Here are the companies that are enjoying the AI gold rush the most. Economist Report
How rich are companies getting in the AI gold rush? The in-depth analysis of the weekly The Economist
Not a day goes by without excitement over artificial intelligence (“AI”) causing another company's market value to skyrocket. Earlier this month the share price of Dell, a hardware maker, jumped more than 30% in one day on hopes that the technology could boost sales. A few days later Together ai, a cloud computing startup, raised new funding at a valuation of $1.3 billion, up from $500 million in November. One of its investors is AI chip maker Nvidia, which itself is experiencing a long bull run. Before the launch of Chatgpt, a “generative” AI that answers questions in an incredibly human way, its market capitalization in November 2022 was around $300 billion, similar to that of Home Depot, a grocery store chain for the House. Today it is $2.3 billion, about $300 billion less than Apple.
THE WINNING COMPANIES OF THE ARTIFICIAL INTELLIGENCE BOOM
The constant flow of AI headlines makes it difficult to understand which companies are truly winners in the AI boom and which will be winners in the long run. To answer this question, the Economist examined where value has accumulated so far and how this fits in with expected sales of products and services in the AI "stack," as technologists call the various layers of hardware. and software that AI relies on to do its magic. On March 18, many companies, up and down the stack, will gather in San Jose for a four-day event hosted by Nvidia. With talks ranging from robotics to drug discovery, the event will showcase the latest innovations in artificial intelligence. Furthermore, it will highlight the furious competition between companies within the layers of the stack and, increasingly, between them.
Our analysis looked at four of these layers and the companies that inhabit them: ai-powered applications sold to companies outside the stack; the ai models themselves, such as gpt-4, the brains behind Chatgpt, and their archives (e.g., Hugging Face); the cloud-computing platforms that host many of these models and some of the applications (Amazon Web Services, Google Cloud Platform, Microsoft Azure); and hardware, such as semiconductors (produced by companies such as AMD, Intel, and Nvidia), servers (Dell), and networking equipment (Arista), responsible for the computing power of clouds.
Technological breakthroughs tend to bring forth new tech giants. The PC boom in the 1980s and 1990s pushed Microsoft, which created the Windows operating system, and Intel, which made the chips needed to run it, to the top of the corporate ladder. According to investment bank Jefferies, “Wintel” captured four-fifths of the PC industry's operating profits in the 2000s. The smartphone era did the same to Apple. Just a few years after the iPhone's launch in 2007, Apple captured more than half of cellphone makers' global operating profits.
The world is still in the early days of the generative-ai era. However, it has already been immensely profitable. Collectively, the approximately 100 companies we looked at have created $8 trillion in value for their owners since its inception – which, for the purposes of this article, we define as October 2022, just before Chatgpt launched. Not all of these gains are the result of the AI frenzy – stock markets have been on a broader run of late – but many are.
At each level of the stack, value is becoming increasingly concentrated in a handful of leading companies. In the hardware, modeling and applications industries, the three largest companies have increased their share of overall value created by an average of 14 percentage points over the past year and a half. In the cloud sector, Microsoft, which has a partnership with Chatgpt maker Openai, has overtaken Amazon and Alphabet (Google's parent company). Its market capitalization now represents 46% of the total cloud trio, up from 41% before Chatgpt's release.
NVIDIA, DELL AND MORE
The value distribution is also not uniform across strata. In absolute terms, the greatest wealth went to hardware manufacturers. This group includes chip companies (like Nvidia), companies that build servers (Dell), and companies that make networking equipment (Arista). As of October 2022, the 27 public hardware companies in our sample were worth approximately $1.5 trillion. Today the figure is 5 trillion dollars. This is what you expect in a tech boom: the underlying physical infrastructure must be built first in order to deliver the software. In the late 1990s, when the Internet boom was starting, suppliers of modems and other telecommunications equipment, such as Cisco and WorldCom, were the early winners.
The San Jose party host is by far the biggest winner so far. Nvidia accounts for approximately 57% of the market cap growth of our hardware companies. According to research firm idc, the company produces more than 80% of all AI chips. It also enjoys a near-monopoly in the networking equipment used to connect chips inside servers to data centers. In the 12 months to the end of January, revenue from Nvidia's data center business more than tripled from a year earlier. Its gross margins grew from 59% to 74%.
Nvidia's chip rivals want a slice of those riches. Established ones, such as AMD and Intel, are launching competing products. And also startups like Groq, which produces super-fast chips, and Cerebras, which produces super-sized ones. Nvidia's biggest customers, the three cloud giants, are also designing their own chips, both to reduce reliance on a single supplier and to steal some of Nvidia's juicy margins for themselves. Lisa Su, CEO of AMD, predicted that revenue from the sale of AI chips could reach $400 billion by 2027, up from $45 billion in 2023. That would be too much for Nvidia to digest alone.
NVIDIA'S FIRM GRIP
As AI applications become more widespread, an increasing share of demand will also shift away from chips needed for model training, which involves analyzing mountains of data to teach algorithms to predict the next word or pixel in a sequence. to those necessary to actually use them to respond to requests ("inference", in technical jargon). Over the past year, about two-fifths of Nvidia's AI revenue came from customers using its chips for inference. Experts expect some inference to start moving from the specialized graphics processing units (GPUs), which are Nvidia's forte, to general-purpose central processing units (CPUs), such as those used in laptops and in smartphones, which are dominated by AMD and Intel. Soon training could also be done on CPU instead of GPU.
However, Nvidia's hold on the hardware market seems secure for the next few years. Unprecedented startups will struggle to convince large customers to reconfigure enterprise hardware systems for their new technology. The deployment of their chips by the cloud giants is still limited. Nvidia has Cuda, a software platform that allows customers to tailor chips to their needs. It's popular among programmers and makes it difficult for customers to switch to competing semiconductors, which cuda doesn't support.
While in absolute terms hardware wins the value accumulation race, it is independent model makers who have seen the largest proportional gains. The collective value of 11 such companies we looked at rose from $29 billion to about $138 billion over the past 16 months. Openai is believed to be worth around $100 billion, up from $20 billion in October 2022. Anthropic's valuation rose from $3.4 billion in April 2022 to $18 billion. Mistral, a French startup founded less than a year ago, is now worth about $2 billion.
HARDWARE AND INTELLECTUAL PROPERTY
Part of this value is related to hardware. Startups buy stacks of chips, mostly from Nvidia, to train their models. Imbue, which like Openai and Anthropic is based in San Francisco, has 10,000 such chips. Cohere, a Canadian rival, has 16,000. These semiconductors can sell for tens of thousands of dollars each. As models become more and more sophisticated, more and more are needed. GPT-4 training reportedly cost around $100 million. Some suspect that training his successor could cost the Openai tenfold.
However, the true value of template creators lies in their intellectual property and the profits it can generate. The true size of these profits will depend on how fierce the competition between model suppliers is and how long it lasts. The rivalry is very heated at the moment, which may explain why the level hasn't gained much value in absolute terms.
Although Openai took an early lead, the challengers quickly made up ground. They were able to tap into the same data as the creator of Chatgpt (i.e. text and images on the Internet) and, like him, for free. Anthropic's Claude 3 is following in the footsteps of gpt-4. Four months after the release of gpt-4, Meta, Facebook's parent company, released Llama 2, a powerful rival that, unlike the proprietary models of Openai and Anthropic, is open and can be modified at will by others. In February Mistral, which has fewer than 40 employees, stunned the industry by releasing an open model whose performance nearly rivals that of gpt-4, despite requiring much less computing power to train and run.
Even smaller models increasingly offer good performance at an affordable price, points out Stephanie Zhan of Sequoia, a venture capital firm. Some are designed for specific tasks. A startup called Nixtla has developed Timegpt, a model for financial forecasting. Another, Hippocratic ai, trained its model on medical school entrance exam data to provide accurate medical advice.
The abundance of models has also allowed the growth of the application layer. The value of the 19 publicly traded software companies in our application group has increased by $1.1 trillion, or 35%, since October 2022. This includes large software vendors that are adding generative AI to their services . Zoom uses this technology to allow users to summarize video calls. ServiceNow, which provides technical, human resources and other support to businesses, has introduced chatbots to help resolve customer IT-related questions. Adobe, maker of Photoshop, has an application called Firefly, which uses AI to edit images.
Newcomers are adding more variety. The “There's An ai For That” website has more than 12,000 applications, up from fewer than 1,000 in 2022. DeepScribe helps transcribe doctors' notes. Harvey assists the lawyers. More idiosyncratically, 32 chatbots promise “sarcastic conversations” and 20 generate tattoo designs. Fierce competition and low barriers to entry mean that some, if not many, applications may struggle to capture value.
THE CLOUD
Then there's the cloud layer. The combined market capitalization of Alphabet, Amazon and Microsoft has increased by $2.5 trillion since the AI boom began. In dollar terms, that's less than three-quarters of hardware level growth and just a quarter in percentage terms. However, compared to the actual revenues that AI is expected to generate for the big-tech trio in the near term, this value creation far exceeds that of all other tiers. That's 120 times the $20 billion in revenue that generative AI is expected to add to cloud giants' sales in 2024. The comparable ratio is about 40 for hardware companies and about 30 for model makers.
This implies that investors believe that the cloud giants will be the biggest winners in the long run. Companies' stock price to earnings ratio, another indicator of expected future profits, tells a similar story. The three big cloud companies average 29. That's more than 50 percent higher than the typical non-tech company in the S&P 500 Big Business Index, and up from 21 at the start. of 2023.
Investor enthusiasm for the cloud can be explained by three factors. First, tech titans have all the ingredients to develop world-class AI systems: truckloads of data, armies of researchers, massive data centers, and plenty of cash. Second, buyers of AI services, such as large corporations, prefer to do business with established business partners rather than untested start-ups. Third, and most importantly, big tech has the greatest potential to control every layer of the stack, from chips to applications. In addition to designing some of their own chips, Amazon, Google and Microsoft are investing in both models and applications. Of the 11 model manufacturers in our sample, nine have the support of at least one of the three giants. These include Microsoft-backed Openai, Anthropic (Google and Amazon) and Mistral (Microsoft again).
The potential profits from controlling more strata are leading even hitherto strata-specific companies to expand. Openai's in-house venture capital arm has invested in 14 companies since its launch in January 2021, including Harvey ai and Ambience Healthcare, another medical startup. Sam Altman, head of Openai, is reportedly looking for investors to finance a pharaonic $7 trillion chip manufacturing venture.
Nvidia is also becoming more ambitious. It has taken stakes in seven model makers and now offers its own ai models. It has also invested in startups like Together ai and CoreWeave, which compete with its large cloud customers. At the San Jose event, the company is expected to unveil a new GPU and perhaps even AI tools from other layers of the stack. The biggest value creator of the AI boom isn't willing to give up his crown.
(Extract from the eprcomunicazione press review)
This is a machine translation from Italian language of a post published on Start Magazine at the URL https://www.startmag.it/innovazione/intelligenza-artificiale-aziende-valutazione-mercato/ on Sat, 23 Mar 2024 06:40:57 +0000.