These days, looking at the stock market, the value of companies is shifting by hundreds of billions of dollars due to AI.

In Korea, the trend of investing in U.S. stocks continues, and Koreans living in the U.S. are showing immense interest in stocks like NVIDIA, Google, Amazon, and Microsoft.

However, after reading the recently released sustainability reports from Google and Amazon, I found some thought-provoking points.

AI can potentially generate a lot of revenue. But it is also an industry that requires significant upfront investment.

This is a crucial point from an investor's perspective.

Beginner investors only look at how much revenue has increased. But seasoned investors first consider costs. Just like if you earn 10 billion but spend 12 billion, you end up in the red, the key with AI is ultimately how much profit is left.

Running a single AI service properly requires more than just adding a few servers.

It involves building data centers worth billions of dollars, purchasing tens of thousands of GPUs, and consuming electricity 24/7. Cooling systems must also run continuously, and a significant amount of water is used.

In the past, software companies were seen as companies that made great computers, but now it feels more like operating a power plant.

This was clearly reflected in the recent reports. Both Google and Amazon reported an increase in carbon emissions. While companies didn't explicitly state it was due to AI, the answers are almost clear when you read the reports thoroughly.

They are building more data centers, buying more semiconductors, and consuming more electricity.

There is one word that investors must pay attention to: margin.

An increase in AI revenue by 30% may not be a cause for celebration.

If data center investment costs rise by 50% and electricity prices continue to climb, the actual profit may be less than expected.

In the future, the competitive edge for AI companies may shift from who can make AI smarter to who can operate it more cheaply.

And what I find even more interesting is electricity.

AI consumes a lot more electricity than expected.

Ultimately, the competition in AI is shifting from a battle of computers to a competition for securing electricity.

Therefore, investment ideas may need to change a bit.

While everyone is focused on NVIDIA, I have started to look at the companies making money alongside it.

Power plants, power grids, transformers, data center construction companies, cooling equipment manufacturers, optical cables, power semiconductors, industrial cooling systems...

These companies may not be glamorous.

But they are industries that are likely to grow as AI expands.

The same goes for semiconductors. Focusing only on GPUs means you are only seeing half the picture.

On the other hand, there are some concerning aspects.

Looking at the amount of money big tech companies are investing in data centers is truly jaw-dropping.

What if the speed of AI monetization is slower than the market expects?

Enormous depreciation and operating costs could pressure performance.

Just like during the internet bubble when too much fiber optic infrastructure was laid, similar issues could arise again.

Environmental regulations are also a variable.

As carbon regulations tighten, the operating costs of data centers are likely to continue rising.

Ultimately, companies will need to buy carbon credits, increase contracts for green energy, and use low-carbon building materials.

In the end, those costs will reflect on corporate performance, and some may be passed on to the prices of AI services.

Ultimately, investors need to look at where the money is flowing rather than the flashy term 'AI.'

The winners in the AI era may not just be the companies that create the smartest AI.

The real winners could be the companies that secure electricity most cheaply, operate data centers efficiently, and control costs effectively.