Summary of Key Points
This news article focuses on the capital, energy, cost, and profitability factors behind the AI revolution, with key insights from Larry Fink, CEO of BlackRock: The current AI phenomenon is not a bubble but a technological revolution that requires substantial infrastructure investment. Key highlights include:
- Cloud providers will see a significant increase in capital expenditure (65% year-over-year by 2025), making data center financing a new area of interest for the financial sector;
- Excessively high computing costs may lead to a divergence in the industry, with large companies reaping benefits while smaller firms being marginalized;
- Power supply is a more limiting factor for AI development than chip availability;
- AI has already begun to generate profits for businesses, shifting from a phase of heavy investment to one of increased profitability.
The overall logic is as follows: Capital supports infrastructure; energy determines the pace of expansion; costs affect the accessibility of technology; and profitability verifies the value of investments.
1. AI Infrastructure Becoming a New Hotspot in the Financial Market
In the past, tech financing mainly targeted software companies, venture capital, or stocks. However, now funds are flowing towards essential infrastructure for AI, such as data centers and chips. The reason? AI is extremely costly to develop. Building a 1GW (1 million kilowatt) data center can cost $50-60 billion—equivalent to the cost of constructing several aircraft carriers. Relying solely on tech companies' own funds is insufficient, so new financing methods have emerged, such as project financing (where multiple parties invest in a single data center project), infrastructure funds (dedicated to investing in such assets), and equipment leasing (cloud providers rent chips without paying the full amount upfront).
Data shows that the top seven cloud providers will increase their capital expenditure by 65% in 2025 and another 16% in 2026, accounting for 26% of their revenue (the highest percentage ever). This indicates that financial markets are treating AI infrastructure as a new investment opportunity, much like they did real estate and stocks before.
2. High Computing Costs May Exclude Small Businesses from the AI Revolution
Fink is not concerned about a decline in AI demand but rather about the high cost of computing resources, which only large companies can afford. Large tech firms (like Google and Amazon) can purchase thousands of GPU chips and build their own data centers to spread costs over a larger scale. Small businesses, on the other hand, must rent computing power from cloud platforms, incurring higher per-use costs. For example, using generative AI for customer service through cloud APIs can result in significant expenses, which may outweigh the revenue generated. This could lead to a two-tiered industry structure: large platforms with capital, chips, and data centers, and smaller companies that rely on them. If computing costs do not decrease, many small firms will struggle to survive despite having good products. Fink calls on cloud service providers to lower costs to make AI more accessible to all businesses.
3. Power Supply is a More Critical Barrier than Chips
Building data centers requires not only chips but also electricity. The power consumption of a 1GW data center is equivalent to that of a medium-sized city. While chip production can be completed in a few quarters, building power transmission lines and generating stations takes years. Fink points out that while the U.S. leads globally in chip technology, its power infrastructure is lagging behind. Countries like China are investing heavily in nuclear and solar energy, while some regions in the U.S. are cautious about new power projects. Data centers must operate 24/7, so a stable power supply is essential. Fink advocates for a “neutral energy source”—whether it's solar, natural gas, or nuclear—as long as it is affordable, reliable, and can be deployed quickly. If power supply cannot keep up with demand, even an abundance of chips will not enable the expansion of AI.
4. AI is Generating Profits, Not Just Burning Money
AI has moved from a phase of heavy investment to one of profit generation. For example, BlackRock has seen its assets increase by $1 trillion without adding new employees, and its profit margin has risen by 2.6 percentage points (260 basis points) due to AI-driven improvements in software development, automated transactions, and enhanced customer service. Other industries are also benefiting: financial firms are reducing back-office costs, manufacturing companies are improving equipment efficiency, and retail businesses are optimizing inventory management. However, not all companies will benefit equally. Those with data, technology, and capital will reap the benefits first, while smaller firms may fall behind due to high computing costs or a lack of digital infrastructure. Fink predicts that market opportunities will become more selective over the next 12 months, with only those able to convert AI into profits experiencing growth.
In Conclusion
The AI revolution relies on a combination of factors: capital, energy, cost control, and profitability. The focus is not on whether it's a bubble but on whether we can build the necessary infrastructure, reduce costs, and enable AI to truly benefit all businesses. This marks the beginning of the golden age of AI.