Anthropic (ANT.US) races toward a U.S. IPO! Goldman Sachs Group, Inc. and Morgan Stanley lead a syndicate of 22 investment banks as underwriters; Alphabet Inc. Class C (GOOGL.US) and Amazon.com, Inc. (AMZN.US) contribute nearly half of revenue; computing power supply chain risks draw attention.

Anthropic (ANT.US) races toward a U.S. IPO! Goldman Sachs Group, Inc. and Morgan Stanley lead a syndicate of 22 investment banks as underwriters; Alphabet Inc. Class C (GOOGL.US) and Amazon.com, Inc. (AMZN.US) contribute nearly half of revenue; computing power supply chain risks draw attention. Artificial intelligence startup Anthropic (ANT.US) is accelerating its initial public offering (IPO), with Goldman Sachs Group, Inc. (GS.US) and Morgan Stanley (MS.US) securing the core underwriting roles for the listing, joined by 20 other investment banks. In addition, the IPO registration filing Anthropic published on Monday revealed the complex business relationships behind its rapid growth: Alphabet Inc. Class C (GOOG.US, GOOGL.US), Amazon.com, Inc. (AMZN.US) and Microsoft Corporation (MSFT.US) are not only important investors, computing power suppliers and sales channels for the company, but also its potential competitors. The company also warned that the highly concentrated global advanced chip manufacturing landscape and its reliance on key suppliers such as Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR (TSM.US) could pose risks to its long-term business development. According to the S-1 registration statement Anthropic filed, Goldman Sachs Group, Inc. was designated as the lead-left underwriter for the IPO, topping the list of underwriters, with Morgan Stanley ranked second and set to serve as stabilizing agent, responsible for taking measures to stabilize the stock price when necessary in the early stages of trading. This arrangement mirrors the underwriting structure of SpaceX's (SPCX.US) record-breaking IPO in June this year, when Goldman Sachs Group, Inc. and Morgan Stanley likewise held the top two core underwriting seats. SpaceX raised US$86.2 billion through its listing, and the 22 banks involved in the deal collectively received a fixed underwriting fee of US$500 million, equivalent to about 0.58% of total proceeds. Of that, Goldman Sachs Group, Inc. and Morgan Stanley each received US$100 million, the highest among all participating banks. Besides these two investment banks, Anthropic's IPO underwriting syndicate also includes 20 banks such as JPMorgan Chase (JPM.US), Citigroup (C.US), Barclays PLC Sponsored ADR (BCS.US), Wells Fargo & Company (WFC.US) and Bank of America Corp (BAC.US), bringing the total number of underwriters to 22, the same scale as SpaceX's previous IPO. Listing window and financial quality Anthropic's listing plan comes at a critical time when the AI industry faces investors re-examining it. Demand for AI technology continues to grow, but massive infrastructure investment, rising computing power costs, and whether companies can achieve sustainable profitability are becoming focal points for public market investors. Financial data disclosed by Anthropic show that the company's second-quarter revenue reached US$11.6 billion and it achieved operating profit of US$974 million, indicating that its AI business is rapidly scaling up. However, earlier this month, reports that Anthropic's third-quarter revenue growth had slowed exacerbated selling pressure on tech stocks, reflecting market concerns that AI capital expenditure growth may outpace actual demand. Notably, Anthropic's rapid growth relies heavily on sales channels provided by large cloud service providers. According to the IPO filing, in the first half of 2026, sales through the cloud marketplaces of hyperscalers such as Alphabet Inc. Class C, Amazon.com, Inc. and Microsoft Corporation accounted for 48% of Anthropic's total revenue, up sharply from 11% in 2023. Under this sales model, cloud service providers charge customers directly, deduct channel service fees and then pay the remainder to Anthropic. As this channel has grown in importance, the fees Anthropic pays to channel partners have also risen rapidly. In the first half of 2026, related fees increased by US$2.7 billion compared with the same period last year. This means that although Anthropic has gained a vast enterprise customer base through the distribution networks of cloud giants, its dependence on these partners is also deepening. Cloud channel dependence and "four-role" risk Anthropic explicitly stated in its prospectus that Alphabet Inc. Class C, Amazon.com, Inc. and Microsoft Corporation simultaneously play multiple roles as the company's "investors, customers, distribution partners and competitors." This complex relationship also brings potential risks. Because cloud service providers are responsible for charging end customers, they are able to control the pricing and commercial contract terms of Anthropic's products. At the same time, these companies also have their own AI models and products, so they may in the future prioritize promoting their own technology rather than Anthropic's Claude models. Anthropic also warned that its close relationships with large cloud computing companies have already drawn the attention of antitrust regulators and may face further scrutiny in the future. Beyond sales channels, Anthropic is also highly dependent on the computing resources provided by these cloud giants for AI model training and inference. According to the prospectus, Anthropic has committed to paying Alphabet Inc. Class C at least US$114.5 billion before 2033, at least US$110 billion to Amazon.com, Inc. before 2036, and US$44.9 billion to Microsoft Corporation. Even if it does not fully use the relevant computing power, the company must still fulfill these minimum spending commitments. The three commitments total US$269.4 billion, underscoring the enormous financial obligations Anthropic has taken on to secure long-term computing power supply. In addition, Anthropic expects to pay an additional US$125 billion to US$150 billion to Alphabet Inc. Class C and other suppliers between 2028 and 2031 to obtain more tensor processing unit (TPU) computing resources. This arrangement also reflects the complex relationship of both cooperation and competition between Anthropic and cloud service providers. In April this year, when Anthropic announced a cooperation agreement with Amazon.com, Inc., it described Amazon.com, Inc.'s cloud unit (AWS) as the primary training and cloud service provider for its key workloads. However, according to the latest disclosure, Anthropic's minimum spending commitment to Alphabet Inc. Class C is not only larger in scale but also has a shorter performance period. Amazon.com, Inc. and Alphabet Inc. Class C are also important shareholders of Anthropic. According to the prospectus, Amazon.com, Inc. has invested about US$18 billion in total since agreeing in 2023 to invest up to US$4 billion in Anthropic, and has committed to investing up to an additional US$15 billion once certain conditions are met. According to media calculations, on a converted fully diluted share basis, Amazon.com, Inc. currently holds about 9% of Anthropic's economic interest. In addition, Amazon.com, Inc. also holds about US$6.2 billion in outstanding convertible notes issued under financing agreements signed in 2023 and 2024, which can be converted in the future into 1.84 billion Anthropic shares, equivalent to an additional roughly 11% equity stake. However, Amazon.com, Inc.'s shareholding is in principle subject to a 12% cap, though the company has the right to waive this limit. As for Alphabet Inc. Class C, it has invested about US$23 billion in Anthropic through multiple rounds of transactions and has committed to investing an additional US$20 billion subject to the achievement of performance targets. According to calculations, Alphabet Inc. Class C currently holds at least about 6.6% of Anthropic's economic interest. In addition, Alphabet Inc. Class C still has US$20 billion of remaining financing capacity under a financing agreement signed in 2026. If Anthropic draws on the relevant funds, it may further issue shares or convertible notes. Under the latest financing agreement, Alphabet Inc. Class C's shareholding cap is 14.99%, but it can likewise choose to waive this limit. Notably, Alphabet Inc. Class C also holds an option to subscribe, through a private placement at the public offering price, for up to US$10 billion of shares concurrently with Anthropic's IPO. If Alphabet Inc. Class C exercises this option, the relevant investment amount will correspondingly reduce the remaining capacity under its existing financing agreement. Massive computing power commitments and shareholder ties In addition to the three cloud computing giants, Anthropic has also established large-scale computing power supply relationships with other tech companies. According to the prospectus, Anthropic's equipment lease commitments with Broadcom Inc. (AVGO.US) total US$161.2 billion, and it has agreed to purchase up to US$120.5 billion of computing resources from SpaceXAI. However, Anthropic's dependence on the computing power supply chain does not stop at cloud service providers and equipment suppliers. The company also emphasized in its IPO prospectus that the highly concentrated advanced semiconductor manufacturing capacity could become an important risk factor affecting its business development. Anthropic noted that the vast majority of the world's most advanced chips are still produced by Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR, whose main advanced manufacturing operations and supply chain are concentrated in Taiwan. This concentration means that any major event affecting local advanced chip production or supply could have a lasting impact on the global AI industry chain and further affect Anthropic's business operations, financial condition and operating results. This risk also highlights the critical position of Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR in the global artificial intelligence industry. AI chip design companies such as NVIDIA Corporation (NVDA.US) and AMD (AMD.US) rely heavily on Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR's advanced manufacturing processes to turn chip designs into actual products. Even tech giants such as Amazon.com, Inc. and Alphabet Inc. Class C that develop their own AI chips still need to rely on specialized wafer foundries to manufacture chips. Anthropic said in its prospectus that restrictions or delays in the supply of advanced chips could adversely affect the company's business, financial condition and operating results. In fact, dependence on Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR's advanced manufacturing capacity has become a common feature of the global AI industry. In addition to Anthropic, companies such as NVIDIA Corporation, Broadcom Inc. and Marvell Technology, Inc. (MRVL.US) have previously highlighted the potential risks posed by semiconductor supply chain concentration in relevant disclosures. Although Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR is actively expanding global capacity, the highly concentrated advanced chip manufacturing landscape is unlikely to change fundamentally in the short term. According to relevant reports, Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR has committed to investing US$265 billion in Arizona to build advanced semiconductor facilities and is considering building a new large-scale production base in Texas. However, Anthropic noted in its prospectus that the buildout of advanced chip manufacturing capacity in the United States is still at a relatively early stage, and related expansion plans may not significantly reduce existing supply chain concentration risks for the foreseeable future. Chip and memory supply chain concentration risk In addition to advanced logic chips, Anthropic also specifically mentioned its dependence on high-end memory chip suppliers. The company noted that the global advanced memory chip market is mainly supplied by three companies: SK Hynix, Samsung Electronics and Micron Technology, Inc. (MU.US). As AI models continue to grow in scale, demand for high-performance memory chips for training and inference workloads continues to rise, and related supply capacity has become an important factor constraining the expansion of AI infrastructure. Although all three memory chip manufacturers are actively expanding capacity to meet growing market demand, Anthropic warned that there is no guarantee that these expansion plans will be able to form effective supply quickly enough to meet future demand. For Anthropic, insufficient supply of advanced chips and memory products could not only delay the deployment of additional computing infrastructure but also limit the training and inference capabilities of the Claude models, thereby affecting the company's product development and commercialization progress. The company also noted that migrating existing AI workloads to other computing power suppliers is not only costly and time-consuming but may also reduce computational efficiency. This means that while Anthropic accelerates the expansion of its AI business, it also faces multiple challenges such as dependence on cloud service providers, massive capital commitments and semiconductor supply chain concentration. On the one hand, Alphabet Inc. Class C, Amazon.com, Inc. and Microsoft Corporation have vast enterprise customer networks, mature cloud service infrastructure and strong financial strength, which can help Anthropic rapidly expand the market coverage of the Claude models. On the other hand, these companies control important sales channels for Anthropic and part of its key computing power supply, while also developing AI products that compete with Claude, leaving Anthropic somewhat constrained in commercial negotiations, pricing and long-term strategy. In addition, as global AI infrastructure buildout continues to accelerate, the supply capacity of advanced chip manufacturing and high-performance memory products will also directly affect whether Anthropic can expand its computing power scale as planned and meet customer demand. As Anthropic prepares to enter the public market, investors will need to assess not only its revenue growth and profitability but also closely watch its dependence on cloud giants, its hundreds of billions of dollars in computing power procurement commitments, and the long-term operating risks that key semiconductor supply chains may bring.