Tech companies like to make two big claims about the future of artificial intelligence: the first is that the technology will bring about a revolution akin to the advent of fire, nuclear weapons, and the Internet, and the second is that artificial intelligence will cost untold amounts of money.
Silicon Valley has already driven tens or even hundreds of billions of dollars of spending on AI, and companies want to spend even more. Their reasoning is simple: These companies have decided that the best way to make generative AI better is to build bigger AI models. And that’s really Really Funding the data centers and related infrastructure that generative AI depends on would require resources and costs equivalent to a lunar mission or an interstate highway system. With a product as important as fire, they say, any expense is worth it. OpenAI CEO Sam Altman described his company as “the most capital-intensive startup in the history of Silicon Valley.” Dario Amodei, CEO of rival startup Anthropik, predicts that it will cost $100 billion to train a single AI model (such as GPT-6) by 2027. The global data center buildout over the next few years will require $100 billion in new revenue streams. Trillion Moody’s July report said money is flowing in from technology companies, utilities and other industries.
Now, some in the financial world are questioning whether all of this investment will pay off: OpenAI could lose up to $5 billion this year, nearly 10 times what the company lost in 2022. informationOver the past few weeks, analysts and investors from some of the world’s most influential financial institutions, including Goldman Sachs, Sequoia Capital, Moody’s and Barclays, have published reports questioning whether huge investments in generative AI will pay off. “If you’re justifying an investment of more than $1 trillion, you have to be really careful,” Jim Covello, head of global equity research at Goldman Sachs, told me. [AI] “We need to solve complex problems and do things that we couldn’t do before,” he said, something that today’s mainstream AI models largely fail to do.
Judged by almost every criterion other than the revolution brought about by electricity and the Internet, generative AI is already doing extraordinary things: accelerating drug development, solving difficult math problems, generating amazing video clips, and so on. But it remains unclear what uses of this technology could actually generate profits. Currently, AI is generally good at performing existing tasks faster and cheaper than humans, such as writing blog posts, coding, and translating. But there is a limit to the value that can be gained from increased efficiency, and while it can stimulate the current economy, it cannot create a new one. For now, Silicon Valley may only be functionally replacing some jobs, such as customer service and form processing, with historically expensive software, but this will not lead to widespread economic transformation.
Even if generative AI has yet to significantly change many people’s lives, proponents say that as the technology advances, it will solve long-standing scientific problems, greatly increase productivity, and create entirely new sectors of the economy. In just a few years, various generative AI models have evolved from writing simple sentences to writing entire essays. Many investors and analysts are getting in on the action. Tony Kim, head of technology investing at BlackRock, the world’s largest asset manager, told me that AI will create one of the biggest technological upheavals in history. “Previous industrial revolutions weren’t about intelligence,” he said. “Here, we can manufacture intelligence.” McKinsey estimates that generative AI could eventually add almost $8 trillion to the global economy each year. One JP Morgan researcher recently said that AI will be more influential “than the Internet or the iPhone.”
Amid all this hype, we must remember that this future is not guaranteed. MIT economist Daron Acemoglu has found that many of the productivity gains expected from AI may not only be overestimated, but also premature. AI products have significant flaws, such as a tendency to fabricate disinformation, which may mean they cannot be used or can only be deployed with strict human oversight in certain environments, such as courts, hospitals, government agencies, and schools. Much of the human labor is manual, and it is difficult to replace it with software. There is much debate over whether scaling up AI models will continue to produce significantly better results. And while likening AI to an atomic bomb is inspiring, it is not a roadmap for sustainable business models. Generative AI is often touted as a truly groundbreaking technology, but it may be more like blockchain. Blockchain is a very expensive tool that is destined to fall short of its promise of fundamentally transforming society and the economy.
But tech companies are spending as if such transformative uses are a given. Barclays researchers recently calculated that tech companies are collectively paying for enough AI computing infrastructure to eventually run 12,000 ChatGPTs. Silicon Valley could well produce a lot of hit-generating AI products like ChatGPT, but “it probably won’t be 12,000,” the researchers wrote. And even if it did, there would be nowhere near enough demand to actually use all of those apps to make money. Sequoia Capital partner David Kahn sees the funding gap in a different light: Some of the biggest tech companies would need roughly $600 billion in annual sales to recoup the money they’re currently spending on AI data centers, and there’s currently a shortfall of roughly $500 billion of that, he said.
Tech advocates have responded to criticism that the industry is putting too much money into it too quickly with something like religious dogma. “It doesn’t matter how much you put into it,” Altman said. “It really doesn’t matter.” In other words, the industry is asking the world for a trillion-dollar tautology. AI’s world-transforming potential justifies any amount of resources being put into it, because AI advocates will put any amount into changing the world with AI. Kim, the BlackRock AI optimist, captured this sentiment perfectly: “You have to believe that these technologies and capabilities are going to keep on advancing, but that requires significant investment,” Kim told me.
The tech industry has long walked a tricky line between grand visions and grandiose delusions. Often, the only difference between the two has been what pays off in the long run. But especially in the age of AI, the lack of clear evidence of a healthy return on investment may not matter. Unlike the companies that collapsed in the dot-com bubble of the early 2000s, Big Tech has been largely okay with spending exorbitant amounts of money. But at some point, especially if the economy worsens, the enormous bank accounts of Microsoft, Google, Amazon and Meta could start to dwindle. If these companies’ balance sheets become unstable, shareholders and investors may lose enthusiasm, Raj Joshi, a senior vice president at Moody’s Investor Service who analyzes the technology sector, told me.
Even if generative AI is a bubble, that doesn’t mean all this investment is for naught. Chatbots are unlikely to generate $600 billion in annual revenue in the next few years, but that doesn’t mean other kinds of AI won’t transform society by 2040, or even the decade after that. This spending frenzy may be too intense and premature. With Amazon, Google, Meta, and Microsoft spending hundreds of billions of dollars building out data centers, that means future tech startups could potentially use those computing resources at a lower cost.
Attitude matters more than any product right now. Technology companies are teeth Their proof that AI can bring profits. And perhaps even more important than a messianic faith in AI in Silicon Valley is a deep fear of being left behind. “In tech, this is driven in part by the feeling that no one wants to be left behind. No one wants to be seen as lagging behind,” Joshi says. Amazon, Google, Meta and Microsoft are protecting their empires. Putting all our efforts into AIIf you think about it, Or someone elseTheir actions betray a “sense of desperation,” Kahn writes: “If we don’t act now, we’ll never get another chance.” Driven by a mixture of unshakeable confidence and all-encompassing fear, huge amounts of money will continue to flow into AI for the foreseeable future.
