虎嗅

"The Glue Behind Chips: Is It Getting More Expensive as AI Grows? Uncovering the Overlooked Resource War"

原文:芯片背后的胶水,AI越火它越贵?揭秘被忽视的资源战

Summary of the Core Content

The main takeaway from this news is that, amidst the AI boom, tin has become a crucial but often overlooked resource in the global computing power supply chain. As AI servers require 3 to 10 times more tin than traditional servers, the sudden surge in demand has led to a skyrocketing price of tin. This makes tin a critical element that can potentially bottleneck the entire computing power industry.

Detailed Analysis

#### Why is Tin Called the “Glue” for Chips?

If you open a smartphone or computer, you’ll see that the chips are not fixed in place by magic. Tin acts as a strong adhesive (more precisely, “solder”) that securely connects the metal pins on the chips to the contacts on the circuit board. Without tin, the chips and the circuit board would be completely isolated, preventing electricity from flowing and rendering the chips useless. In short, tin serves as the “connector” that enables chips to function; without it, even the most advanced chips would be nothing but useless pieces of metal.

#### Why Do AI Servers Use 3 to 10 Times More Tin than Traditional Servers?

AI servers are much more power-consuming compared to regular servers. Regular servers handle everyday data and have fewer, smaller chips, while AI servers need to run large models and process massive amounts of data, which requires many high-performance chips (such as NVIDIA GPUs). Each chip must be soldered using tin, and since AI chips are typically larger and have more pins, the amount of tin needed is significantly higher. For example, a regular server might use 10 grams of tin, while an AI server could require 30 to 100 grams. With the global demand for AI servers surging, it’s no wonder tin prices have skyrocketed.

#### The Logic Behind the Tin Price Spike

The increase in tin prices is not due to speculation but a genuine imbalance between supply and demand:

  • Demand: The AI boom has led to a sudden and persistent surge in tin consumption, as AI servers continue to be deployed on a large scale.
  • Supply: Tin mining is not a quick process; it takes several years to discover new deposits, build mines, and obtain the necessary permits. Moreover, the world’s main tin reserves are concentrated in countries like Indonesia, China, and Peru, where production growth is slow. It’s like suddenly having 100 people demanding bread when a bakery can only produce 50 per day—prices inevitably rise.

#### Why Can Tin Become a Bottleneck for the Global Computing Power Chain?

The core of the computing power chain is AI servers, and tin is essential for their operation. Without sufficient tin, servers cannot be assembled. Moreover, the supply of tin is highly concentrated in a few countries. If a major tin-producing country (such as Indonesia) restricts exports or if a mine experiences an accident, global tin supplies could be disrupted, halting AI server production. This is analogous to a critical pipe being blocked, stopping the entire flow of water.

#### What Does This Resource Competition Mean?

  • For Companies: Server manufacturers (like Dell and Huawei) and chip companies need to act quickly to secure tin supplies. They either need to stockpile it in advance or sign long-term contracts with mining companies to avoid production disruptions.
  • For Governments: Countries should prioritize strategic reserves of tin, similar to oil reserves, to prevent being dependent on other countries for this vital resource.
  • For Consumers: AI-related products (such as AI-powered laptops and smart speakers) may see price increases due to the rising cost of tin. It also highlights how seemingly insignificant metals like tin, cobalt, and lithium are essential foundations of high technology; without them, even the most advanced technologies cannot function effectively.

In One Sentence

The more popular AI becomes, the more important tin becomes. Although it may not be a high-profile component, tin is crucial for enabling chips to work effectively. It has already become an invisible but critical factor in the global computing power supply chain.