There are many ways to contextualize this prediction and subsequent reaction. The sales beat analyst expectations by 53% and surpassed the company’s all-time record set in March by 33%. The first-day rally was the third-largest in U.S. history, and the two-day rally surpassed the market capitalization of all but 48 stocks worldwide.
Among the companies dwarfed by Nvidia’s $200 billion skyrocketing valuation are two of the most important enablers of the AI revolution. Of these, South Korea’s SK Hynix Inc. and Boise-based Micron Technology Inc. account for his 52% of the global market for dynamic random access memories. Their combined value is just $140 billion. Its sole rival, Samsung Electronics, with a 43% share of the DRAM industry, is just one of at least four global sectors it leads, with $317 billion in transactions.
If the generative AI sector takes off, as Nvidia and its customers believe, incumbent giants like Microsoft Corp. and upstarts like OpenAI will knock on the gates of Samsung, SK Hynix and Micron. It will be
Machines that process massive amounts of data, analyze video, audio, and text for patterns, and spew out replicas of human-generated content will need memory chips. In fact, AI companies may buy more DRAM than any technology sector in history.
The reason for the demand for memory chips is very simple. Nvidia’s AI chips differ from standard processors by sucking in tons of data all at once, crunching numbers at once, and spewing out results all at once. However, to realize this power advantage, information must enter the computer quickly and without delay. That’s where memory chips come in.
The processor does not read data directly from the hard drive. This is too slow and inefficient. The first option is to store it in temporary storage within the chip itself. But there is not enough space here to hold many things. Chip makers like to dedicate this precious area to numerical processing functions. So the next best option is to use her DRAM.
When processing billions of information at once, that data needs to be at your fingertips and delivered quickly. Without adequate DRAM in the system, the computer would slow down significantly, invalidating the value he spends $10,000 on the best processor to run advanced chatbots. This means that every time you buy a high-end AI processor, you could end up with as much as a terabyte of DRAM. That’s 30 times more than high-end laptops.
Such memory demand means that DRAM sold for servers is expected to exceed the amount in smartphones this year, according to Taipei-based researcher TrendForce.
These systems also need to be able to store large amounts of output nearby for immediate reading and writing. This is done on NAND Flash, the same chip used in smartphones and modern laptops. Samsung is the world leader in this space, followed by Japan’s Kioxia Holdings (a spinoff from Toshiba) and SK Hynix.
Combined, DRAM and NAND accounted for $8.9 billion in Samsung’s revenue last quarter, well ahead of Nvidia’s $4.3 billion from its data center business, which includes products for AI. But to put this in context, this is the worst performance for Samsung’s memory division in seven years, and the company’s AI-related memory sales represent just a fraction of its total revenue.
Both numbers are expected to increase. For every high-end AI chip sold to customers, 12 more DRAM chips will be shipped, which means more revenue for Samsung, SK Hynix and Micron. As Nvidia grows, so will these three of his companies, who jointly control his 95% of the DRAM market.
The AI revolution is definitely upon us, with the makers of cool chatbots, ubiquitous search engines, and high-powered processors being the biggest winners. But those who churn out boring old memory chips won’t be left behind either.
Details from Bloomberg Opinion:
• I got my chips back.But neither equal nor for all: Tim Calpin
• Don’t fall into the rabbit hole of AI long-termism: Palmy Olson
• US Chip Regulation Highlights Cracks in China’s AI Strategy: Tim Kalpan
This column does not necessarily reflect the opinions of the editorial board or Bloomberg LP and its owners.
Tim Calpan is a Bloomberg Opinion columnist covering technology in Asia. Previously, he was a technology reporter for Bloomberg News.
More articles like this can be found at bloomberg.com/opinion.
