Summary of Key Points
In the past year, global memory prices have skyrocketed, with increases ranging from 300% to 500% on average, and in extreme cases, even by a factor of 20. The fundamental reason behind this is a severe imbalance between supply and demand: the explosive growth in AI demand has exhausted the available memory, while the supply side has been unable to keep up due to previous production cuts during industry cycles. Additionally, the cyclical nature of the memory industry itself makes prices prone to sharp fluctuations, which is the key factor contributing to these price spikes.
Detailed Explanation
#### 1. The Direct Cause of the Price Surge: Surging Demand with Stagnant Supply
The main reason for the dramatic rise in memory prices is that there are too many buyers and too little supply.
- Demand Side: The primary driver is AI. Large models such as ChatGPT and Wenxin Yiyuan require massive amounts of memory for training and operation—just like you need a large amount of paper to solve complex math problems; AI needs to temporarily store billions of parameters and vast amounts of data in memory to process them quickly. Besides AI, the consumer electronics market (smartphones, computers) is also gradually recovering, resulting in a combined demand that exceeds the available supply.
- Supply Side: The memory industry has a peculiar cycle: when prices are low, manufacturers lose money and cut production (for example, Samsung and SK Hynix have reduced their capacity). Now that demand has surged suddenly, it will take at least 1 to 2 years for manufacturers to expand production by building new factories, purchasing equipment, and adjusting production lines. As a result, supply cannot keep up with demand, leading to soaring prices.
#### 2. The Intimate Relationship Between AI and Memory: AI is the Major Consumer of Memory
Why does AI consume so much memory? Simply put:
- Training Phase: AI models need to process large amounts of data (tens of millions of articles or images), which must be temporarily stored in memory for quick calculations and parameter adjustments. The larger the model (e.g., GPT-4), the more memory it requires. For instance, training GPT-3 may require hundreds of terabytes of memory, and GPT-4 likely needs even more.
- Operation Phase: When using AI for communication, it needs to quickly retrieve previous conversations and model parameters, which also relies on memory. Insufficient memory can cause lagging responses or even system crashes.
Currently, global AI companies (such as Microsoft, Google, and ByteDance in China) are purchasing large amounts of memory, pushing prices to new heights.
#### 3. The Cyclical Nature of the Memory Industry: Price Fluctuations Are Common
This is not the first time memory prices have skyrocketed; it's a recurring issue in this industry.
The memory industry is highly cyclical:
- Surplus Supply: When demand falls (e.g., during the decline in smartphone and computer sales a few years ago), memory prices plummeted (DDR4 memory dropped below 100 yuan in 2019), and manufacturers cut production to avoid losses.
- Shortage Demand: When demand surges (e.g., with the AI boom), but production has not yet recovered, prices rise sharply.
This cycle of "production cuts → shortages → price increases → expansion → oversupply → price drops → production cuts" is the root cause of memory price fluctuations. This time, due to the rapid growth in AI demand, the cycle has been amplified, leading to particularly extreme price increases.
#### 4. The Impact on Ordinary People
The increase in memory prices affects us as well:
- Smartphones/Computers: If you wanted to buy a laptop with 8GB of memory last year, it might have cost 3000 yuan. This year, due to higher memory costs, manufacturers may raise the price by several hundred yuan or reduce the memory capacity to 6GB (under the guise of a "basic version"). Choosing a model with 16GB of memory could result in a significant price difference.
- Other Devices: Memory is used in smartwatches, routers, gaming consoles, and other devices, so cost increases will be passed on to consumers. However, the increase is usually less severe for consumer-grade memory compared to server memory used by AI systems. Manufacturers may also use cheaper older models (e.g., DDR4 instead of DDR5) to reduce costs, though this might result in a lower quality experience.
#### 5. Will Prices Fall in the Future?
- Short Term (within 1 year): Inventory is already low, and production expansion has not kept up with demand, so prices are likely to continue rising or remain high.
- Long Term (1–2 years): Manufacturers like Samsung and SK Hynix have announced plans to expand DDR5 memory production. Once new capacity is available, supply will increase, and prices should gradually decline. However, if AI demand grows faster than production, prices may still stay high.
In summary, the surge in memory prices is a result of a combination of supply-demand imbalance, industry cycles, and the explosive growth in AI. For ordinary people, if you need to buy electronic devices soon, it's best to act now or opt for models with more memory to avoid future shortages. If you're not in a hurry, you can wait for half a year to a year before making a decision.
By understanding these factors, you should have a clearer picture of the reasons behind the price increases. The core issue is the sudden surge in demand that has outpaced supply, coupled with the cyclical nature of the memory industry, leading to these extreme price changes.