The relationship between Amazon and Anthropic has evolved from a straightforward investment into a complex organizational arrangement that now fundamentally shapes Amazon's reported financial performance and competitive positioning. Amazon's carrying value of Anthropic preferred stock stands at $32.0 billion 45 — a figure that has become a material driver of reported earnings, a source of mark-to-market volatility, and a structural lens through which investors must evaluate the sustainability of the entire AI infrastructure buildout.
What emerges from an examination of the available claims is a portrait of an industry grappling with an extraordinary mismatch between capital deployed and revenue generated. The key players — Amazon, Google, Meta, Microsoft, and the AI laboratories they fund — are locked in a circular, mutually reinforcing ecosystem of spending that carries both breathtaking upside potential and systemic fragility.
This matters for Amazon because the company's financial results are now materially shaped by non-cash valuation adjustments tied to a single private AI company, while simultaneously, Amazon Web Services derives a meaningful portion of its AI infrastructure demand from that same entity. Understanding the Anthropic relationship is no longer a niche concern for analysts — it is central to evaluating Amazon's earnings quality, competitive positioning, and exposure to an industry-wide correction whose contours remain incompletely visible.
The Revenue Trajectory: Extraordinary Growth, Disputed Methodology
Anthropic's reported revenue growth has been nothing short of remarkable. A heavily corroborated set of claims establishes that the company's annualized recurring revenue stood at $1 billion at the start of 2025 5, reached $9 billion by December 2025 4,5,14,18, and hit $19 billion by the end of February 2026 5. The pace appears to have accelerated further, with Anthropic reportedly adding $11 billion in ARR during March 2026 and the first seven days of April 2026 5. This would imply a 9-to-10x expansion over the historical period cited 4,18,20,22.
Looking ahead, one projection forecasts Anthropic's revenue at approximately $10 billion for calendar year 2025 and $100 billion by 2027 22, a pace that the analysis suggests would be faster than ByteDance's historical trajectory 22. The market has responded accordingly: the 90th percentile forecast for Anthropic's IPO market capitalization was revised upward from $873 billion on March 31 to $1.04 trillion on April 8, following the ARR announcement 22.
However, these numbers must be interpreted with significant organizational caveats. OpenAI has publicly disputed Anthropic's revenue recognition accounting, alleging that Anthropic inflates revenue by grossing up revenue shares with Amazon and Google 6, and asserting that stated revenue is inflated by approximately $8 billion due to differing methodologies 6. The crux of the dispute is structural: OpenAI reports revenue after deducting a 20% cost or cut attributable to Microsoft, whereas Anthropic reports revenue without such an adjustment 24.
Some commentators suggest that when accounting for revenue shared with cloud providers on a consistent basis, Anthropic's ARR is roughly comparable to OpenAI's 5, while another cited figure places Anthropic's ARR at $21 billion versus OpenAI's $5 billion 24 — a discrepancy that is almost certainly an artifact of differing methodologies rather than genuine competitive dynamics. This lack of comparability 5 introduces a material risk of misvaluation surprises for investors who have not adjusted for these accounting differences.
The Circular Capital Flow: A Structural Vulnerability
Multiple commenters have identified circular capital flow between Anthropic, AWS, and potentially NVIDIA as a major systemic risk 42. The organizational logic is worth examining closely. Amazon and Google invest billions in Anthropic; Anthropic uses that capital to purchase cloud compute from AWS and Google Cloud; those cloud providers then book revenue and backlog from Anthropic; and the cloud providers' investments in Anthropic are marked up as the private company's valuation rises. This circularity creates a feedback loop where each round of investment and spending justifies the next, but also introduces what analysts have characterized as "systemic fragility" 9.
The numbers involved are staggering. Mizuho analyst Vijay Rakesh estimates Broadcom revenue tied to Anthropic at $21 billion in 2026 and $42 billion in 2027 3,5. Broadcom supplies custom AI chips — TPUs and Trainium — to the hyperscalers, meaning Anthropic's demand for compute indirectly drives Broadcom's AI revenue. Google's projected AI infrastructure capital expenditure for 2026 is $75 billion or more 11. Meta updated its 2026 capital expenditure guidance to between $125 billion and $145 billion, up from a prior range of $115 billion to $135 billion 12,15,17,19,29,33,34,37,38,41. Commenters estimate combined hyperscaler AI infrastructure spending at $700 billion 25, with industry-wide AI spending approaching that same figure for 2026 40. One projection sees total hyperscaler AI capital expenditure exceeding $1 trillion by 2027 44, and global AI spending could potentially reach $1 trillion by 2027 48 or $4 trillion by 2030 22.
The structural concern is straightforward: if Anthropic represents a disproportionate share of Trainium utilization — AWS's custom AI chip — then the demand story becomes partly circular 43, introducing concentration risk in AI infrastructure demand signals 43. An organizational arrangement that appears robust during expansion could unwind in a destabilizing manner if growth decelerates.
The Capex-Revenue Mismatch: Organizational Realities
One of the most striking tensions in this cluster of claims is the enormous gap between AI infrastructure spending and AI-associated revenue. A Semper Augustus investor letter reports that AI-related capital expenditure in 2025 totaled approximately $400 billion 28, while 2025 AI-associated revenues were only between $30 billion and $50 billion 28. This indicates a fundamental mismatch 28 — roughly ten times more capital deployed than revenue generated. One commentator noted that $15 billion in revenue would not cover depreciation on $325 billion of planned spend 14. The $200 billion capital outlay figure (likely referencing Meta's or combined hyperscaler spend) spooked investors and fanned worries about an industry bubble 50.
OpenAI itself pushed back against such criticism, stating that "that caution looks less like discipline and more like underestimating how fast demand would arrive" 10. Yet the organizational question remains: can this revenue ever justify the spending?
There are counterpoints worth considering. Physical AI — AI integrated with robotics and autonomous systems — offers a massive total addressable market. Morgan Stanley estimates it at $5 trillion 35, Citigroup at $7 trillion 35, and another estimate places the 2026–2040 TAM at $3.26 trillion to $7 trillion 35. Potential artificial general intelligence could generate annual profits of $2 trillion, implying a potential market capitalization of $40 trillion to $60 trillion 27. The bull case rests on these addressable markets materializing; the bear case rests on them being speculative and distant. The organizational strategist recognizes this as a classic high-conviction bet where the structural logic is clear but the timing and magnitude remain deeply uncertain.
The Amazon Financial Impact: Earnings Quality Under Scrutiny
Amazon's Q1 2026 financial results were significantly impacted by non-cash valuation adjustments on its Anthropic investment. The company recorded a $12.3 billion upward valuation adjustment on Anthropic nonvoting preferred stock plus $4.5 billion in reclassification gains 45. The carrying value of Anthropic preferred stock reached $32.0 billion 45. However, these gains are explicitly volatile and reversible 45.
The impact on Amazon's reported EPS was material. Approximately 60% — $3 per share — of the reported $5.11 EPS was attributable to unrealized investment gains related to SpaceX and Anthropic 13. This means Amazon's operations contributed only about $2.11 per share of earnings, with the remainder coming from mark-to-market adjustments on private company stakes. Private market valuations of SpaceX and Anthropic are introducing significant mark-to-market volatility into reported earnings 13, complicating earnings analysis for even the most disciplined investor.
Zoom Video Communications also benefited, reporting a $406 million boost to net income from appreciation on its Anthropic investment 21, illustrating the widespread impact of Anthropic's valuation on multiple public companies' financials. The organizational implication is clear: a correction in Anthropic's private valuation would have cascading effects across the public company earnings landscape.
Operational Constraints and Capacity Issues
Despite its revenue trajectory, Anthropic faces severe operational constraints that introduce supply-side risk into its growth narrative. The company is "completely overwhelmed and out of compute capacity" 10, compute-constrained and throttling its users 10,49. Anthropic is reducing compute usage for its Claude models by making Claude "think less by default," decreasing token allowances for each plan, and reducing usage during peak periods 26. Some development teams are switching from Anthropic to competitors due to budget constraints 22, creating potential customer concentration risk that could produce cascading effects 22. Anthropic's growth is constrained by available compute capacity, creating a supply-demand imbalance risk 5.
In a telling data point, the company more than doubled its public estimate of average daily Claude Code spending from $6 to $13 per developer 47, suggesting either rapidly increasing usage or a deliberate repricing strategy. Anthropic's Managed Agents cost $0.08 per session hour on top of standard Claude API token rates 46, and Anthropic requires a minimum $20 Claude balance for initial testing 46. Approximately 80% of Anthropic's token usage occurs via API rather than through its consumer application 10, underscoring the enterprise focus.
Some commentators attributed part of Anthropic's revenue growth to the company's shift requiring most customers to use its more expensive API access tier instead of subscription plans 5, suggesting that pricing changes — rather than pure volume growth — may be driving some of the revenue expansion. This distinction matters for assessing the sustainability of the revenue trajectory.
Regulatory and Legal Overhangs
The U.S. Department of Defense has labeled Anthropic a "supply-chain risk" 1,2,7,9,16,23,31,32, a designation that became public in February 2026 5. Anthropic is legally challenging this designation 9,30,36, but the dispute signals potential government scrutiny that could affect the company's operations or customer base. Reddit has also filed a lawsuit against Anthropic 9, and the U.S. government is engaged in ongoing litigation with the company 23. These legal and regulatory overhangs introduce additional uncertainty into the investment thesis at a time when the organizational structure is already under considerable strain.
Implications for Organizational Assessment
-
Earnings quality requires careful parsing. Amazon's Q1 2026 earnings included approximately $3 per share in non-cash, reversible gains from private company investments 13. Investors should model Amazon on an adjusted basis excluding these gains, using the $180 to $210 DCF range 8 as a reference point for operational valuation. The $12.3 billion upward adjustment and $4.5 billion reclassification 45 will not recur in the same magnitude indefinitely.
-
The circular capital flow is the most important systemic risk to monitor. If Anthropic's growth decelerates or its valuation corrects, the impact would be twofold for Amazon: reduced AWS revenue from Anthropic compute spend and reversal of prior valuation gains in reported earnings. The fact that Broadcom's $21 billion in 2026 AI revenue is tied to Anthropic 3,5 and that Anthropic's $100 billion deal is not yet in Amazon's backlog 39 means the magnitude of exposure is still not fully visible to outside observers.
-
The capex-revenue gap demands resolution. The industry is spending more than $400 billion annually on AI infrastructure against $30 to $50 billion in associated revenue 28. Either revenue must grow tenfold, or capex growth must slow materially. Amazon and its hyperscaler peers are making a high-conviction bet that the former outcome prevails. Investors should monitor the trajectory of Anthropic's ARR growth — which reached $19 billion by February 2026 5 — as the leading indicator of whether this bet is paying off.
-
Anthropic's operational constraints represent both risk and opportunity. The company's compute shortage 10,49, customer throttling 10, and customer churn due to budget constraints 22 suggest that its revenue trajectory may be constrained by supply rather than demand. If AWS can solve Anthropic's compute needs profitably, the upside could be substantial. However, the Pentagon's supply-chain risk designation 2,7,9,16,32 and litigation with Reddit 9 introduce regulatory and legal overhangs that could disrupt this narrative at precisely the moment when organizational coordination is most critical.
Sources
1. The Pentagon is threatening to use the Defense Production Act to force Anthropic into military align... - 2026-02-28
2. Here is your AI summary of the week: 1/5 The AI sector saw major geopolitical tension this week. An... - 2026-03-14
3. Anthropic signs biggest compute deal yet with Google and Broadcom as run rate hits $30bn | TNW - 2026-04-07
4. Anthropic ARR hits $30 billion - 2026-04-07
5. Broadcom agrees to expanded chip deals with Google, Anthropic - 2026-04-06
6. OpenAI touts Amazon alliance in memo, says Microsoft has ‘limited our ability’ to reach clients - 2026-04-13
7. Anthropic ups compute deal with Google and Broadcom amid skyrocketing demand - 2026-04-07
8. TSLA at $190 is not a prediction, its just math. bear with me - 2026-04-12
9. Google to invest $10B in Anthropic at $350B valuation with up to $30B more tied to AI growth targets - 2026-04-24
10. OpenAI Misses Key Revenue, User Targets in High-Stakes Sprint Toward IPO - 2026-04-28
11. Google puts AI agents at heart of its enterprise money-making push - 2026-04-22
12. Big Tech Earnings Test AI Spending - 2026-04-29
13. GOOGL Quarterly Revenue $109.9 billion (up 22% YoY) - 2026-04-29
14. TSMC Quarterly Revenue US $36 billion (up 41% YoY) - 2026-04-16
15. Ad engines power Big Tech: Alphabet ads hit $77 billion, Meta surges 33%, Amazon crosses $70 billion run rate - 2026-04-30
16. Google, Nvidia and other tech titans sign AI deal with the Pentagon ->Los Angeles Times | More on "T... - 2026-05-01
17. Microsoft, Meta, and Google just announced billions more in AI spending—and only one got punished - 2026-04-29
18. Google Backs Anthropic With $40B and 5 Gigawatts - 2026-04-24
19. Meta shares slide as plan to spend billions more on AI spooks investors - 2026-04-30
20. is anyone actually making money from AI or is it just the chip sellers? - 2026-04-24
21. Is Zoom Communications a buy after shifting to an AI-first strategy with almost $8 billion in cash? - 2026-04-18
22. I legitimately think Anthropic is worth at least $100B more than it was a week ago - 2026-04-09
23. GOOGL’s $40B Anthropic bet, A strategic move toward $400/share? - 2026-04-25
24. Okay! One more Microsoft post. - 2026-04-09
25. Intel is killing themselves and the market is celebrating - 2026-04-25
26. Amazon to invest up to another $25 billion in Anthropic as part of AI infrastructure deal - 2026-04-21
27. Does investing in upcoming LLM Stocks even make sense longterm? - 2026-04-11
28. Is AI’s real impact on stocks about margin expansion, not revenue growth? Looking for flaws in this thesis. - 2026-04-18
29. Alphabet checks boxes, Meta raises AI worries, says investor - 2026-04-30
30. NSA Tests Anthropic Mythos on Microsoft Software - 2026-05-01
31. Why Anthropic's new Mythos AI model has Washington and Wall Street worked up - 2026-04-14
32. The Priest Who Helped Write Claude's Conscience - 2026-04-09
33. 💬: How does TikTok’s European infrastructure investment compare to peers like Meta Platforms and Alp... - 2026-04-08
34. $META, $GOOGL - Meta is finally overtaking Google in digital advertising Meta $243.46B, Google $239... - 2026-04-13
35. $AMD Inference Queen to win in Physical AI 🤖 As we stand at the dawn of the agentic AI and physical... - 2026-04-19
36. Alphabet plans up to $40B investment in Anthropic: report | artificial intelligence | CryptoRank.io - 2026-04-24
37. Stocks climb to new record high as traders digest Big Tech earnings - 2026-04-30
38. Ad engines power Big Tech: Alphabet ads hit $77 billion, Meta surges 33%, Amazon crosses $70 billion run rate - 2026-04-30
39. We're raising our price target on Amazon after its all-around killer quarter - 2026-04-29
40. Amazon earnings beat expectations with strong cloud growth - 2026-04-29
41. Investors still trust Google more than Meta when it comes to spending their money on AI - 2026-04-29
42. Anthropic commits $100 billion to Amazon's AWS over next 10 years - 2026-04-23
43. Amazon’s $200B AI Bet Signals Shift in Data Center Buildout - 2026-04-16
44. AI boom: Big Tech capital expenditures now seen topping $1 trillion in 2027 - 2026-04-30
45. SEC 10-Q for AMZN (0001018724-26-000014) - 2026-04-29
46. Anthropic wants to be the AWS of agentic AI - 2026-04-29
47. E-commerce Industry News Recap 🔥 Week of May 4th, 2026 - 2026-05-04
48. What happens to the index if AI infra spending slows down? Which is inevitable - 2026-05-02
49. Amazon CEO Jassy defends $200 billion AI spend: "We're not going to be conservative" - 2026-04-09
50. Amazon CEO Jassy says company could sell AI chips, raising stakes for Nvidia, AMD - 2026-04-09
