It seems set to be the biggest stock market debut in history, potentially worth $1.5-trillion, promising to inject the power of AI into an investment portfolio near you.
But should South Africans be thinking of investing in Anthropic, the San Francisco company founded in 2021 by siblings Dario and Daniela Amodei, who used to work at OpenAI, the company that made ChatGPT?
You’ll know Anthropic as the company that launched Claude, the large language model that seems to be everywhere, as part of its mission to “secure the benefits of AI while mitigating its risks”.
Its listing, expected in mid-November, comes at an inflection point: AI has escaped the chatbot box.
It is writing code, designing products and handling tasks that once required a procession of humans, spreadsheets and coffee. The most consequential AI products don’t feel like products at all but more like the quiet software layer inside a business, making its people faster, processes cheaper and, potentially, its jobs unnecessary.
That is why Anthropic’s listing matters beyond the usual Silicon Valley spectacle.
The company is selling a claim on a technology that could alter how businesses operate, while asking public investors to help pay for the immense computing power needed to build and run it. The question at the heart of it is: will the growth and eventual profits justify the price — and the continuing bill for staying ahead?
In its prospectus, the Claude maker reported second-quarter revenue of $11.6bn, up 1,368% on the same quarter a year earlier. First-half revenue was $16.3bn, against $1.2bn in the first half of 2025.
We don’t yet know the listing price — the S1 filed on October 26 leaves the number of shares and price range blank — but some analysts are talking of a value between $1.5-trillion and $2-trillion, more than double the $965bn of its funding round six months ago.
But the figures contain both fuel for the excitement and a reason to keep your wallet in your pocket. In particular, Anthropic reported a $693m operating loss for the first half of 2026, though its own definition of “adjusted operating income” was positive at $1.62bn. The net loss was an astounding $80.7bn, but this was driven by a noncash accounting charge linked to changes in the value of convertible notes.
The bottom line is that the losses are real, as far as accounting results go, but that doesn’t mean Anthropic spent $80.7bn running the business.
The FM spoke to four investment professionals about how South Africans might approach the listing. Their recommendations range from high enthusiasm to shorting the share to buying after the first rush of excitement has passed.
Yet there is one useful point of agreement: all four are wary of the listing process and the prospect of investors confusing a famous name, a historic technology and a first-day price spike with a guaranteed investment return.
So this is how the experts see it.
‘A complete no-go’
James Gubb is a retired CFA and author of a bearish AI market analysis. He takes a step back from Anthropic and looks at the whole AI investment boom: the enormous spending on chips, data centres and computing capacity, and the assumption that this will eventually generate returns large enough to justify it.
Gubb’s concern is that frontier AI is an unusually costly business to be in.
The leading companies must keep investing in research and computing power to stay competitive, while the technology keeps changing. If the winning position must be bought again and again through fresh investment, shareholders may find that a successful business is also a very expensive one.
“The bet is not that AI fails; it is that a commodity earns commodity returns on record-priced capital,” he tells the FM.
Anthropic’s prospectus gives his concern a fresh peg. It says computing and infrastructure expenses will increase in absolute dollars and remain its largest operating expense for the foreseeable future. It also says computing commitments will grow as the business scales and that securing enough capacity is directly linked to access to capital. That is the treadmill Gubb is watching: more users and more ambitious models require more computing power, which requires further investment.
There is, though, an important counterweight. In the first half of 2026, Anthropic’s infrastructure costs rose 283% year on year, while revenue rose 1,300%.
These compute costs — the money needed for the physical hardware — also fell as a share of revenue. This suggests that Anthropic was getting more revenue from each dollar spent on the hardware over that period.
Gubb’s warning is therefore about the long-term economics of continually funding the race, not evidence that Anthropic’s latest cost trend is already running away from its revenue.
“Don’t get me wrong, I love the product,” he says. But Claude has a six-month moat at best, he warns. “Their product is unpatentable, 80%–90% copyable/distilled at around one-20th of the cost within six months, with competition hot on its heels.”
And then, he says, there is the apocalyptic warning of the 10% chance that its product wipes out not only the user but all humanity, with fewer guardrails than milk production currently has. “This all makes it a no-go for me,” he says.
‘Wait and see’
The advice given by Piet Viljoen, the acclaimed value investor, is less dramatic but perhaps harder to follow: sit on your hands.
Viljoen sees a business environment in which the leading models look and feel increasingly similar, and new versions can overtake the leader within months. That makes it tricky to assess and the IPO difficult to price with confidence.
“These things are capital-intensive businesses with no network effect and no moat,” he says. “The one that spends the most capital leads until somebody else comes along and spends more capital.”
That is why Viljoen is in no hurry to buy Anthropic as it lists.
You didn’t have to buy Amazon when it was listed; you could have waited five years until Jeff Bezos had established himself and the business had begun to prove what it could become, he says.
Viljoen’s recommendation is to wait for the industry to sort itself out, then buy the businesses that have shown they can endure. Then investors will get a chance to “buy the stuff that drifts to the top”.
He is careful not to make governance the centre of his objection.
Anthropic has an unusual structure in which the founders control 50.1% of the shareholder vote, whatever happens, which critics say is “deviously self-serving” because it means public investors are largely powerless.
But while this structure may give investors pause, Viljoen says you should rather watch how the people in charge behave. Google had a similarly unusual control structure and proved successful.
Rather, Viljoen’s caution is about the business, its capital needs and the uncertainty of the competitive race. He is not saying that AI is unimportant or that Anthropic must fail. He is saying that urgency is not an investment thesis.
‘Positive for the world’
Paul Theron, the founder of investment firm Vestact, is hugely enthusiastic about AI as an economic force. Applications are appearing everywhere and people are only beginning to see what AI can do.
“Around the world, across the global economy, people are now finally applying themselves to how this could be used,” he says. Often, they find themselves “amazed and delighted”.
But his optimism about AI does not translate into a large, concentrated bet on Anthropic. Theron prefers businesses such as Google and Meta, which have existing platforms and cash-generating operations to support their AI spending. “It’s helluva useful to have a big balance sheet and this very strong cash-generative business so that you don’t entirely have to wing it,” he says.
Frontier labs, by contrast, are in a race where the leader also inherits a daunting bill. Theron makes the point crisply: “Once you’ve got a winner on your hands, then you in fact have a problem because you’ve got to provide the inference assets so that it’ll work.”
In other words, success creates demand for more computing capacity, and serving that demand costs money.
Even so, he expects to buy some Anthropic shares.
“I’ll probably buy a couple of hundred thousand dollars’ worth of Anthropic anyway for the hell of it, honestly,” he says.
He adds that Vestact sometimes invests alongside clients to see how something plays out. That’s an expression of interest — but it’s not the same as shifting substantial sums out of established holdings to buy a frontier lab.
When Theron asks clients to fund an Anthropic position, the money must not come from within their portfolio but should be fresh capital.
Theron’s position, then, is not “buy whatever price the IPO opens at”.
Rather, it is: AI is a major investment theme, Anthropic is interesting enough to own in a limited way, and investors should not treat a listing allocation as a prize worth chasing at any cost.
‘A huge fan’
Magda Wierzycka, the founder of asset manager Sygnia, is the most bullish of the four experts on Anthropic itself. She likes its early focus on enterprise customers and sees value in embedding it in the systems and workflows of large organisations.
Wierzycka reckons that once a company’s security, data and processes are built around an AI provider, there may be substantial friction in switching away. Asked which AI company she would choose, she says: “I would have chosen Anthropic.”
Her long-term expectation is striking. “I reckon this company has more than a sporting chance of being a $4-trillion company within a couple of years,” she says.
She describes Anthropic as a “five-year holding from now” and says she has a large personal stake.
Wierzycka’s enthusiasm is not simply a bet that Claude is a clever chatbot. She thinks trust, enterprise integration and access to computing power could matter as much as the model itself.
“I’m a big believer of first-mover advantage to a brand name you trust,” she says. She argues that the products built around the models — the agents, tools and systems that make them usable inside organisations — will be difficult to replicate.
Yet there is a distinction between Wierzycka’s bullish view of Anthropic and a recommendation to buy at whatever price. Her own route into Anthropic was through private investments before the public listing. About 25% of the Sygnia Life Attentia Fund, currently worth between R350m and R400m, is already invested in Anthropic through the private investment. The fund is also invested in OpenAI and SpaceX.
For a new investor, it’s different. She expects a familiar pattern: the initial excitement, a sharp rise and then a more sober price.
“It’s one of these ‘immediate bubble burst, huge excitement, price goes through the roof, lands down a few weeks later to a reasonable level’ situations. And that’s the point I buy and hold,” she says.
That is Wierzycka’s clearest practical advice: she may have bought early, but she does not tell a new investor to rush in on day one. She says she will not sell her own shares at the IPO and calls herself “a huge fan of Anthropic” and “that entire space”.
When the dust settles
The four positions look like a simple ladder, from short to buy. In practice, the advice is more conditional.
Gubb sees the capital requirements as a source of risk and argues that the market may be pricing in a much smoother future than the industry can deliver.
Viljoen wants investors to wait until the competitive landscape is clearer.
Theron is positive on AI and willing to own some Anthropic, but keeps the position at the margin and prefers businesses with existing cash flows.
Wierzycka has the strongest belief in Anthropic’s long-term value, yet still expects the IPO excitement to give way to a better entry point.
Yet they are all, in different ways, wary of the listing. The differences are about what happens after the dust settles: whether Anthropic’s growth can outrun the expense of staying at the frontier; whether its enterprise position gives it staying power; and whether the public market valuation will leave shareholders room to be right.
For South Africans, the first decision may be less cinematic than the technology itself. The share count and offer price are not yet set. A successful listing would still not tell investors whether the price was sensible.
The question is not whether AI will matter. It is whether Anthropic can turn that importance into returns for shareholders after paying the continuing bill for compute, talent and the next leap forward.