Nvidia isn’t the only company profiting from the AI ​​gold rush

The building belongs to Equinix, a company that leases data center space. The equipment inside belongs to enterprise customers, who are increasingly using it to run artificial intelligence (AI) systems. The AI ​​Gold Rush, fueled by the astonishing sophistication of “generative” models such as his ChatGPT, a popular virtual conversation list, promises enormous benefits to those who harness the potential of technology. But like any other early Gold Rush, it’s already started. Cast a fortune to the seller of the necessary pickaxe and shovel.

Nvidia, which designs the semiconductors of choice for many AI servers, beat analysts’ revenue and earnings estimates for the three months to April, expecting revenue of $11 billion this quarter, May 24. announced that it is That’s half of Wall Street’s turnover. I predicted. On May 29, NVIDIA President Jensen Huang declared that the world was at a “tipping point in a new computing era.” The next day, the company’s market value soared by 30% after earnings, at one point he reached $1 trillion.

Other chip companies, from design buddies like AMD to manufacturers like Taiwan’s TSMC, are also getting caught up in the AI ​​excitement. So do other computing providers of his infrastructure. This includes everything from colorful cables, noisy air conditioning units, and data center floor space to software that helps run AI models and organize data. Since ChatGPT’s launch in November, an equally weighted index of more than 30 such companies is up 40%, while the tech-heavy NASDAQ index is up 13% (see chart). ). “New tech stacks are emerging,” sums up Daniel Jeffries of AI Infrastructure His Alliance, a lobby group.

At first glance, the AI ​​Gavin seems less appealing than the clever “large language model” behind ChatGPT and its many rapidly expanding competitors. But they need a lot of computing power now in the AI ​​pie of the future, as the builders of models and the creators of applications piggybacking on them are vying for a piece of it.

Modern AI systems, including generative sorting, are much more computationally intensive than older systems, let alone non-AI applications. Amin Vahdat, his head of AI infrastructure at Google Cloud Platform, the internet giant’s cloud computing division, observes that model sizes have increased tenfold each year for the past six years. His GPT-4, the latest version of ChatGPT’s enhancement, analyzes data using the 1trn parameter, which is probably 5 times more than his previous version. As models become more complex, the computational need to train them increases accordingly.

AI, once trained, requires less computation to use. However, given the range of applications offered, this “reasoning” also cumulatively demands a lot of processing power. The creator of ChatGPT, Microsoft has more than 2,500 customers for services that use technology from his OpenAI, nearly half of which are owned by software giants. Google’s parent company, Alphabet, has six of his products with over 2 billion users worldwide and plans to power them with generative AI.

The most obvious winners from the surge in demand for computing power are chip makers. Products from companies such as Nvidia and AMD, whose chips are designed and manufactured in his foundries such as TSMC, are in huge demand, especially from the big providers of cloud computing that power most AI applications. As such, AI benefits from more powerful chips (which tend to yield higher margins) and more chips, so it’s a boon for chip designers. Bank UBS estimates that AI will add $10 billion to $15 billion to demand for specialized chips known as graphics processing units (GPUs) over the next year or two. Nvidia’s data center revenue, which accounts for 56% of sales, could double. AMD will be launching new GPUs this year. The company, a much smaller player than his Nvidia in the GPU design game, is poised to profit even if it “just takes the wreckage” of the market, says Stacey Rasgon of brokerage firm Bernstein. To tell. Cerebras and his Graphcore are making a name for themselves, and his data provider, his PitchBook, includes about 300 such companies.

Naturally, some of the inventory also accrues to manufacturers. In April, TSMC boss CC Wei was cautious about the “gradual upside of AI-related demand.” Investor interest is growing. The company’s stock price jumped 10% following NVIDIA’s latest results, adding about $20 billion to its market capitalization. Less obvious are the companies that allow more chips to be packaged into a single processor, said Pierre Ferrag of analyst firm New Street Research. The company controls his three-quarters of the precision adhesive market, and its stock has risen more than half this year.

UBS estimates that GPUs account for about half the cost of specialized AI servers, compared to one-tenth the cost of standard servers. But that’s not the only gear you need. GPUs in a data center also need to communicate with each other to operate as a single computer. It requires advanced switches, routers and specialized chips. The market for such networking gear is expected to grow 40 percent annually over the next few years, reaching nearly $9 billion by 2027, according to research firm 650 Group. Nvidia also sells similar kits, which account for his 78% of global sales. But rivals like California-based Arista Networks have also caught the eye of investors, with the company’s shares up nearly 70% over the past year. Broadcom, which makes chips that help operate networks, said annual revenue from these will quadruple to $800 million by 2023.

The AI ​​boom is also good news for assemblers of the servers that go into data centers, said Peter Rutten of IDC, another research firm. Already one of his analyst firms, the Dell’Oro Group, said data centers in the world have increased his share of AI-dedicated servers from less than 10% today to about 20% within five years. I predict it will. Server utilization rises from about 20% to 45%.

This will benefit server makers such as Taiwan’s Wistron and Inventec, which primarily produce custom-built servers for giant cloud providers such as Amazon Web Services (AWS) and Microsoft’s Azure. Small manufacturers should do well too. The management of Wiwynn, another Taiwanese server maker, recently said that AI-related projects account for more than half of its current orders. US company Supermicro said AI products accounted for 29% of its sales in the three months to April, up from an average of 20% over the past 12 months.

All this AI hardware requires specialized software to work. Some of these programs are provided by hardware companies. For example, Nvidia’s software platform called CUDA allows a customer to get the most out of his GPU. Other companies are creating applications that allow AI companies to manage data (Datagen, Pinecone, Scale AI) and host large language models (HuggingFace, Replicate). PitchBook counts about 80 such startups. More than 20 companies have raised new funding so far this year. Pinecorn counts two big names in venture capital, Andreessen Horowitz and Tiger Global, as investors.

As with hardware, many of the main customers for this software are cloud giants. Together, Amazon, Alphabet and Microsoft are planning capital spending from $78 billion in 2022 to about $120 billion this year. Much of that goes to expanding cloud capacity. Yet the demand for AI computing is so high that even they are struggling to keep up.

That paved the way for challengers. In recent years, IBM, Nvidia, and Equinix have started offering access to GPUs “as a service.” AI-focused cloud his startups are also booming. In March, one of them, Lambda, raised his $44 million from investors including Gradient Ventures. Brannyn McBee said in a deal with Greg Brockman, a member of Google’s venture arm and co-founder of OpenAI, that the company was valued at around $200 million. The CoreWeave co-founder argues that a focus on customer service and an infrastructure designed around AI will help the company compete with cloud giants.

The final group of AI infrastructure winners, who are closest to providing a real shovel, are data center landlords. That property is filling up as the demand for cloud computing surges. Data center vacancy in the second half of 2022 will hit a record low of 3%. Specialty firms such as Equinix and rival Digital Realty are competing with big asset managers who want to add data centers to their real estate portfolios. In 2021, private market giant Blackstone paid $10 billion to QTS Realty Trust, one of the largest data center operators in the US. Blackstone’s Canadian rival Brookfield, which has invested heavily in data centers, acquired French data center company Data4 in April.

As the AI ​​infrastructure stack continues to grow, it may face constraints. One is energy. Big investors in data centers say they expect the development of new power in hubs like Northern Virginia and Silicon Valley to be delayed as power assets tap into so much power. However, the shift from giant AI models and cloud-based inference to smaller systems could slow demand. This is because, like Google’s recently unveiled scaled-down version, it requires less computing power to train and inference can be performed on smartphones. palm model.

The biggest question mark rests on the permanence of the AI ​​boom itself. Despite the popularity of ChatGPT and its likes, the beneficial use cases for this technology remain obscure. In Silicon Valley, hype can suddenly turn into disappointment. Nvidia’s market value skyrocketed in 2021 as the company’s GPUs proved to be perfect for mining Bitcoin and other cryptocurrencies, but fell as the crypto boom collapsed. And if the technology lives up to its transformative claims, it could be cracked down by regulators. Policy makers around the world, concerned about the potential for generative AI to kill jobs and spread misinformation, are already considering guardrails. Indeed, on May 11, EU lawmakers proposed a set of rules to limit chatbots.

All of these limiting factors could slow AI adoption, and in doing so could weaken the prospects for AI infrastructure companies. But maybe just a little. Even if generative AI is not as revolutionary as its proponents claim, it will almost certainly be more useful than cryptocurrency. There are many other non-generative AIs that require a lot of computing power. A global ban on generative AI is the only way to stop the gold rush, but that is not imminent. And as long as everyone is in a hurry, pickaxe and shovel hawkers will make money.

To stay on top of the biggest business and technology news, Conclusiona weekly subscriber-only newsletter.

© 2023, The Economist Newspaper. all rights reserved. Published under license from The Economist. Original content is available at www.economist.com.