Emerging Dependence on Major Tech Giants
Recent filings indicate that Anthropic is increasingly aligning its sales of artificial intelligence technologies with tech titans such as Google, Amazon, and Microsoft.
As outlined in the Initial Public Offering (IPO) registration document released by Anthropic on Monday, nearly half of the company’s total revenue—specifically, 48%—generated through the cloud marketplaces of these three “hyperscalers” during the first half of 2026, a stark increase from just 11% in 2023.
These prominent cloud providers are tasked with processing customer payments before remitting the remaining funds to Anthropic, following the deduction of their respective fees.
Notably, the fees paid by Anthropic to its distribution partners—which encompass the aforementioned cloud giants—escalated by a staggering $2.7 billion in the first half of this year compared to the same timeframe last year.
This model of collaboration has propelled Anthropic’s swift ascension. The company reported sales of $115.55 billion in the second quarter, alongside an operating profit of $974 million.
Anthropic’s impending IPO is poised to serve as a critical barometer for investor sentiment within the artificial intelligence sector, which, as of late, seems to have magnified concerns regarding the exorbitant costs associated with large-scale AI infrastructure development.
Reports earlier this month hinted at a deceleration in Anthropic’s revenue growth during the third quarter, intensifying a broader tech stock sell-off amid apprehensions that the growth of AI investments may surpass actual market demand.
Moreover, Anthropic explicitly highlighted these risks in its S-1 filing, depicting its cloud service partners as simultaneously fulfilling four roles: “investors, customers, distribution partners, and competitors.”
The company cautioned that these cloud providers possess the capability to scrutinize its pricing structures and business terms.
The document accentuated the possibility that these major providers might prioritize the promotion of their proprietary AI models; concurrently, Anthropic’s close association with these entities has attracted the scrutiny of antitrust regulators, hinting at potential regulatory challenges in the future.
The “Circular Deal” Phenomenon
Previously, Anthropic faced criticism for its so-called “circular deals” within the tech landscape: entities that financially support Anthropic’s balance sheet simultaneously reap revenue from the company.
As a significant client of Amazon Web Services (AWS), Anthropic employs Amazon’s data centers and specialized chips to train and operate its Claude model, with Amazon in turn delivering Claude to clientele via its Bedrock platform.
Similarly, Anthropic leverages Google’s bespoke AI chips, while Google marketizes Claude through its own cloud infrastructure. Thus, these firms serve as Anthropic’s investors, suppliers, sales channels, and competitors all at once.
Anthropic’s dependency extends to computational resources as well, committing to pay a minimum of $114.5 billion to Google by 2033, $110 billion to Amazon by 2036, and $44.9 billion to Microsoft.
Intriguingly, these minimum commitments persist irrespective of whether Anthropic eventually utilizes these computing capacities.
A Complex “Frenemy” Dynamic with Cloud Titans
These statistics illuminate the intricate “frenemy” dynamics between Anthropic and the prominent cloud services.
In April of this year, upon solidifying its agreement with Amazon, Anthropic designated AWS as its “primary training and cloud service provider for mission-critical workloads.”
Yet, examining the scale of contracts reveals that Anthropic’s minimum purchase obligation to Google is more substantial and involves a shorter duration.
Additionally, Anthropic anticipates incurring further costs ranging from $125 billion to $150 billion for supplemental TPU (Tensor Processing Unit) computational capacity.
This supplementary power is expected to be sourced from Google and other suppliers, slated for delivery between 2028 and 2031.
Significant Shareholding by Amazon and Google
Both Amazon and Google hold considerable shares in Anthropic; however, details regarding their exact shareholdings remain undisclosed in the IPO documentation.
A key reason for this is that their shares do not carry voting rights, thus excluding them from the disclosure of stakeholders holding more than 5% of Anthropic’s voting securities.
Nonetheless, Anthropic clarified in other segments of the document that after the conversion of preferred shares and certain convertible notes during the IPO process, both Amazon and Google will possess shares exceeding 5% of the company.
Since committing to invest up to $40 billion in 2023, Amazon has cultivated a profound investment and business rapport with Anthropic.
Data from the IPO document reveals that Amazon has already invested approximately $18 billion in Anthropic and has pledged to infuse an additional amount of up to $150 billion, contingent upon fulfilling specified conditions.
Moreover, Amazon retains convertible notes from purchase agreements made in 2023 and 2024, allowing for conversion into a maximum of 1.84 billion Anthropic shares, accounting for roughly 10.9% of Anthropic’s fully diluted shares as of June this year.
This figure excludes other equity interests currently held by Amazon. However, a restriction applies to these convertible notes: if conversion would elevate Amazon’s total shareholding beyond 12%, the conversion cannot proceed unless voluntarily waived by Amazon.
The portion exceeding the 12% threshold will automatically convert into new notes, which may be converted in the following quarter but will still be subject to the same 12% shareholding cap.
In contrast, Google has earmarked about $230 billion for investment in Anthropic across various financing rounds and has committed to an additional $200 billion contingent on achieving specific performance benchmarks.
Under the latest financing arrangement, Google’s upper shareholding ceiling is set at 14.99%, although Google possesses the discretion to waive this limit.
Anthropic’s Vital Connections with Cloud Giants
The IPO documentation elaborates on how Anthropic has intricately entwined its future prospects with these cloud computing behemoths.
Anthropic asserted that migrating workloads to rival cloud service providers would be “costly and time-intensive,” potentially jeopardizing the efficiency of its computing operations.
This assertion underscores that Anthropic’s reliance on Google, Amazon, and Microsoft transcends mere supplier partnerships, embedding itself firmly within the company’s foundational architecture and operational framework.
However, Anthropic’s dependencies are not limited to these three dominant cloud players; the company engages other suppliers as well.
For instance, Anthropic has entered into a $161.2 billion equipment lease agreement with Broadcom and has committed up to $120.5 billion for computing resources from SpaceXAI.
In essence, while Anthropic is constructing an expansive AI infrastructure system, this endeavor concurrently amplifies its reliance on a select cadre of considerable tech giants and chip suppliers.
Per insights gleaned from the IPO document, this dependency not only serves as a catalyst for Anthropic’s rapid expansion but also emerges as a potential source of future costs, supply chain vulnerabilities, competitive challenges, and antitrust inquiries.
Source link: Eu.36kr.com.
