Summary of Key Points
This article discusses how the increasing cost of smartphones and computers, as well as potential rising electricity prices, are due to the "third wave of inflation" triggered by the massive development of AI. AI requires specialized memory (HBM) and substantial amounts of power, which has led to a shortage of memory for consumer electronics, with the resulting costs being passed on to everyone. Economists disagree on whether AI will ultimately cause inflation or deflation, but currently, we are in a phase where significant investments are being made without immediate tangible benefits. Ordinary people are effectively paying an "AI tax," while companies that control physical resources such as memory and power are reaping the profits.
1. AI is Taking Your Memory, Leading to Higher Prices for Phones and Computers
The training of large AI models necessitates fast-data-transfer memory (HBM), which is extremely difficult to produce—similar to stacking DRAM wafers thousands of layers high and drilling at the micron level, resulting in a high failure rate. There's a "3:1" rule in the industry: for every piece of HBM used in AI, three pieces of regular memory for phones or computers are sacrificed.
Why do manufacturers still participate? Because AI giants like Google and Microsoft are willing to pay exorbitant prices! While the profit margin on regular memory is 20%-30%, the profit margin on HBM is so high that it's irresistible. As a result, memory manufacturers are dedicating their best facilities and equipment to producing HBM, leaving less capacity for consumer electronics. By 2026, 70% of the global memory production capacity will be used by data centers (compared to just 20% a few years ago).
The consequence is that consumer electronics manufacturers struggle to obtain sufficient memory, leading to price increases. Contracts used to be quarterly or annual, but now prices are updated within days before the goods even arrive. Additionally, it takes 3-5 years to build new memory factories (Micron's new plant will only start mass production in 2027), and this shortage is expected to last at least three years. As long as AI giants continue to demand HBM, prices for phones and computers are unlikely to fall.
2. Data Centers Are Power-Hungry: Rising Electricity Costs Affect Everyone
AI data centers consume enormous amounts of power, and there can be no outages. The International Energy Agency predicts that global data center electricity consumption will double in the next few years, equivalent to the annual electricity usage of Japan. In some parts of the United States, half of the new power generation capacity over the next five years will be dedicated to data centers.
The question arises: who will pay for the infrastructure upgrades? Power companies won't bear the cost alone and will likely raise electricity prices, affecting everyone within their service area (residents, restaurants, hospitals, etc.). Goldman Sachs predicts that electricity prices in the U.S. will increase by about 6% annually between 2026 and 2027.
The situation is compounded by a chain reaction: higher electricity costs for restaurants lead to higher food prices; higher electricity costs for hospitals result in increased medical expenses; and higher costs for cold storage affect the prices of meat, eggs, and dairy products. These are all expenses that ordinary people face daily. Moreover, electricity costs are regressive, meaning wealthy individuals are less affected by price increases (a 10% rise may not be significant for them), while low-wage families, whose expenses are heavily reliant on electricity and food, will feel the impact more severely—effectively subsidizing AI development.
3. Economists Are Divided: Inflation or Deflation Due to AI?
There are two opposing views among economists, both with compelling arguments:
- Short-term Realists (e.g., the Fed's hawks): Stop talking about the future and focus on the present! AI development is competing for resources such as copper, wires, transformers, and skilled electricians (the hourly wage of electricians in the U.S. has increased by 6.5%, compared to only 3.6% in other industries), all of which are driving up prices and wages, increasing the "cost of capital" in the economy. The Fed cannot afford to let inflation run wild and may continue to raise interest rates.
- Long-term Optimists (technology scholars + doves): AI represents a positive supply-side shock! Historically, major infrastructure projects like railways, power grids, and the internet have led to initial price increases, but later reduced overall costs. For example, AI could save millions of hours of labor, which should eventually lower inflation.
The key factor is time: Currently, AI is in a phase where investments are being made without immediate returns (similar to preparing a restaurant before it starts operating). If this phase lasts only two years, people can tolerate it; however, if it takes 5-8 years for benefits to emerge, the real economy will face significant challenges.
4. Who Is Making Money from AI? Ordinary People Are Paying the "AI Tax"
The distribution of profits from AI development is clear:
- Profitors: Companies that control physical resources essential for AI, such as copper mines (needed for wires), manufacturers of high-voltage transformers, memory giants (like Micron and Samsung) that monopolize HBM technology, and nuclear power plants that supply data centers. They convert the funds invested by tech giants into their own profits.
- Losers: Consumer electronics manufacturers (unable to obtain affordable memory) and ordinary people (paying higher prices for products and increased electricity costs), effectively paying an "AI tax."
When will things improve? It depends on whether engineers can quickly realize the productivity benefits of AI, such as reducing costs for restaurants or improving hospital efficiency. Until then, you might consider your electricity bills as a contribution to the evolution of AI.
This article clearly outlines the hidden costs associated with AI development: AI is not a free benefit; it requires substantial physical resources, and these costs will eventually be borne by everyone until it begins to provide tangible benefits.