Dear Readers,

Anthropic has officially gone public under the ticker "ANT", targeting an active price range of $70 to $78 per share. Beyond the headline excitement of a blockbuster listing, this debut is a historic test of how public markets value the future of artificial intelligence "ANT"

In this week's bonus Byte, I bring you the whole story: The Brain, The Business Model, and The Trillion-Dollar math behind Anthropic.


The IPO Offering

It is offering 825 million Class A shares in its public debut (801.35 million new shares and 23.65 million from existing stockholders), split between two main pools:

  • Up to 20% to Retail (165M Shares): Anthropic is setting aside up to 20% of the 825 million shares ~ 165 million shares (worth ~$12.2 billion at the $74 midpoint) for individual investors through participating brokerages like Schwab, Fidelity, Robinhood, SoFi, and E*Trade.
  • 80% to Institutions (660M Shares): The remaining 80% (~660 million shares) goes to mutual funds, hedge funds, sovereign wealth funds, and pension plans managed by the underwriting banks.
  • Concurrent $26B Strategic Placement: Separately from the public offering, seven major tech partners: Google ($10B), Nvidia ($10B), AMD ($2.5B), Broadcom ($1B), Microsoft ($1B), Micron ($1B), and TSMC ($500M) are buying $26.0 billion in non-voting Class N shares directly from Anthropic.

Before deciding whether to invest, it helps to understand what Anthropic actually builds, how it generates revenue, and the structural risks hiding inside its prospectus.


The Core Business: What is Anthropic?

Founded by seven ex-OpenAI researchers in 2021, Anthropic is a Frontier Model Developer. They construct "the brains" of artificial intelligence. Their core asset is the Claude ecosystem, a family of foundational AI models that can reason, write code, analyze data, and manipulate desktop interfaces.

They sell access to these models by the token. Approximately 80% of revenue is metered usage through its Application Programming Interface (API) consumption.

How API Consumption Actually Works for Anthropic

When a business "consumes" Anthropic's API, the process happens in milliseconds:

  • The Trigger: A customer on a banking app clicks "Analyze my spending habits."
  • The Request: The bank's software packages up the user's financial data and automatically shoots a secret text message through the API pipeline to Anthropic.
  • The Processing: Anthropic’s model reads the data, generates a smart response, and shoots it back through the pipeline.
  • The Delivery: The bank's app reads the text and displays it to the customer. The customer has no idea they just used Claude; they think the bank built the tool.

Why Usage-Based Billing Matter

In traditional software (like Microsoft Word or Netflix), you pay a flat monthly subscription. It doesn't matter if you watch 1 movie or 100; the price is the same.

Anthropic’s API business model is a usage-based utility, exactly like electricity or water.

  • Businesses are charged strictly by the volume of text (called "tokens") flowing through the pipeline.
  • They pay a specific rate for every 1 million tokens sent into Claude (prompts) and a higher rate for every 1 million tokens Claude sends out (answers).

When institutional investors look at Anthropic’s $81 billion revenue run-rate (Oct’26 Revenue $6.75B X 12), they aren't looking at consumer subscriptions. They are looking at massive enterprises consuming billions of tokens every single hour through the API pipeline.


Where Anthropic Sits in the AI Stack (And Why It Matters)

Think of the AI industry as a four-step assembly line from raw hardware to finished software:

  • Layer 1 - Silicon & Servers (Nvidia, AMD, Micron, Sandisk, Supermicro): Build the specialized chips, high-speed memory, and server racks that power AI.
  • Layer 2 - Cloud Hyperscalers (AWS, Google Cloud, Microsoft Azure): Buy those servers, plug them into massive data centers, and rent out the computing power.
  • Layer 3 - Frontier Model Developers - Anthropic’s Seat (Anthropic, OpenAI, Google, Meta, xAI): Lease raw computing power from the cloud giants to train foundational AI models like Claude.
  • Layer 4 - Applications (Claude Code, ChatGPT, Grok, Microsoft Copilot): Package that AI intelligence into software tools for businesses and consumers.

So What? The Circular Trap in the Middle

Sitting in Layer 3 leaves Anthropic squeezed from both sides. Anthropic does not own its own data centers, meaning it must rent computing power from AWS and Google Cloud who happen to be Anthropic’s largest suppliers, major investors, distribution partners, and direct competitors (building rival models like Google's Gemini) all at the same time.

Every dollar tech giants invest into Anthropic largely flows right back to them in cloud compute fees. To avoid being trapped as a middleman beholden to its landlords, Anthropic is pushing aggressively into Layer 4 with direct enterprise products like Claude Code so it owns the end-customer relationship directly.


The Financial Landscape: Hyper-Growth on a Treadmill

Anthropic is scaling at a speed never seen in corporate history. Revenue grew from $4.59 billion in 2025 to an expected $52.0 billion for full-year 2026, 79% of which was driven by its metered consumption and remainder by subscription.

 

44% of nine-month revenue ran through cloud marketplaces. The clouds/hyperscalers (layer 3 of AI eco system) are investors, suppliers and resellers at once.

Who are its customers:

Their customer base is enterprise - More than 11,900 customers spend over $100,000 a year as of Sept. 30, up from about 950 a year earlier. More than 1,900 of those customers actually spend over $1 million a year, and more than half are actually Fortune 500 companies or in other words, they have yet to capture remainder of the Fortune 500 companies.

Customer Concentration – In last 9 months of 2026, no customer was above 10% and government was below 2% of revenue.


Anthropic’s Operating Costs Outweigh Its Revenue in 9 months ended Sep 30th, 2026

Compute Consumes More Than Half of Anthropic’s Revenue. 

In nine months ended Sep 30th 2026, Anthropic spent $18.8 billion on compute and infrastructure, a 329% increase compared to nine months ended Sep 30th, 2025. The increase was driven by the delivery of committed compute throughout the first nine months of 2026.

As revenue has scaled over time, compute costs have fallen from $10.56 in FY2023 to just $0.57 per revenue dollar, reflecting higher utilization of existing capacity and exponential revenue growth.

While compute costs may appear well-controlled today, they are expected to grow materially. 

The compute primarily comprises cost of training and inference and both are expected to increase as Anthropic continues to train larger, more advanced models. Additionally, Anthropic has also contracted significant compute capacity to meet rapidly increasing demand. Therefore, it is good to presume compute per revenue dollar to also increase as its massive compute commitments come online, leading to some margin compression.

The reason for signing large commitments is pretty obvious. Demand for frontier AI models far exceeds the existing supply of AI compute, with AI labs aiming to bring more capacity online to capture incremental revenue and grow their businesses. 

However necessary for sustaining growth amid growing demand, these massive future commitments pose a significant risk for Anthropic. CEO Dario Amodei emphasized “I could buy $1 trillion of compute that starts at the end of 2027. If my revenue is not $1 trillion, if it’s even $800 billion, there’s no force on Earth, there’s no hedge on Earth, that could stop me from going bankrupt.”


The Cost of Elite Talent

In the nine months ended Sept. 30, 2026, Anthropic spent $4.86 billion on personnel, or 15 cents per revenue dollar. Roughly 68% ($3.3 billion) was stock-based compensation, leaving cash pay at roughly $1.56 billion (5 cents per revenue dollar) across 5,300 staff.

That is a remarkably lean cash payroll for its revenue scale. Trailing 12-month revenue per employee exceeded $6.6 million as of Sept. 30, 2026, up from $1.4 million a year earlier and well above top software peers.

This tells me 3 things:

  • Preserving vital cash - Paying 68% of compensation in stock keeps actual cash payroll at just six cents per revenue dollar.
  • The dilution cost for investor - Handing out roughly $3.3 billion in stock compensation in just six months directly dilutes public shareholders. The company is essentially paying its employees with your future ownership stake, which puts downward pressure on the stock price once lock-ups expire.
  • Scaling without headcount - Generating $6.6 million in revenue per employee is an elite efficiency metric that beats almost every major software company in history. It proves Anthropic does not need to hire a massive army of salespeople or support staff to grow.

When Does Anthropic Expect to be Profitable?

Anthropic is not profitable yet, and heavy data center bills will keep it burning cash until 2028. For the nine months ended September 30, 2026, the company reported a headline net loss of $104.0 billion on $33.24 billion in revenue, though $94.3 billion of that was a non-cash accounting charge leaving an adjusted loss of $9.7 billion.

  • Core Operations Are Nearing Breakeven. Stripping out paper adjustments narrows the nine-month operating loss to $1.26 billion, highlighted by Q2's first-ever operating profit of $974 million (an 8.4% margin) before Q3 slipped to a $567 million operating loss on extra compute purchases.
  • Cash Burn Continues Through 2027. Per WSJ reporting on internal investor road maps, cash burn is projected to fall to roughly 9% of revenue in 2027, i.e. Anthropic will spend $1.09 for every $1.00 it brings in (a net cash outflow of 9 cents per sales dollar) as its $451.6 billion in signed compute leases begin coming online.
  • Full-Year Profitability Target. Management expects the company to reach full-year breakeven for the first time in 2028.

The Share Count Math

Standard financial media evaluates Anthropic using its 13.09 billion basic common shares, yielding a ~$1.02 trillion market cap. But due to aggressive Stock-Based Compensation (SBC) used to recruit elite AI scientists, the fully diluted share count sits at roughly 18.00 billion shares, generating a massive $1.404 Trillion Fully Diluted Valuation (FDV).

  • Basic Price-to-Sales: 12.6x based on the October revenue run-rate (At $78: 1,021.0B\$81.0B).
  • Fully Diluted Price-to-Sales: 17.3x, which represents the true economic cost you are paying for a claim on future earnings (At $78: 1,404.0B\$81.0B).

The 13.09 billion basic shares already include converted prior private rounds, the new IPO shares, and the $26 billion strategic partner placement. The bridge to 18.00 billion fully diluted shares adds roughly 4.9 billion potential shares from employee stock options, unvested RSUs, reserved hiring equity pools, and Amazon convertible notes.


The Investment Thesis: Bull vs. Bear

Bull Case for Anthropic – rapid adoption of Claude code and increased market share

In the bull case, I estimate Anthropic’s five-year revenue growth rate to be north of 60%, with a 2027 price/sales multiple of 26 times and its equity value to be ~$2.16T.

The bull case is based on the fact that Anthropic already reached an $81.0 billion annualized run-rate in October 2026. That makes hitting the $83.20 billion Year-1 target in 2027 highly probable, leaving four years of steep compounding to prove out.

Lead underwriter Goldman Sachs actually models an even steeper 90.5% CAGR through 2030, projecting $685.0 billion in annual sales. The 60% growth assumption is more conservative than the bank's roadshow pitch.

Bear Case for Anthropic – slower enterprise adoption and competition from open- weight model providers like Meta and Google

Using a $52.0B 2026 baseline and 18.00 billion diluted, a true bear case keeps IPO buyers underwater:

The bear case assumes revenue scaling more gradually due to slow enterprise adoption, as existing switching costs and the risks associated with replacing traditional workflows with AI outweigh the potential AI-related benefits in a typical enterprise’s strategic calculus.

In addition to the slow enterprise adoption, model capabilities also scale more gradually, opening fewer target markets for Anthropic and slowing the pace of adoption of Anthropic’s models.


The Bottom Line: At $70–$78 per share, it appears Anthropic is reasonably close to its fair value. If you secure an initial IPO allocation, you can ride the near-term momentum. Just be sure to protect your gains ahead of the 6-month insider lock-up expiration, when employees and early investors can begin selling their shares on the open market.

If you plan to hold for the long run, keep an eye on three key talking points:

  • Revenue Growth & Adoption: Watch whether more enterprises continue adopting Claude and finding new ways to apply the models, or if corporate spending starts to slow down.
  • Market Share & Pricing Power: Because cloud hyperscalers act as Anthropic's suppliers, investors, and rivals all at once, track who is winning market share: Anthropic, OpenAI, or the hyperscalers and who can still charge premium prices.
  • Compute Costs vs. Cash Runway: Make sure fixed compute bills do not outpace usage revenue and squeeze margins. While Anthropic holds $131.7 billion in cash today and the IPO plus a $26 billion partner placement will push reserves past $200 billion to cover the next 1–2 years of commitments, its $895.4 billion in multi-year compute costs are legally fixed while revenue is not, meaning the company still faces real bankruptcy risk if projected growth fails to materialize.

I hope you found this Byte helpful as you evaluate this much-anticipated IPO.

Until then…

Invest Smart with Pooja :)

Subscribe to InvestSmartWithPooja and never miss a Byte. Your weekly dose of smart investing and financial education.

This isn't a recommendation to buy or sell. It's about understanding the principles and process behind long-term investing.