By Khumbulani Kunene
Anthropic is preparing for one of the largest initial public offerings (IPO) in history, targeting proceeds of around $84.6 billion at the midpoint of the price range. The company is selling 801.4 million new Class A shares, while existing shareholders are selling a further 23.6 million. Immediately after the listing, Google, NVIDIA, AMD, Broadcom, Micron, Microsoft and TSMC have agreed to buy $26 billion of non-voting Class N shares at the IPO price.
Business model
Anthropic develops frontier artificial intelligence (AI) systems under the Claude brand and has experienced one of the fastest enterprise adoptions of any technology in history. The company was founded in 2021 by a team of former OpenAI researchers, including siblings Dario Amodei (chief executive officer [CEO]) and Daniela Amodei (president). Its customers span governments, healthcare, education and financial service sectors across the world. Anthropic is incorporated as a public benefit corporation (PBC), meaning it is legally required to balance shareholder returns with its stated mission to "responsibly develop and maintain advanced AI for the long-term benefit of humanity".
Anthropic launched its main product, Claude AI, commercially in the first quarter of 2023 and has since grown quarterly revenue from zero to $17 billion in the three months to 30 September 2026. Revenue is mainly consumption based, meaning customers pay per usage (tokens processed, model calls, etc.) to build Claude into their own products, where larger models cost more than smaller ones. Consumption-based fees made up 79% of revenue in the nine months to September 2026, with subscriptions contributing the remaining 21%. This model allows revenue to scale quickly as customers embed Claude into business-critical workflows, but it also makes revenue more sensitive to changes in usage than a traditional software subscription. Sales through cloud service provider (CSP) marketplaces, such as AWS and Google Cloud, made up ~44% of revenue in the first nine months of 2026, highlighting both the reach of these partnerships and Anthropic's reliance on them.
Claude AI
Claude is a family of large language models (LLM), which are AI systems trained on large amounts of text and other data, so they can read, write, analyse, reason, write code and work with documents and images. Claude comes in tiers that trade off capability against speed and cost: Haiku (smaller and faster), Sonnet (balanced), and Opus (more complex, deep reasoning), with a newer Mythos tier above these.
Anthropic delivers Claude to customers through three channels:
Claude's scale and growth creates self-reinforcing loops that makes Claude better over time which has driven durable competitive advantages:
Rapid enterprise adoption has solidified Claude's position at the frontier of AI. In 3Q26, the company reported more than 11 900 customers with $100 000 or more in revenue on a trailing twelve months (TTM) basis and more than 1 900 customers with $1 million or more in revenue on a TTM basis as well as ~250 customers with $10 million or more of revenue on a TTM basis.
From research lab to the centre of the AI race
Frontier AI: AI systems at the current leading edge of capability and safety, performing at or near the highest levels of any existing system across a broad range of tasks.
Powerful AI (also Artificial General Intelligence [AGI]): A highly capable form of AI with cognitive capabilities matching or exceeding those of top human experts, able to navigate all interfaces available to a human working virtually, including text, audio, video, and direct computer control.
While we are still firmly in the frontier AI era, the rise of AI agents is evidence of the industry's push toward powerful AI, directed at some of humanity's hardest problems: compressing decades of scientific progress into years to accelerate drug discovery, disease detection and personalised therapy, as well as tailored education, expert financial and legal advice, and more efficient power grids and battery design. For this reason, demand from frontier labs such Anthropic and OpenAI has been a key driver of the extraordinary surge in hyperscaler capital expenditure (see graph below).
After a pause in FY23, the four largest hyperscalers (Amazon, Alphabet, Microsoft and Meta) stepped up spending sharply from FY24, with combined capex expected to exceed $600 billion in 2026, more than four times FY22 levels. Amazon leads with 2026 capex of ~$215 billion, more than three times its FY22 level. NVIDIA's own capex is comparatively small, as it is primarily a chip designer rather than an infrastructure operator, though it has still grown ~5.7 times over the period.
This buildout has in turn fuelled Anthropic's own growth, creating an iterative loop with Anthropic at the centre. What sets Anthropic apart is that most of the AI supply chain now has a financial stake in its success. Of the above mentioned, Amazon and Google are its largest shareholders (holding about 15% and 13% of the company after the offer respectively), its main compute providers and, alongside Microsoft, its largest distribution channels. Chip designers and manufacturers NVIDIA, AMD, Broadcom, Micron and TSMC are also participating in the private placement, to secure downstream demand, recycle capital into their own hardware orders, and hedge against value shifting from chips to AI models. These companies have invested tens of billions into Anthropic, which Anthropic recycles straight back into those same companies through its multi-year cloud and silicon commitments.
AI ecosystem dynamics
Anthropic has ultimately landed in the centre by becoming deeply integrated across AWS, Google Cloud, and Microsoft Azure simultaneously, making it the default hedge for enterprises avoiding single cloud lock-in. Anthropic therefore sits at the intersection of the cloud platforms, the chip ecosystem and enterprise customers, rather than depending on any single partner.
Anthropic has ultimately landed in the centre by becoming deeply integrated across AWS, Google Cloud, and Microsoft Azure simultaneously, making it the default hedge for enterprises avoiding single cloud lock-in. Anthropic therefore sits at the intersection of the cloud platforms, the chip ecosystem and enterprise customers, rather than depending on any single partner.
Competitive landscape
Claude faces asymmetric threats from both large incumbent technology companies and well-funded AI-native startups. Unlike pure-play rivals, incumbents like Alphabet, Microsoft, Meta, and Amazon hold up to $5.2 trillion in market caps, trade at disciplined forward multiples (vs. Anthropic's ~13 times annualised revenue multiple), and can subsidise AI models via core cloud, search, advertising, or software franchises. Every major cloud partner now competes with Anthropic at the model layer (Amazon Nova, Google Gemini, Microsoft/OpenAI) while simultaneously relying on Claude API usage to drive high-margin cloud infrastructure consumption.
A less visible but potentially greater threat comes from open-weight models such as DeepSeek, whose freely available model weights allow companies to run capable AI on their own infrastructure at a fraction of the cost. If open-weight models close the capability gap, frontier pricing could come under pressure, compressing the premium that underpins Anthropic's valuation.
Financial performance
In FY25 ended 31 December 2025, revenue increased to $4.59 billion (FY24: $386 million) while the adjusted operating income was a loss widened to $4.9 billion (FY24: loss of $2.79 billion). Much of this stemmed from then higher compute requirements, which is the most significant operating expense in the business. Compute efficiency is therefore a critical determinant of the groups operating leverage and revenue generation, longer-term.
More recently, 3Q26 revenue surged to $16.96 billion (3Q25: $1.38 billion), and while compute and infrastructure expenses climbed 382% to $9.86 billion, adjusted operating income grew to $2.5 billion from a loss of $1.3 billion in the prior period. The margin came in at 15% (3Q25: margin loss of 93%). Compute and infrastructure expenses dropped from 148.1% of revenue in 3Q25 to 58.2% of revenue in 3Q26, reflecting both model/inference efficiency and rapid revenue scaling absorbing fixed training and R&D overhead.
Run-rate revenue, which annualises current usage and subscriptions, reached ~$81 billion as at 31 October 2026, up $10 billion from a month earlier. Management cautions that this is not a forecast and that growth rates are expected to decline over time.
A potential $238.9 billion gap by 2030
Anthropic enters the public market with a massive cash pile, holding $131.8 billion as of September 2026 and rising to ~$212.7 billion after the IPO and private placement. While the company has raised $141.1 billion since inception, it has signed ~$901 billion in long-term compute and infrastructure deals. Only $320 billion of that total is locked into fixed, unconditional payments, while the remaining ~$580.9 billion is tied to variable pricing, actual usage, or future construction milestones.
In the short term, Anthropic's runway is well protected, and its post-IPO cash balance covers the $70.7 billion due over the next 12 months three times over, leaving a $142 billion cushion. However, cumulative compute bills reach $451.6 billion by 2030, creating a $238.9 billion funding gap before accounting for salaries, R&D, and overhead. Closing this 2030 shortfall (and up to $688.2 billion across the full commitment period if all milestones are delivered) means Anthropic must turn its rapid revenue growth into massive free cash flow, or else repeatedly borrow money and issue new shares, thus exposing investors to rising costs, financing risk, and future dilution.
Governance structure post IPO
Post-listing, Anthropic's co-founders collectively are expected to retain 50.1% of the total voting power (but no economic rights) through a special Class F share, ensuring majority control over business strategy.
As a result, Anthropic will be a "controlled company" under Nasdaq rules. Control rests with the founders and the Long-term Benefit Trust (LTBT), which is designed to keep the company focused on its mission even when this may not maximise short-term shareholder returns.
Investment case
Risks
Of the 94 pages of risks highlighted in the listing statement, the key risks to the investment case include:
Valuation and SPM view
At the $74 midpoint, Anthropic's implied market capitalisation of approximately $968 billion places it on roughly 27 times trailing 12-month revenue of $35.3 billion and around 13 times run-rate revenue of ~$81 billion.
The valuation prices in continued hypergrowth and successful execution on enormous compute commitments. Anthropic offers rare exposure to a frontier AI leader with rapidly scaling revenue and a clear enterprise focus, but sustaining this opportunity depends on maintaining its technical lead and translating growth into durable cash generation while managing substantial compute commitments and regulatory risk. The risk embedded in this business is therefore regarded as high.
A free float of ~6% is low for a mega-cap public listing, with 20% ($12.2 billion at the midpoint of the offer range) of the capital raise allocated to retail investors. With a small proportion of the company's total equity available to the public, institutional demand has reportedly outstripped available shares, and the offer has been oversubscribed. This tight supply is therefore likely to cause significant upward price volatility immediately following the listing with uncertainty over whether this will reflect the current fundamentals of this business.
NB: While FNB is not participating in the capital raise, shareholders can gain access in the secondary market through the Global Trader platform once shares list on the Nasdaq stock market.
A Local Trader account is required before opening a Global Trader account.
If you do not have a Global Trader or Local Trader account, you can open one on the FNB App: Navigate to "Invest" tab > Select "Add accounts" > Select "Invest in Shares & Gold" > Select "Local Trader" or "Global Trader" > Select "Apply now" > Complete the steps and accept the Ts and Cs.