View in browser
During prime-time football last Sunday, OpenAI debuted its biggest ad campaign yet, emphasizing three possible uses for ChatGPT: making dinner recipes, creating workout routines, and planning road trips. As romantic as pasta à la ChatGPT sounds, the simplicity of these chatbot queries raises a question: Is the company really worth the hundreds of billions of dollars its investors say it is?
The market certainly thinks so. Yesterday morning, OpenAI became the most valuable privately held company in the world, worth $500 billion. (OpenAI has a corporate partnership with The Atlantic.) That’s thanks to a new stock sale and recent agreements with the chipmaking giant Nvidia, which plans to invest up to $100 billion in OpenAI, and the cloud-computing company Oracle, to which OpenAI will pay $300 billion for a massive chunk of its computing power. The result is a curiously recursive cash flow, because Oracle already committed to buying chips from Nvidia in May. In other words, Nvidia will pay OpenAI, which will pay Oracle, which will pay Nvidia.
All this while OpenAI is reportedly set to hemorrhage more than $1 billion this year. Much like the tech unicorns of decades past (think Uber, WeWork, and Spotify), the company is confident enough in its eventual dominance that it’s willing to operate at a loss in the near term. ChatGPT’s arrival, in 2022, was revolutionary, and it was easy to see why its parent company secured such astronomical investments and major partnerships during the initial funding frenzy. An AI chatbot that does homework, writes emails, ships code, and doles out free therapy? There’s a reason it has, according to the company’s tally, 700 million active users each week.
In the years since, things have gotten a bit more complicated. AI still “hallucinates” wildly in its responses (in other words, it tends to make things up), and there are now numerous reports about the dangers it poses to users’ mental health: The Wall Street Journal wrote in August that sycophantic ChatGPT became a “trusted sidekick” to a paranoid man who eventually killed his mother and himself (OpenAI preempted the article with a blog post about its focus on “people using ChatGPT in the midst of acute crises,” and told the Journal that the bot had encouraged the man to seek professional help).
But none of this has slowed the breakneck speed of the AI race. In the case of OpenAI in particular, the attitude is growth at all costs, a path to long-term dominance that those betting on OpenAI say could mirror what Amazon pulled off in the early 2000s. (OpenAI is now also dipping a toe in social media with a powerful new video-generation app, indicating yet another area of interest for the company.) On their own, none of these features is generating the kind of return that would justify OpenAI’s sky-high valuation—at least not yet. But if the company gets its way, preparing a recipe or planning a vacation without an assist from the chatbot might start to feel alien, similar to the way people have come to rely on real-time directions from Google Maps or blazing-fast grocery delivery from Instacart. Whether enough people will choose to pay directly for ChatGPT to meet investors’ expectations is an open question.
Generative AI is still very much a bet on the future; concrete returns are potentially years away. If the gamble doesn’t pay off, the market for AI could collapse. Many have compared the hype over AI to early internet hype and the infamous dot-com bubble of the late 1990s. Altman and Meta CEO Mark Zuckerberg, another true believer in AI, have acknowledged that a hypothetical AI bubble could pop too. As of now, consumer spending doesn’t even come close to meeting institutional investment: Research from the VC firm Menlo Ventures suggests that only about 3 percent of AI users pay for any sort of service—that’s about $12 billion, while the industry is projected to spend up to $3 trillion by 2028.
But to hear a venture capitalist tell it, the promise of AI is powerful enough on its own to justify the sort of long-term value that would preclude a bubble. The investors going all in on early-stage AI start-ups are broadly confident that AI can save companies time, and therefore money, across industries. The problem is figuring out how to implement that technology in a way that can juice profits without confusing or impeding workers. Researchers at McKinsey recently put out a report diagnosing a generative-AI “paradox”: the contradiction between companies’ high AI-adoption rates and how little that seems to improve their bottom line. AI chatbots can still do only so much—human beings are often needed to correct their work. According to a report published by S&P Global this year, 42 percent of the companies that have tried out AI pilot projects abandoned the majority of them.
The result is a disconnect between what goes in and what comes out. OpenAI is worth more than Norway’s GDP; its employees are raking in princely sums, and its expenditures on everything from development to operations to raw computing power are, by any metric, extreme. But there’s no guarantee that AI chatbots will be the revenue-inflating productivity accelerators their biggest boosters hope they will be. There’s also no guarantee that chatbots will offer the kind of life-changing support for everyday users that these companies claim. Still, the world’s biggest venture capitalists and tech conglomerates are willing to take that chance. Potential regulatory hurdles, implementation bottlenecks, bubble doomersayers, social backlash—none of that seems to matter, at least not yet.
Related:
