Source: Content compiled from CNN

The U.S. government has imposed new export controls on sales to China of high-tech memory chips used in artificial intelligence (AI) applications.

The rules apply to U.S.-made high-bandwidth memory (HBM) technology as well as foreign-produced technology.

Here's everything you need to know about these cutting-edge semiconductors, which are in soaring demand as the world's craze for AI grows.

What is High Bandwidth Memory?

High-bandwidth memory (HBM) is basically a bunch of memory chips, small components that store data. They can store more information and transfer data faster than older technology DRAM (dynamic random access memory).

HBM chips are commonly used in graphics cards, high-performance computing systems, data centers, and self-driving cars.

Most importantly, they are essential for running increasingly popular AI applications (including generative AI), which are powered by AI processors such as graphics processing units (GPUs) produced by Nvidia (NVDA) and Advanced Micro Devices (AMD).

"Processors and memory are the two fundamental components of AI," G Dan Hutcheson, vice president of chip research firm TechInsights, told CNN. "Without memory, it's like having a logical brain but no memory."

How will these restrictions affect China?

How these restrictions will affect China The latest export restrictions were announced on December 2, following two rounds of restrictions on advanced chips announced by the Biden administration in three years, aimed at preventing China from acquiring key technologies that could give it an advantage. In retaliation, Beijing imposed new restrictions on exports of germanium, gallium and other material elements necessary to manufacture semiconductors and other high-tech equipment.

Experts say the latest export restrictions will slow China's development in AI chips and at most hinder its access to HBM. Although China's current ability to produce HBM lags behind South Korea's SK Hynix and Samsung and the United States' Micron (MU), it is developing its own HBM capabilities.

Jeffery Chiu, CEO of Ansforce, a professional network consulting firm in the technology field, told CNN: "The US export restrictions will reduce China's access to higher-quality HBM in the short term. But in the long run, China will still be able to produce independently, albeit with less advanced technology." Country?

Why is HBM so important?

The power of HBM chips lies in their larger storage space and faster data transfer speeds compared to traditional memory chips.

Since AI applications require a lot of complex calculations, these features ensure that these applications run smoothly without delays or glitches.

More storage space means that more data can be stored, transferred and processed, which improves the performance of AI applications because large language models (LLMs) enable them to have more parameters to train.

Think of increasing data speed, or bandwidth in chip terms, like a highway. The more lanes a highway has, the less likely it is to have bottlenecks, so more cars can fit.

"It's like the difference between a two-lane highway and a 100-lane highway. You don't get stuck in traffic jams,"

Who are the top manufacturers?

Currently, the global HBM market is dominated by three companies.

According to a research report released by Taipei market research firm TrendForce, Hynix will account for 50% of the total HBM market share by 2022, followed by Samsung (40%) and Micron (10%). The two Korean companies are expected to occupy a similar share of the HBM market in 2023 and 2024, accounting for a combined share of about 95%.

Micron aims to increase its HBM market share to between 20% and 25% by 2025, according to Taiwan's official Central News Agency, citing Micron executive Praveen Vaidyanatha.

The high value of HBM has led all manufacturers to devote a considerable portion of their manufacturing capacity to more advanced memory chips. Starting in 2024, HBM is expected to account for more than 20% of the total market for standard memory chips, and may exceed 30% by next year, said Avril Wu, senior vice president of research at TrendForce.

How is HBM made?

Imagine stacking multiple standard memory chips on top of each other like a hamburger. That's basically how HBM is made.

It sounds simple on the surface, but it's not easy to do, and that's reflected in the price. HBMs cost several times more per unit than conventional memory chips.

That's because HBMs are about the height of six human hairs, which means that each layer of standard memory chips stacked together also needs to be very thin, a feat that requires a state-of-the-art manufacturing technique called advanced packaging.

"Each memory chip needs to be ground down to half a hair's worth of thinness before it can be stacked together, which is very difficult," says Qiu.

In addition, before these memory chips can be stacked together, holes need to be drilled in them for wires to pass through, and the location and size of these holes need to be extremely precise.

"When you try to build these devices, you have a lot of failure points. It's almost like building a house of cards," says Hutchison.

View more at EASELINK