TechFlow Logo
Login/ Sign up
ETH Gas
Gwei
Fear
gas
Anthropic Is the AWS of AI: Why It Turns Cutting-Edge Intelligence Into Enterprise AI's Default Infrastructure

Anthropic Is the AWS of AI: Why It Turns Cutting-Edge Intelligence Into Enterprise AI's Default Infrastructure

2026.09.09
Share
Anthropic will create enterprise-level lock-in, with long-term gross margins expected to exceed 60% or even higher, and build an extremely lucrative business like AWS, while its ARR will surpass AWS by 2027.

Author: Artemis Analytics

Compiled by: TechFlow

TechFlow Summary: As the market debates whether open-source models will turn cutting-edge AI into a commodity, Artemis offers a counterintuitive thesis: Anthropic's true moat isn't the models themselves, but its enterprise-grade lock-in capabilities akin to AWS. This analysis breaks down how to read Anthropic's S-1 through three dimensions: compute capacity, gross margins, and channel structure, offering valuable reference for investors tracking how AI narratives map onto crypto and tech stock valuations.

Core Thesis: Anthropic is the AWS of AI

In 2015, skeptics viewed AWS as an extremely unprofitable cost center internal to Amazon, offering a commoditized product akin to electricity, with no one able to clarify just how large the cloud market truly was.

To make matters worse, the market worried that competitors like Google and Microsoft would crush AWS's pricing power.

Fast forward to 2026, and Amazon is the undisputed winner in cloud computing, having built a powerhouse business far exceeding its initial S3 offering through a one-stop product matrix, scale economies, and a developer ecosystem.

By 2026, Anthropic faces similar skepticism: despite raising and burning billions of dollars, their cutting-edge models could still be commoditized by open-source players.

We believe that with an exceptional team, Anthropic will not only continue pushing the boundaries of cutting-edge models but also build a highly sticky, one-stop AI product matrix that enables enterprises to easily build, scale, and maintain AI systems across every vertical industry globally.

These products include, but are not limited to, domain-specific cutting-edge models (e.g., drug discovery models) and tools ensuring AI safety, compliance, and performance optimization.

Anthropic will create enterprise-grade lock-in, with long-term gross margins expected to reach over 60% and potentially higher, building an extremely profitable business akin to AWS, with ARR surpassing AWS by 2027.

Full financial model available at: https://www.artemis.ai/anthropic-thesis

The Essence of Anthropic (What to Look For in the S-1)

Anthropic's core comes down to three things: 1) securing top-tier researchers to develop cutting-edge models; 2) acquiring and locking in compute at low cost ($10M to $15M per megawatt); 3) charging a premium for tokens via API (e.g., $50M ARR per megawatt).

Key metrics breakdown:

  • Demand side: We estimate monthly net new ARR between $10B and $15B. Even as token prices decline, token usage continues to rise, and enterprise demand for cutting-edge models and deployed agents has become significantly de-risked.
  • Supply side: Locked total compute (GW). Reports indicate Anthropic has locked up 15–16 GW of compute by 2030. Supporting $1 trillion in ARR requires at least 20 GW of training compute, which we believe is achievable by 2030 (OpenAI is reportedly targeting at least 30 GW of total locked compute by 2030).
  • Gross Margins: The spread between the ARR Anthropic can charge per megawatt (in millions) and the compute cost per megawatt (in millions) is key to determining Anthropic's long-term profitability and whether it constitutes a high-quality business.

Additionally:

  • Anthropic is fundamentally an enterprise API business. 90% of ARR comes from APIs. The $10T terminal value hinges on Anthropic's ability to sell models directly, as well as through AWS Bedrock, Gemini Enterprise Agent Platform, and Microsoft Foundry. Consumer business is negligible.
  • Focus on Net ARR, not Gross ARR. Investors must deduct the 15–20% revshare paid to AWS, Gemini, and Microsoft, and exclude revenue from Meta and Chinese AI labs distilling Anthropic's models.

What Could Go Right

Several variables could align favorably.

In our base case, Anthropic locks up over 30.8 GW of total compute by 2030, supporting $1 trillion in ARR, with total compute costs at $18M per megawatt and inference ARR at $50M per megawatt.

This implies a stable-state gross margin of 66%, EBIT margin of 30%, and total training/R&D expenditure falling to approximately 25% of revenue.

At this stage, Anthropic generates massive profits and grows rapidly as agents continue to penetrate over 300 million enterprises globally.

What Could Go Wrong

Ranked by probability:

  • High Risk: Codex and OpenAI Astra steal share from Anthropic. We've observed engineers around us migrating from Claude Code to Codex. After trialing Astra and Fable last weekend, we found the gap has narrowed significantly, and customers may pivot to OpenAI. Here are direct customer quotes: Head of Blockchain Data at Artemis: “From personal experience, I’ve been using Codex exclusively. I prefer its interaction style. Some treat them as complementary—Claude Code for planning, Codex for execution—but I view them as substitutes and stick solely to Codex. I used to be a heavy terminal user of Claude Code, but now I prefer the Codex experience within the ChatGPT app.”
  • Fintech Analyst at Artemis: “OpenAI’s Codex app is incredibly strong. It visualizes what agents are doing and is clearly built for coding. Claude Code’s first-gen CLI experience was solid, but Codex’s GUI massively outperforms it when observing agent behavior. Claude Code pioneered the first-gen coding experience in CLI, but the second gen will undoubtedly emerge in GUI-based applications.”
  • Product Manager at a $3B company: A close contact and die-hard supporter of Claude Code and Claude Harness shared over the weekend: “Wow, guys, I have to say, Astra is insanely good.”
  • High Risk: Low switching costs between models. In Q2 2026, 25% of gross revenue flowed through third-party channels like AWS Bedrock and Gemini Enterprise Agent Platform. Via AWS Bedrock, Anthropic’s APIs can be easily swapped by enterprises for OpenAI’s Astra and other open-weight models. If Anthropic loses its cutting-edge edge, enterprises can switch effortlessly, potentially shifting revenue significantly.
  • Medium Risk: Harnesses like Grokbot and Instinct don’t require Anthropic. Grokbot runs on x.ai’s cutting-edge models, directly challenging Claude Cowork and Claude’s harness offerings. If Instinct and new AI apps are built on open-source or alternative cutting-edge models, it forces Anthropic to move up the stack—acquiring harnesses/apps or being compelled to innovate similarly to Claude Code, Claude Design, or Claude Cowork.
  • Low Risk: Compute cannot be locked down. Anthropic has already secured nearly 15 GW and needs another 15 GW to support $1T ARR. An IPO above $100B is more than enough capital to secure compute supply. That said, Amazon, NVIDIA, Google, and SpaceX are simultaneously suppliers, shareholders, and competitors (Amazon has Titan, SpaceX has Grokbot, and NVIDIA is moving upstack via its HuggingFace partnership). We consider this scenario highly unlikely.
  • Low Risk: Anthropic halts cutting-edge model releases. Either researchers become wealthy enough post-IPO to step away, or they hit a technical bottleneck. Currently, incentives are perfectly aligned, but what if researchers hold $150M–$300M in equity? A $5M–$10M stake at a $100B–$200B valuation becomes nine figures at a $3T valuation. We deem it highly unlikely that Anthropic will stop releasing frontier models.

Valuation and Closing Thoughts

Even against high-growth software and AI peer companies, Anthropic trades cheaply at a $2T valuation based on EV/ARR and EV/NTM revenue multiples (not unreasonable even at $3T).

We expect Anthropic to list in mid-to-late October at a $2T valuation, with retail and institutional demand pushing it toward $3T–$4T.

Honestly, Anthropic might be a buy regardless.

  • Anthropic reported $65B net ARR in July 2026, implying 15x YoY growth. Our estimate of $90B ARR by end-September 2026 also reflects 15x YoY growth—a generational-scale expansion.
  • Furthermore, our forecast (more conservative than SemiAnalysis’s $300B ARR projection for 2027) sits at $125B ARR by end-2026 and $275B ARR by end-2027, representing 20x and 7.2x multiples against expected 2026 and 2027 ARR, respectively.
  • For one of the fastest-growing public companies ever, plus the clear winner in enterprise AI, it’s hard to argue against going long $ANTHR.

How should you value a company that could develop AGI and sell it to over 300 million enterprises globally, all eager to replace or augment white-collar workers with intelligence? Rough math: 300M enterprises × 1 replaced worker each × $120 ACV = $36T in global potential AI spend. What if Anthropic captures 20% ($7.2T ARR)? 30% ($10.8T ARR)?

Ramp estimates 56% of US enterprises currently spend on AI, yet the median is merely $12/month. We attribute this to enterprises paying for ChatGPT or Claude individual subscriptions.

Our long-term thesis is that agents—not chatting with agents via Claude or ChatGPT—are the key to accelerating global, broader AI adoption. Currently, agent penetration in enterprises remains under 1%.

In short, our base case of $1T ARR by 2030 might actually be too conservative.

What if superintelligence arrives and Anthropic builds models with its own models, drastically lowering research costs and expanding margins? What if superintelligence is only accessible through Anthropic, payment willingness spikes, and the $50M ARR/megawatt assumption proves too conservative after all?

Time will tell. These frontier models are inching closer to superintelligence every day.

Finally, I’m reminded of these words from Anthropic’s CFO: “Humans mostly think linearly and incrementally. I’ve been at [Anthropic] for two years. It was a paradigm I had to break for myself. Stop thinking linearly. Start thinking exponentially.”

Congratulations to Anthropic. Looking forward to the S-1 filing and the exponential growth along the path to superintelligence.

Disclosure:

This article represents the personal views of Jon Ma and is published for informational and educational purposes only. It does not constitute investment, financial, legal, or tax advice, nor does it offer a recommendation or solicitation to buy, sell, or hold any securities (including any interest in Anthropic).

Artemis Analytics is not a registered investment advisor or broker-dealer. The author may hold a position in Anthropic via an SPV. Always conduct your own research and consult licensed professionals before making any investment decisions.

Anthropic is a private company. Its shares are unregistered and not publicly traded, secondary market interests lack liquidity and can only be sold through restricted channels, and third-party “marks” or secondary market prices cited herein (including any implied valuations) may not reflect actual transaction prices. Any references to a future IPO, including timing or pricing, are speculative; there is no guarantee an IPO will occur or under what terms.

Conflict of Interest. The author and Artemis Analytics hold economic interests in Anthropic directly and/or through pooled investment vehicles, and would benefit from a rise in Anthropic’s valuation. We may trade these interests at any time without updating this page. Artemis Analytics is also a customer of Anthropic; our products are partially built on Anthropic’s models, and Artemis personnel are quoted in this article as users of Anthropic’s products. Readers should assume the author is not a neutral observer.

Forward-Looking Statements & Estimates. This page contains estimates, forecasts, scenarios, and price targets (including ARR, gross margins, compute capacity, and 2030 valuation data), reflecting the author’s own views or sourced from third parties, including Artemis’ internal estimates and published research from firms like SemiAnalysis. These figures are not financial results reported by Anthropic, have not been verified or endorsed by Anthropic, and are inherently uncertain. Actual results may differ materially. Bull, base, and bear cases are illustrative scenarios, not predictions or guarantees.

Information from third parties is believed to be reliable but has not been independently verified; we make no representations regarding its accuracy or completeness. All investments involve risk, including the potential loss of entire principal amounts; past performance does not guarantee future results. Conduct your own due diligence and consult licensed financial advisors, attorneys, or tax professionals before investing. Anthropic and Artemis Analytics are unaffiliated, and neither has reviewed or approved this content.