Anthropic (ANT.US) pushes for US IPO! Goldman Sachs and Morgan Stanley lead 22 investment banks in underwriting. Google (GOOGL.US) and Amazon (AMZN.US) contribute nearly half of revenue. Risks in the computing power supply chain attract attention.

According to English Finance APP, artificial intelligence startup Anthropic (ANT.US) is accelerating its initial public offering (IPO), with Goldman Sachs (GS.US) and Morgan Stanley (MS.US) securing the lead underwriting positions. Twenty other investment banks are also participating in the deal. Additionally, Anthropic’s IPO registration documents revealed the complex business relationships underpinning its rapid growth: Google (GOOG.US, GOOGL.US), Amazon (AMZN.US), and Microsoft (MSFT.US) are not only major investors, computing power suppliers, and sales channels for the company, but also potential competitors. Anthropic further warned that the global concentration of advanced chip manufacturing and its reliance on key suppliers such as TSMC (TSM.US) may pose risks to its long-term business development.

According to Anthropic’s S-1 registration filing, Goldman Sachs was designated as the IPO’s primary lead underwriter, topping the list, with Morgan Stanley ranked second and appointed as the stabilization agent responsible for stabilizing the stock price during the early stages of trading if necessary.

This arrangement mirrors the underwriting structure for SpaceX’s (SPCX.US) record-breaking IPO in June this year. At that time, Goldman Sachs and Morgan Stanley also served as the top two core underwriters. SpaceX raised $86.2 billion through its IPO, and the 22 participating banks received a total of $500 million in fixed underwriting fees, about 0.58% of the total amount raised. Among them, Goldman Sachs and Morgan Stanley each received $100 million, the highest among all banks involved.

In addition to these two investment banks, Anthropic’s IPO syndicate also includes JPMorgan (JPM.US), Citigroup (C.US), Barclays (BCS.US), Wells Fargo (WFC.US), and Bank of America (BAC.US) among 20 others, bringing the total number of underwriters to 22, the same scale as SpaceX’s previous IPO.

IPO Timing and Financial Quality

Anthropic’s IPO comes at a pivotal moment as the artificial intelligence sector faces renewed scrutiny from investors. Demand for AI technology continues to grow, but massive infrastructure investments, rising compute costs, and the question of sustainable profitability are becoming central concerns for public market investors.

Anthropic’s disclosed financials show second-quarter revenue of $11.6 billion and operating profit of $974 million, indicating that its AI business is rapidly scaling.

However, earlier this month, reports of slowing third-quarter revenue growth for Anthropic increased selling pressure on tech stocks, reflecting market concerns that capital expenditures in AI may be outpacing actual demand.

Notably, Anthropic’s rapid expansion heavily relies on sales channels provided by major cloud service providers. According to IPO documents, in the first half of 2026, sales via hyperscale cloud vendors such as Google, Amazon, and Microsoft accounted for 48% of Anthropic’s total revenue, a significant jump from 11% in 2023.

Under this model, cloud providers charge customers directly, then pay Anthropic the remaining amount after deducting channel service fees. As this channel grows in importance, expenses paid to channel partners have also surged. In the first half of 2026, these expenses increased by $2.7 billion compared to the same period last year. This means that although Anthropic has gained a large enterprise customer base through cloud giants’ distribution networks, its dependency on these partners continues to deepen.

Cloud Channel Dependency and “Quadruple Role” Risks

Anthropic’s prospectus explicitly states that Google, Amazon, and Microsoft simultaneously serve as “investors, customers, distribution partners, and competitors.”

This complex relationship brings potential risks. Since cloud providers handle end-customer billing, they control Anthropic’s product pricing and contract terms. At the same time, these companies have their own AI models and products and may prioritize promoting their proprietary technology over Anthropic’s Claude model in the future.

Anthropic also warns that its close ties with major cloud companies have already attracted antitrust regulatory attention, and it may face further scrutiny going forward.

Beyond sales channels, Anthropic also relies heavily on these cloud giants for computing resources used in AI model training and inference. According to its prospectus, Anthropic has committed to spend at least $114.5 billion with Google by 2033, $110 billion with Amazon by 2036, and $44.9 billion with Microsoft. Even if it does not fully utilize the contracted compute, the company must fulfill these minimum spending obligations. The total of these three commitments is $269.4 billion, highlighting Anthropic’s massive financial obligations to guarantee long-term compute supply.

Additionally, Anthropic expects to pay Google and other suppliers an extra $125–150 billion between 2028 and 2031 for additional TPU compute resources. This arrangement further illustrates the cooperative yet competitive relationship between Anthropic and its cloud partners.

In April, Anthropic announced a cooperation agreement with Amazon and referred to AWS as its key workload training and cloud service provider. Yet, according to more recent disclosures, Anthropic’s minimum spending commitment to Google is not only larger but also due sooner.

Amazon and Google are also major shareholders of Anthropic. According to the prospectus, since Amazon agreed to invest up to $4 billion in Anthropic in 2023, it has cumulatively invested about $18 billion and has promised to invest up to $15 billion more upon meeting certain conditions.

Media calculations show that, on a fully diluted post-conversion basis, Amazon currently holds around 9% of Anthropic’s economic interest.

Additionally, Amazon has about $6.2 billion in outstanding convertible notes, which under financing agreements signed in 2023 and 2024 can be converted into 1.84 billion Anthropic shares, representing an additional approximately 11% equity. However, Amazon’s stake is generally subject to a 12% cap, but it has the right to waive this restriction.

Google, through multiple transactions, has invested about $23 billion in Anthropic, and is committed to an additional $20 billion based on performance goals. Estimates suggest Google currently holds at least 6.6% of Anthropic’s economic interest.

Furthermore, Google has another $20 billion in unutilized financing as per a 2026 agreement. If Anthropic uses these funds, it may issue more shares or convertible notes.

Under the latest financing agreement, Google’s ownership limit is 14.99%, but it too may choose to waive this cap. Notably, Google also holds an option to subscribe up to $10 billion in stock via a private placement simultaneous to the Anthropic IPO at the public offer price. If exercised, this investment amount will be deducted from Google’s current financing agreement balance.

Massive Compute Commitments and Shareholder Ties

Besides the three major cloud giants, Anthropic has established large-scale compute supply relationships with other tech companies. According to the prospectus, Anthropic’s equipment rental commitment with Broadcom (AVGO.US) totals $161.2 billion, and it agreed to purchase up to $120.5 billion in compute resources from SpaceXAI.

However, Anthropic’s dependence on the compute supply chain extends beyond cloud providers and equipment vendors. The company also highlights in its IPO documents that advanced semiconductor manufacturing is highly concentrated, which may become a major risk factor for its business development.

Anthropic points out that the vast majority of the world’s most advanced chips are currently produced by TSMC, with its primary advanced manufacturing operations and supply chain centered in Taiwan. This concentration means any significant event affecting advanced chip production or supply in the region could have lasting effects on the global AI industry chain and further impact Anthropic’s business operations, financial health, and performance.

This risk underscores TSMC’s critical role in the global AI industry. AI chip designers such as Nvidia (NVDA.US) and AMD (AMD.US) heavily rely on TSMC’s advanced manufacturing process to turn chip designs into final products. Even tech giants like Amazon and Google that develop their own AI chips still depend on professional foundries to manufacture them.

Anthropic states in its prospectus that any constraints or delays in the supply of advanced chips may adversely affect its business, financial status, and operational performance. In fact, dependence on TSMC’s advanced manufacturing capability has become a shared trait among global AI companies. Other firms, including Nvidia, Broadcom, and Marvell Technology (MRVL.US), have previously emphasized the potential risks of supply chain concentration in their disclosures.

Although TSMC is actively expanding capacity worldwide, in the short term, the highly concentrated nature of advanced chip manufacturing is unlikely to fundamentally change. According to reports, TSMC has committed $265 billion to build advanced semiconductor facilities in Arizona and is considering establishing new large production bases in Texas.

However, Anthropic states in its prospectus that the construction of advanced chip manufacturing capacity in the U.S. is still in early stages, and these expansion plans may not significantly mitigate supply chain concentration risks in the foreseeable future.

Concentration Risks in the Chip and Memory Supply Chain

Besides advanced logic chips, Anthropic also highlighted its reliance on high-end memory chip suppliers. The company notes that the global advanced memory chip market is mainly supplied by SK Hynix, Samsung Electronics, and Micron Technology (MU.US). As AI model sizes grow, demand for high-performance memory chips for training and inference workloads continually increases, making supply capability a key constraint on AI infrastructure expansion.

Despite all three memory chip manufacturers actively ramping up production to meet rising demand, Anthropic warns that it is not yet certain whether these expansion plans will quickly yield effective supply to meet future needs.

For Anthropic, inadequate advanced chip and memory supply could not only delay new compute infrastructure deployment but also limit the training and inference capacity of the Claude model, thereby affecting product development and commercialization progress.

The company also points out that migrating existing AI workloads to other compute suppliers is costly, time-consuming, and may reduce computing efficiency. This means as Anthropic accelerates its AI business expansion, it faces multiple challenges including reliance on cloud providers, massive capital commitments, and a concentrated semiconductor supply chain. On one hand, Google, Amazon, and Microsoft possess vast enterprise customer networks, mature cloud service infrastructure, and substantial financial resources that can help Anthropic quickly expand the market coverage of its Claude model.

On the other hand, these companies not only control key sales channels and critical compute supply, but also develop competing AI products, putting constraints on Anthropic in business negotiations, pricing, and long-term strategy.

Furthermore, as global AI infrastructure construction accelerates, the supply capacity of advanced chip manufacturing and high-performance memory products will directly impact Anthropic’s ability to scale compute and meet customer demand as planned.

With Anthropic about to enter the public markets, investors will need to assess not only its revenue growth and profitability, but also its dependency on cloud giants, multi-billion-dollar compute procurement commitments, and potential long-term operational risks from concentrated semiconductor supply chains.