One of the most popular questions in the investment world these days is whether spending on artificial intelligence will decrease. Will companies like Nvidia, AMD, Broadcom have difficulty selling chips? My short answer to this question is no, on the contrary, spending will increase explosively. Large technology companies, governments, and investors will continue to pour billions of dollars into artificial intelligence. Some of this transfer will go to infrastructure, some to software, but more money will continue to pour in. Because brand new artificial intelligence models that are much smarter and also real energy monsters await us. The situation is roughly this; artificial intelligence no longer works just like a chatbot. It has been about 2 years since we met Chat GPT. We were basically using them as a chatbot, an advanced search engine. But now, large language models are evolving into reasoning models and artificial intelligence agents. These new systems do not just answer our questions. They think about the problem we present to them step by step.

For example, let's say you want to plan a weekend vacation for yourself. I want to go on holiday in London, tell me your AI suggestions, old AIs used to offer you a set of lists. If you impose restrictions, for example if you say I will go with my child, they make that list a little clearer. New AIs no longer offer such a list, they only compare hotel prices for you. They look at the weather, offer suggestions according to your budget and ensure that your trip to London is great. In other words, this is a thinking AI and thinking AIs are not cheap at all. For example, Nvidia says that reasoning models use 100 times more processing power than old AI models. This also means that they use 100 times more energy. Because they try to reason in the background. AIs ask themselves hundreds of questions. Which hotel is better? Which is more suitable for the person asking the question? What happens if it rains? What will the traffic be like in London that day? They have to think about all these questions. This means that more chips, more electricity and more data centers are needed.

This change accelerated a lot last year. First, Open AI introduced its first reasoning model, O1, in September 2024, and the race began. For example, this model would tell you the logic sequence it followed step by step while solving a mathematical problem. This is called chain of thought. Then DeepSeek's R1 came from China and the world was shaken. People got into a fight in the world of artificial intelligence, everyone was surprised. Because this new model could be trained for much less money. Although it later turned out that it was not trained for as cheap a price as the Chinese exaggerated. But still, with some good tricks done by DeepSeek, it seemed that the training of artificial intelligence had become cheaper. What does this mean; fewer chips and fewer data centers are needed. The market was so scared that Nvidia lost nearly 15% of its value in just one day that week. They went through artificial intelligence companies. Not only chip manufacturers, but also software manufacturers, energy producers, companies that cool artificial intelligence centers, everyone fell hard. But later on, it gradually became clear that it was not like that. In fact, the opposite is true. Companies need more infrastructure for these energy-consuming monster models.

Elon Musk's XAI, for example, has built one of the largest data centers and launched Grok 3 last week. I am currently using it. I am extremely pleased. Musk's model seeks answers to crazy questions such as how to settle on Mars. I asked Grok 3 many questions, and he gave incredibly detailed answers. I felt like an artificial intelligence model that actually behaved like a thinking person. I have never been so deeply affected by artificial intelligence before. On the other hand, Open AI's boss Sam Altman also says that the next update will be great, will bring a brand new reasoning ability, and of course, all this does not come for free. Let's look at the financial aspect of the matter. Big technology companies continue to pour money into this area.

By 2025, Google, Microsoft, and Meta expect to spend at least $215 billion on capital expenditures alone. Most of that money will go to AI data centers. That’s a 45% increase from the previous year. In the last quarter alone, Amazon, Microsoft, Alphabet, and Meta spent a combined $75 billion on equipment and property. That’s the size of the money spent, and I’m talking about just four companies. There are dozens more to add. So why is so much money being spent? Because these reasoning models require more power not only when they’re being trained, but also when they’re being used. In older models, there was a trained AI at the center. Our questions would be directed to it, and it would answer them. But now, both the device in our hands and the AI ​​at the center need to think. To put it simply, this means that they produce 100 times more words in the background for every question you ask. For example, if you ask what to cook today, the model analyzes 10 different recipes, thinks of 1,000 different words, and finally gives you the answer to make chicken. That’s all you see. But there are many processes going on in the background and they literally exploit electricity and chips in all these processes.

Now you may say, didn't DeepSeek's R1 show us that these jobs could be done with less energy and fewer microprocessors? In fact, their first claim was that this job could be done with one-tenth the cost. Also, that 1/10 thing is not true. Even if it is true, there is a problem. If DeepSeek's model really reduces the cost of artificial intelligence to 1/10, then there will most likely be more demand for artificial intelligence. Reasoning models consume 100 times more power than other models with each query, and thus the total power requirement still increases. We see examples of this in daily life.

Experts say that the efficiency of artificial intelligence may increase even more in the coming period. It may become 1000 times better, but demand may increase even faster than that. For example, it may increase by a trillion times in 10 years. Because we will benefit from AI in every question we write to WhatsApp, every job we try to do, every problem we try to solve. Our car will tell us that we are low on gas, low on energy, go to that station. It will make the reservation there for us. All of these will happen with AI. That's why Google is already building giant data centers. Microsoft is after nuclear power plants to generate electricity. The market can become 1000 times bigger than it is now, and if you want to be ahead in this race, you must continue to invest, that's for sure.

That's why AI spending is increasing. As efficiency increases in AI, we use it more. As AI does smarter things thanks to thinking models, we make them do more work. As a result, our production and productivity increase. But more AI is needed for this. I think AI companies will continue to grow at full speed in this context. I think the world's energy needs will increase much more.

The market cannot fully shape these in its mind for now. That's why I want to continue to be a long-term AI investor. Of course, we cannot know the short-term price movements, but in the long term, the biggest change in history continues to accelerate. I have seen so many developments, especially in the last 4-5 months, such as DeepSeek, Open AI, Elon Musk's chat GPT, and Cloud, that I am sure we are still at the beginning of this race. I hope it caught your attention, if you liked it, a like would be great.

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.