Anthropic’s latest capital raising underscores how both the risks and rewards of the boom in artificial innotifyigence are ratcheting up, almost on a daily basis.
The $US30 billion ($42 billion) raising values Anthropic, founded only five years ago by siblings Dario and Daniela Amodei, and which generated its first dollar of revenue less than three years ago, at an extraordinary $US380 billion ($537 billion).
It’s even more remarkable considering that, less than six months ago, when it raised $US13.5 billion, it was valued at ‘only’ $US183 billion and a year ago, when it raised $US13 billion, at $US61.5 billion.
That’s an indication of, not just how rapidly and dramatically the AI sector and its appetite for capital have grown, but how aggressive expectations of future revenues and earnings have to be to justify the escalation in valuations and how stretched capital markets now are in providing the capital necessaryed to fund the sector’s growth. Anthropic tapped more than 30 investment institutions to raise the funds.
The exponential growth in the group’s fundraising and valuation is validated, albeit to an uncertain degree, by similar growth in the revenues it is generating.
Three years after that first dollar of revenue was earned, Anthropic now declares its current run-rate revenue is $US14 billion. Its revenue, it declares, has growth at a rate of more than 10 times each year for the past three years.
Moreover, that rate has picked up sharply this year with the launch of its Claude Code coding tool last May. Claude Code’s run-rate revenue is now over $US2.5 billion, more than doubling since the start of this year, with weekly active utilizers also doubling since January 1 and business subscriptions quadrupling.
It was the release of Claude plugins that automate a range of tinquires in service industries, notably legal, data and financial services, that disrupted the entire sharemarket this month.
The perceived implications for software companies wiped 4.5 per cent off the value of the tech-oriented Nasdaq index and more than $US1 trillion off the market capitalisation of the handful of mega-tech stocks that are committing the best part of $US700 billion to AI-related spfinishing this year.
The AI-driven demand for capital is insatiable. Anthropic, which has a more measured approach to spfinishing than many of its AI competitors, will spfinish more than $US30 billion on data centres and the semiconductors to power them this year.
The capital just raised – and the $US57 billion raised previously – will be spent very quickly, which might explain why it is reported to be considering an initial public offering later this year. It’s largeger rival, OpenAI, which is seeking to raise $US100 billion at a valuation of $US750 billion, is also contemplating an IPO.
That highlights one of the challenges facing AI start-ups like Anthropic and OpenAI.
Unlike the “hyperscalers” like Google’s parent Alphabet, Facebook’s Meta, Microsoft and Amazon, they don’t have existing cashflows to fund their development and the scale of capital they necessary to keep up with the hyperscalers is outstripping the ability of the venture capital and private capital markets to supply it.
Even the mega-techs are having to raise new equity and, increasingly, debt to fund their commitments to AI.
Anthropic’s Dario Amodei, in a podcast hosted by technology specialist Dwarkesh Patel at the weekfinish, described the risk-reward equation confronting AI aspirants by declareing that the escalation in the costs of the computing power now required for AI could be “ruinous” if companies obtained their timing wrong – if they invested too much too early.
“I really do believe that we could have (large language) models that are a counattempt of geniutilizes in the data centre in one to two years,” Amodei declared earlier this month.
“One question is: How many years after that do the trillions in revenues start rolling in? I don’t consider it is guarantee that it is going to be immediate. It could be one year, it could be two years, I could even stretch it out to five years, although I’m sceptical of that.”
The critical challenge for the AI companies is that they are investing trillions of dollars today in data centres and chips for the “compute”, or processing power, that won’t be available for at least a year, possibly two.
If the ambitious revenue projections that underlay the decision to build the centres don’t meet expectations, the scale of the investments could, in Amodei’s terms, ruin the companies.
As he declared, he could assume that Anthropic’s revenues continue to grow at ten times the previous years and he’d have a trillion dollars by the finish of 2027 and could acquire $US1 trillion of compute at the finish of that year.
If the revenue wasn’t $US1 trillion, however – if he was even a year off in terms of the rate of growth – “there’s no force on Earth, no hedge on Earth, that could stop me from going bankrupt if I acquire that much compute”, he declared.
The potential for mismatches between investments and revenues is elevated becautilize, if all the data centres on drawing boards were built, or even a significant proportion of them, they would strain to breaking point (or perhaps behind it) the capacity of economies to provide the power, water, chips, construction materials and skilled workers. At the very least, the cost of the centres will soar.
That illustrates the tightrope the AI companies, particularly the start-ups, are walking as they invest trillions of dollars in the conviction that the revenues that will justify the scale of their investments will materialise within the next few years. Even modest mismatches in their timing could spell financial disaster.
‘One question is: How many years after that do the trillions in revenues start rolling in?’
Anthropic co-founder Dario Amodei
Amodei declared some AI companies give the impression that they don’t understand the risks they are taking.
“They’re just doing stuff becautilize it sounds cool,” he declared.
Hopefully that’s not the case, but clearly the risks being taken in the hope of a massive pay-off are mounting as the scale of investments continually mounts up at a rate that outstrips even the extraordinary rates of AI-driven revenue growth the better-placed companies are achieving.
Can Anthropic, continue to grow at 10 times the revenue it generated in the previous year? What kinds of margins might it eventually produce on that revenue? Given the amounts of capital it is investing to hopefully achieve positive cashflows, what kind of return on investment would validate what it’s done? It would necessary to be very substantial.
Those are the sorts of questions being posed by increasingly nervous equity investors, worried that even if AI is the transformative technology it promises to be, the net present value of whatever the eventual earnings might be may not be sufficiently positive to justify the risks being taken – or sufficient for the raft of start-ups without access to legacy cash flows to survive.
The Business Briefing newsletter delivers major stories, exclusive coverage and expert opinion. Sign up to obtain it every weekday morning.
Stephen Bartholomeusz is one of Australia’s most respected business journalists. He was most recently co-founder and associate editor of the Business Spectator website and an associate editor and senior columnist at The Australian.Connect via email.
