The AI ​​industry has invested $150 billion into Nvidia chips, but it may never recoup its investment

Companies are spending huge amounts of money to buy the hardware required to build and operate generative AI chatbots as if it will become obsolete.

Why? According to an analysis by Sequoia Capital, cloud service providers and others could invest $300 billion in AI hardware in 2024. Tom’s Hardware. Nvidia will get about half of the investment, with providers of servers, cooling hardware, energy and other data center components getting the other half.

Will these investments pay off? The answer depends primarily on two things:

• How do you define rewards?
• Will there be a killer app for generative AI?

If that payoff is defined as making enough profit to offset the investment, Sequoia estimates that tech giants like AWS would need to generate $600 billion in generative AI revenue, but such a payoff could be decades away.

To put Khan’s estimate into perspective, major tech companies such as Alphabet, Amazon, Apple, Meta and Microsoft are estimated to have budgeted $400 billion in capital expenditures, mostly for AI-related hardware and research and development. economist.

With these five tech giants adding $2 trillion to their market capitalization over the past year, investors are expecting their sales to increase by $300 billion to $400 billion. economist report.

There’s a big gap between the additional revenue from generative AI that their rising stock market capitalizations suggest and what the five companies will deliver in 2024. “Even the most bullish analysts think Microsoft will only make around $10 billion in generative AI revenue this year,” he says. economist.

If that payoff is defined as enough additional revenue to sustain better-than-expected growth and boost the tech giants’ stock prices, that payoff could come in the next few years.

In my view, such an outcome will only come about if a killer app for generative AI emerges, much like electronic spreadsheets were to personal computers, or the iTunes store was to iPods.

Sequoia forecasts AI hardware investments

Sequoia Capital analyst David Kahn estimates that AI companies would need to generate roughly $600 billion in revenue per year to cover the costs of AI hardware.

AI companies that have invested include “AWS, Google, Meta, Microsoft, and many others,” and the company is noted for spending the money on AI hardware for applications such as OpenAI’s ChatGPT. Tom’s Hardware.

To arrive at the $600 billion revenue figure, Khan made the following assumptions:

• Buying an Nvidia GPU. By the fourth quarter of 2024, Nvidia will generate $150 billion in annual GPU revenue, he predicted.
• Purchasing other AI hardware. Other data center costs (energy, buildings, backup generators, etc.) will require investments equal to twice the cost of Nvidia’s AI GPUs. According to Cahn, AI data center hardware costs will total $300 billion in the fourth quarter of 2024.
• Software margin. Software has a 50% profit margin, he noted.
• The software you need to get a return on your AI hardware investment. Khan estimated that companies would need $600 billion in newly generated AI revenue to generate the $300 billion in profits needed to pay off the $300 billion in costs of AI data centers.

While this model is a useful start, we believe there is room for improvement. Here are some suggestions:

• Extend payback period. Kahn assumes companies can recoup their AI hardware investments in one year; a more typical payback period is three to five years. “If we assume an aggressive three-year period, $600 billion comes to $200 billion per year,” analyst Jon Peddie wrote in a July 8 email.
• Slowing AI hardware investment. Kahn assumed that GPU spending growth will increase 66% by the end of 2024. At some point, the growth rate will slow as demand for AI chatbots slows, and this slowdown is likely to happen unless a killer app for generative AI emerges.
• Different groups of companies are investing in AI hardware, each with different profit margins. Kahn’s model assumes that software revenues will generate the profits needed to recoup investments in AI hardware. Indeed, perhaps the largest group of companies investing in AI hardware are cloud service providers, whose profit margins are lower than those of software companies.

That said, Khan estimates AI revenues will be $500 billion less than what’s needed to recoup $600 billion in AI hardware investments.

The gap assumes that AI revenue will reach $100 billion in 2024. Google, Microsoft, Apple and Meta are estimated to generate $10 billion in annual revenue from AI each, while companies like Oracle, ByteDance, Alibaba, Tencent, X and Tesla are projected to generate $5 billion each. Tom’s Hardware.

based on Economist With generative AI revenue estimates, the actual difference could be even larger.

Finally, Kahn speculated that if this $500 billion gap isn’t filled, the bubble will burst.

Companies that invest in AI hardware aren’t financing their purchases with debt, so there’s no urgency to pay back their investment. Put simply, AWS and other cloud service providers are paying cash for AI hardware.

These companies aren’t under pressure from investors to withdraw cash because many of them, including Microsoft and Meta Platforms, are so profitable that investors are likely not worried about them burning through cash.

Rather, the market may be more concerned about the risk that these companies will not be able to create or keep up with the fastest-growing market opportunities.

Will generative AI value networks provide a return on investment?

Various companies in the Generative AI Value Network are purchasing AI hardware, including:

• AI cloud service provider.
• AI data centers;
• A large-scale language model supplier.
• Data management, application performance monitoring and network technology provider.
• According to my new book, generative AI application developers should: Brain Rush.

Each of these groups has different levels of investment in AI hardware, different cash balances, and varying average profit margins. For example, in 2023, Microsoft and Meta had cash balances of $111 billion and $66 billion, respectively, high profit margins of 34% and 29%, respectively, and relatively modest debt. The Wall Street Journal.

Amazon’s AWS is projected to generate $25 billion in revenue in the first quarter of 2024, with an operating margin of 38%. Registry Amazon has a healthy cash balance, but a much lower net profit margin and large long-term debt. Specifically, analysts note that in 2023, Amazon has an $87 billion cash balance, a 5.3% net profit margin, and $136 billion in long-term debt. journal.

Given the wide difference in the relative profitability of each company, the additional generated AI revenue that Microsoft and Meta need to recoup their investments in AI hardware will likely be smaller than the additional AI revenue that Amazon needs.

Some of these companies report how much capital they are investing in AI hardware and other assets, but they don’t quantify their AI revenue in dollar terms, and it could be a long time before AI revenue is high enough to be reported as a separate line item.

History has shown us that it may take years for a killer app for generative AI to emerge and generate significant revenue: for example, the first PC, the Kenback 1, was introduced in 1971, but it wasn’t until 1979 that the first electronic spreadsheet, Visicalc, hit the market, giving many people a strong reason to buy a PC.

Another killer app was the iTunes store. In 2001, Apple released the iPod, but it didn’t see any significant demand until the iTunes store launched in April 2003. CNN I got it.

The killer app for generative AI will rise to the top of the value pyramid, as we wrote about in our June 2024 report. Forbes Post. The pyramid consists of three levels.

• Overcoming creator’s block. At the bottom of the pyramid, there are many ways people can use AI chatbots to initiate activities, such as writing emails or reports, taking photos or videos, coding software, etc. By helping people overcome creator’s block, AI chatbots can make people more productive. This is the easiest way for businesses to get value right now. However, any benefits from using AI to overcome creator’s block are temporary, as other businesses can do the same.
• Improve customer service and sales productivity. The middle tier of the value pyramid can increase a company’s productivity. For example, in the staffing industry, generative AI can drastically reduce the number of candidates that recruiters send to a company. And AI chatbots can help companies resolve customer questions more quickly. It’s unclear whether companies will pay a high enough price for such AI applications to allow their suppliers to grow profitably.
• Create a new growth curve. The top of the pyramid is the most valuable. AI technology suppliers aim to create a new growth curve by accelerating their customers’ growth. Generally, the faster growth that generative AI drives, the higher the stock prices of companies deploying AI-powered applications, which in turn increases companies’ willingness to pay a higher price for the technology.

There is a lot more real activity going on at the first and second levels of the pyramid than at the third level. Brain Rush Note.

The generative AI bubble hasn’t burst yet

Bubbles are exacerbated when companies grow by borrowing. So far, the most significant debt that companies are taking on for generative AI is so-called technical debt, which refers to “all the expected (and unexpected) costs that companies incur when adopting new technology,” he noted. job.

Examples include “the costs of fixing software bugs that were not addressed at launch, installing patches after unknown vulnerabilities are discovered, and modernizing legacy technology infrastructure.” job report.

While it doesn’t appear the company is taking on a lot of debt to buy generative AI hardware, investors should be concerned about what I call “revenue expectation liability.”

Simply put, unless the billions of dollars being invested in AI hardware translate into a significant increase in revenue growth, companies won’t meet investor expectations.

Such disappointment could drag down shares of Microsoft, Google, Amazon and Meta Platforms, as well as their GPU supplier, Nvidia.