Before I get into this week’s post, it occurred to me that some of you might be frustrated at how often I write about AI — things like whether it makes sense to think of AI engines as conscious, whether we should be afraid of it destroying humanity as we know it, whether it’s bad that people sometimes fall in love with chatbots, whether AI data centres use too much electricity and water, etc. Perhaps you are sick of hearing about AI all the time! Or it’s possible that you have already made up your mind that it is bad. For me, this newsletter is a way of thinking out loud about this type of thing, which is why I rarely come down hard on one side (something I was chastised for recently by a reader). Also, the name of the newsletter is The Torment Nexus (explanation here, for those who aren’t aware of the reference) and what could be more tormented or nexus-like than artificial intelligence? Anyway, I encourage you to stick with me as I explore some of these AI-related questions, unless it is just too much for you (either the topic or my indecisiveness on it), in which case bon voyage!
When reading about the recent funding by Anthropic — which appears to have won the race with OpenAI for who is going to do an initial public stock offering first — I confess I find it hard to wrap my head around the numbers being floated, whether in Anthropic’s recent funding round, its unofficial results, or the coverage of its proposed IPO. This company — which according to New York Times writer Kevin Roose was just “a 160-person start-up in Jackson Square with no products and no revenue” three years ago — is expected to have a market value of close to $1 trillion US dollars, unless something drastic happens between now and issue time. I’ve been writing about technology stocks since Netscape first went public in 1995, and this is unlike anything I’ve ever seen, not just in terms of magnitude but in terms of the speed with which Anthropic has reached this (theoretical) valuation. The company — which was created by Dario Amodei, his sister Daniela, and a number of other former OpenAI staffers — is less than six years old.
Just a little over two months ago, OpenAI — which jump-started the AI gold rush with the release of ChatGPT in 2022 — announced that it had raised $122 billion in a funding round that put its value at $730 billion, a number that the New York Times notes it took roughly a decade to achieve. Anthropic has eclipsed it in half the time. Since I’ve already mentioned 1995, which marked the beginning of one of the great tech-stock IPO runs in history (at least history up to that point) it’s worth noting that according to technology analyst Benedict Evans, the $965-billion theoretical value of Anthropic’s IPO is more than the total market cap at issue of every single venture-backed IPO in the USA from 1995 to 2000. That encompasses the entirety of the dot-com bubble, which at the time was described as a completely unjustified orgy of stupidity (among other things).
Note: This is a version of my Torment Nexus newsletter, which I send out via Ghost, the open-source publishing platform. You can see other issues and sign up here.
Anthropic’s theoretical value is more than three times as much as as Time Warner vaporized after merging with AOL, and that was a legendary bonfire of value I thought would never be exceeded in my lifetime. So the question that springs to my mind (and perhaps to yours), is “What the hell are Anthropic investors smoking?” I know greed is a powerful motivator, but this is ridiculous. Or is it?
According to Anthropic’s self-reported numbers, its “run rate” — what its annual revenues would be if its most recent month of business was multiplied by 13 — is currently $47 billion. In April it was just $30 billion, meaning it has grown by more than 50 percent in a single month, and last year its annualized run rate was just $9 billion, for an increase of 400 percent. If anything, the latest run rate is likely to be significantly understated given the company’s recent blistering pace of growth, something that the investors who paid for a piece of the latest $65 billion funding round (including Google and Amazon) are no doubt counting on. This rising tide also appears to be lifting many boats, as investment types like to say: the share price of Nvidia climbed because of its investment in Anthropic, and so did the stocks of chip companies, on the assumption that the AI company will need more cloud servers. PC maker Dell climbed for a similar reason.
Note: In case you are a first-time reader, or you forgot that you signed up for this newsletter, this is The Torment Nexus. Thanks for reading! You can find out more about me and this newsletter in this post. This newsletter survives solely on your contributions, so please sign up for a paying subscription or visit my Patreon, which you can find here. I also publish a daily email newsletter of odd or interesting links called When The Going Gets Weird, which is here.
Blowing bubbles
As The Atlantic noted recently, the AI sector was definitely looking like a rapidly-inflating bubble as recently as a few months ago — like the dot-com bubble, or the rapid inflation in anything to do with fibre-optics before that, or the railway bubble of the 1800s, or maybe even the notorious frenzy over tulips in the 1600s (which didn’t happen quite the way it is sometimes portrayed) or the South Sea Bubble, where even Isaac Newton lost a pile of dough. There was a lot of hand-wringing about the circular financing arrangements that companies like OpenAI and Nvidia and Microsoft were signing seemingly every few days, agreeing to pour hundreds of billions of (borrowed) dollars into each other’s pockets. Even OpenAI cofounder and CEO Sam Altman admitted last year that investors might be “overexcited about AI” (at the time, I pointed out that bubbles like the fibre-optic frenzy can in some ways be a good thing). Here’s The Atlantic:
Today we’re in a very different world. Software developers are adopting AI tools en masse and reporting astronomical productivity benefits. The worry that the country is building too many data centers now coexists with the fear that we won’t have enough of them to satisfy the public’s growing appetite for these products. And the company previously known as OpenAI’s junior competitor has become possibly the fastest-growing business in the history of capitalism. Anthropic’s revenue is increasing faster — much faster — than Zoom’s during the pandemic, Google’s during the early 2000s, and even Standard Oil’s during the Gilded Age. If the company’s current growth rate were to continue, then by early next year it would be taking in more money than any other company in the world.
So what has changed since everyone was convinced we were in a giant AI bubble? For one thing, Claude Code was released — a special version of Anthropic’s AI engine that is designed to help with programming. The company launched it in November, and according to analysts it has been the main driver of the company’s revenue growth (it’s worth noting that Anthropic reports its revenue as a gross amount, including the money it pays to suppliers and partners, while OpenAI — which is said to have an annualized run rate of about $30 billion — reports revenue minus the payments it makes to partners). Anyone with a programming or automation project can give Claude Code an assignment in plain English, give it access to the tools it needs, and then have an entire project, app, service or feature built literally in minutes. Ethan Mollick, a professor at Wharton who specializes in AI, got Claude Code to build a complete, playable version of an online role-playing game, complete with character guide, manuals, and notes.
Other companies have also released updates to their own coding tools, such as OpenAI’s Codex and Anysphere’s Cursor, and the combination of those three has catapulted AI into a new era that some call the “agentic” era, where software agents or assistants like Claude Code can finish complete tasks with minimal input. “This really was a step change,” Mollick, co-director of the Generative AI Lab at the University of Pennsylvania, told The Atlantic. “For years now, we’ve been in an era of chatbots that mostly just say things. Now we’ve officially crossed into the era of agents that can actually do things.” Boris Cherny, the creator of Claude Code — which started as just an in-house project but eventually came to be used by almost everyone at the company — told Casey Newton at Platformer that he hasn’t written a line of code in more than six months, and that for the kind of work he does inside Anthropic, coding is effectively “solved.”
Obviously, just because Anthropic’s coding agent can pump out Python scripts in seconds or build apps on the fly doesn’t mean entire companies can hand over their software operations to AI. It still makes mistakes — although a lot fewer than it used to — and the code needs to be checked by humans not just for accuracy but to make sure it works well with other things a company is doing. But if you think about the number of companies both in tech and outside that rely on programming of various kinds, you can understand where the $47 billion in revenue that Anthropic reported might be coming from. In what AI supporters might see as an amazing statistic and AI skeptics might see as a frightening one, Anthropic said in one of its regular updates that 90 percent of the code that runs Claude is being written at least in part by Claude itself.
Punch-drunk on FOMO
According to The Atlantic, researchers at Goldman Sachs interviewed staff at 40 software companies about their AI use in mid-April, and found that many were overrunning their initial budgets for AI tools by orders of magnitude. Anthropic recently announced that it was throttling or limiting its customers’ use of its coding tools during what it calls “peak hours,” because it doesn’t have enough infrastructure to accommodate them all (hence the fact that people are also investing in server companies and data-center owners). OpenAI shut down Sora, the AI-powered video-generation app that was a big hit not that long ago, because it too is resource-constrained. The chips that server farms and data centers use are in such high demand that even Nvidia’s fourth-best AI chip, released back in 2022, costs more today than it did three years ago.
It’s important to point out that talk of a bubble hasn’t gone away completely — some financial experts are convinced that Anthropic could be the leading edge of a massive bubble that could burst and imperil the entire economy. The Telegraph quoted a number of financial analysts as saying it reminded them of Global Crossing, a fibre-optic telecom company that was once worth $47 billion and eventually went bankrupt when demand for fibre dried up (partly because of the dot-com crash). Joachim Klement, an analyst at stockbroker Panmure Liberum, warned about what he called the “impossible maths” surrounding some of the major AI players, and their ballooning valuations. Softbank, the Japanese investment company known for big bets, recently passed the market value of automaker Toyota, based largely on its AI holdings.
Michael Burry, the investor who became famous for foreseeing — and betting against — the massive inflation in US house prices and the leverage that investment banks had in the early 2000s housing bubble, says that he doesn’t see how Anthropic is worth anything even close to $1 trillion, despite its rapid revenue growth. Burry said in a subscriber chat that he’s skeptical the AI startup will ever warrant that price tag: “There is no guarantee, and not even a strong likelihood, that Anthropic is long-term worth anywhere near $1 trillion,” he wrote. According to Business Insider, Burry added that Anthropic’s business of developing cutting-edge AI models is “far too expensive, too much brute force,” as over time, he expects computing power “will be commoditized, like internet use.” What is happening now is “a false demand signal,” he wrote, echoing his recent warning that the so-called “tokenmaxxing” trend won’t last.
An author who won a Nobel Prize for his book on the stock-market mechanics that led to the Great Depression wrote in Fortune that the AI spending boom “terrifies me.” Here’s how he put it:
The buildout of the fiber-optic network from 1995 to 2000, which provided the backbone of the modern internet, saw IT investment climb from 3% of GDP to nearly 5% at its peak. The investment surge of the Roaring Twenties — driven by the mass adoption of the automobile, the expansion of the electricity grid, and a construction boom in housing — was another. Finally, and most significantly, there was the railroad boom of the late 19th century. The fear is that, if the past is anything to go by, the AI boom will follow a similar arc to these other technology-driven infrastructure booms: a flood of speculative capital will flow in, leading to massive overinvestment, asset price bubbles and ultimately a crash as euphoria collides with a disappointing reality.
My friend and former boss Om Malik wrote about the AI funding in his newsletter, and said the numbers being tossed around “are so large and so ridiculous that you cannot tell whether anything is real.” Investors, even professional ones, are buying press releases, not financial results, he said. You might believe that investors pouring billions into Anthropic are doing their due diligence, he adds, but “my experience says they are not. There is nothing more dangerous than an investor punch-drunk on FOMO.” Anthropic may be growing faster than any other company in recent memory, and its products may be used by every programmer you know every minute of the day, but will that always be the case? Probably not. But if you are planning to bet against it in some way, please remember the immortal words of John Maynard Keynes, the father of macroeconomics: “The market can remain irrational longer than you can remain solvent.”
Got any thoughts or comments? Feel free to either leave them here, or you can also reach me on Twitter, Threads, BlueSky or Mastodon. And thanks for being a reader.