichai says that his company's 7 important products with more than 2 billion users include Google's AI Gemini model.
In the last quarter of September, Alphabet, Microsoft and Amazon increased their capital expenditures by 62%, 51% and 81%, respectively, compared to the previous year. When we add Meta to these companies, the total investment size reaches 230 billion dollars per year. Microsoft says that we have a capacity constraint for Azure AI services. So we don't have enough AI GPU servers to meet the current demand. Then someone says, "Short Nvidia a little bit, I don't recommend it." Microsoft says we can't find GPUs. Alphabet and Amazon say they will spend even more money next year. Microsoft also plans to increase its AI capacity in the first half of 2025. Meta also said they plan to make significant capital expenditures for 2025. The stock had already fallen a little after this statement. Because Wall Street says find a way to make money without spending money, what's the point of spending money on AI?
There is one product that all companies investing in AI are after, and that is Nvidia's gb200 nvl 72 AI system. This is actually a server, think of it like a small computer, it has 72 GPUs on it. This package sets Nvidia apart from all its competitors. Vertiv CEO Albertazzi says that the nvl 72 is a great value proposition. Customers want the nvl 72 because it is much more efficient and powerful than previous models, allowing companies to save money on overall AI model training and querying.
Nvidia's new system provides 30 times more performance than the previous system, the h100, according to Nvidia's claim. The company also says that this new system provides 4 times more speed for training artificial intelligence models. The excitement created by GB200 is so intense that Microsoft, Open AI and Google shared photos of the Nvidia server on their social media accounts last month and said that they were able to get it. We know that Elon Musk also intervened to get these GPUs as soon as possible.
Yes, developments in artificial intelligence are like this. There are more investment opportunities. Of course, it is necessary to choose the right companies here. The values of some companies increased just because they had artificial intelligence in their names. Those bubbles will probably burst there, but the companies that really contribute to the artificial intelligence revolution still have a long way to go.
The information, comments and recommendations contained herein are not within the scope of investment consultancy. Investment consultancy services are provided within the framework of the investment consultancy agreement to be signed between brokerage firms, portfolio management companies, banks that do not accept deposits and customers. The comments in this article are only my personal comments and these comments may not be appropriate for your financial situation and risk return. For this reason, investments should not be made based on the information and comments in my articles.
