Securing vast computing resources in anticipation of long-term growth
Expanding long-term contracts requiring payment even if usage declines
Slower growth could increase purchase burdens and disrupt investment recovery

Anthropic has placed a massive bet on the long-term growth of the artificial intelligence (AI) market by securing vast computing resources in advance. Buoyed by surging revenue, it has accumulated long-term purchase commitments, but many contracts require payment even if usage declines, leaving it exposed to mounting financial burdens if growth falters. Suppliers are also lending customers the money to fund purchases, creating a risk that failed investments could send shock waves through business partners and creditors. With equipment obsolescence shortening the time available to recoup investments and China's DeepSeek intensifying its low-cost offensive, uncertainty is growing over whether the company can generate enough revenue to support its enormous spending.

Long-Term Purchase Commitments Remain Daunting Despite a 15-Fold Revenue Surge

According to The Wall Street Journal (WSJ) on the 22nd (all dates hereafter are local time), Anthropic is pursuing plans to secure up to USD 750 billion (approximately USD 773 billion) in AI computing resources over the next several years. A draft of Anthropic's confidential initial public offering (IPO) registration statement reviewed by the WSJ outlines plans to secure large-scale computing capacity through cloud services and semiconductor leases. Infrastructure procurement obligations under contracts already signed exceed USD 600 billion (approximately USD 619 billion). Deals currently under discussion could add another USD 150 billion (approximately USD 155 billion) if completed. The plan involves securing in advance the computing resources customers will use over the coming years while taking on enormous payment obligations. Given the long time needed to recoup these investments, the company's financial capacity appears to hinge on the accuracy of its assessment of future demand.

The scale of the investment riding on future demand is substantial even considering the recent surge in revenue. Revenue in the first half of this year reached USD 16.3 billion (approximately USD 16.8 billion), 15 times the USD 1.1 billion (approximately USD 1.13 billion) recorded a year earlier. Third-quarter revenue, disclosed to investors last week, also rose 45% from the previous quarter to USD 16.8 billion (approximately USD 17.3 billion). In just three months, the company generated about 3.7 times last year's full-year revenue of USD 4.6 billion (approximately USD 4.74 billion). Against this backdrop of rapid revenue growth, Anthropic has signed massive purchase commitments to be fulfilled over the next 7–10 years. It is expanding its business by locking in future spending on the assumption that its current growth trajectory will continue for a long time.

However, many of the commitments make it difficult to cut spending even if demand falls short of expectations. Anthropic's draft registration statement outlines plans to spend at least USD 518 billion (approximately USD 534 billion) with six partners over the next 10 years, with about 80% covered by contracts that either cannot be canceled or require payment regardless of usage. Minimum spending commitments to major companies include USD 111.1 billion (approximately USD 115 billion) to Google, USD 110 billion (approximately USD 113 billion) to Amazon, and USD 31.4 billion (approximately USD 32.4 billion) to Microsoft. Under the agreements with Google and Amazon, Anthropic must pay the difference if its actual usage falls short of the committed amount. Equipment lease obligations related to Broadcom also total USD 161.2 billion (approximately USD 166 billion) and cannot be canceled except under contractual exceptions such as default. Because Anthropic must pay a substantial portion of the agreed amounts even if it cannot fully utilize the capacity it secures, the burden will remain with the company if its demand forecasts prove wrong.

Table 1. Anthropic's AI Computing Resource Procurement Plans and Purchase Commitments

Computing Resource Costs Consume Half of First-Half Revenue

As long-term payment obligations accumulate, immediate infrastructure spending is also rising rapidly. Computing and infrastructure spending in the first half of this year reached USD 9 billion (approximately USD 9.28 billion), four times the amount a year earlier. In just six months, it exceeded last year's full-year spending of USD 7.3 billion (approximately USD 7.53 billion) and accounted for more than half of revenue of USD 16.3 billion (approximately USD 16.8 billion) over the same period. The first-half net loss totaled USD 80.7 billion (approximately USD 83.2 billion), but USD 72 billion (approximately USD 74.3 billion) of that was a noncash accounting expense reflecting changes in the value of financial instruments such as convertible bonds. Valuation expenses resulting from the company's rising valuation greatly inflated its losses on paper. Separately from these valuation expenses, which involve no cash outflow, actual spending to secure and operate computing resources is increasing sharply.

As infrastructure spending surged, Anthropic also entered into an agreement to borrow a large sum from a computing resource supplier. Earlier this month, Broadcom agreed to lend up to USD 42 billion (approximately USD 43.3 billion) to support Anthropic's infrastructure spending. Broadcom may designate a separate financing partner, and the debt securities used to raise the funds are structured to be convertible into Anthropic shares in the future. If fully drawn, the facility could cover about one-third of the USD 125.2 billion (approximately USD 129 billion) committed to leasing tensor processing unit (TPU) computing capacity over five years. Semiconductor supply, equipment leasing, and financing are thus bundled together in the transactions with Broadcom. The registration statement includes provisions requiring the immediate payment of a substantial amount of lease charges if certain payment or contractual performance obligations are breached, while potentially restricting the use of the loan facility to cover those payments.

U.S. Treasury Secretary Remains Optimistic Despite Warnings of Cascading Losses

The transaction structure, under which Anthropic's financial difficulties could lead to losses for suppliers and creditors, also connects to the debate over risks across AI investment. In its Global Financial Stability Report this month, the International Monetary Fund (IMF) analyzed 14 major AI-related deals since 2023 involving USD 300 billion (approximately USD 309 billion) in equity investment and USD 650 billion (approximately USD 670 billion) in debt. The companies receiving the funds had committed to future service and semiconductor purchases totaling USD 1.4 trillion (approximately USD 1.44e+3 billion). Because investors also count the companies they finance as customers, a deterioration in those customers' businesses could simultaneously disrupt investment recovery and revenue generation. The IMF warned that if a core investee company falters, the value of equity holdings and loan claims could fall while expected revenue from that company is also impaired. This means that if Anthropic cannot make its promised payments, both the companies supplying computing resources and the creditors lending it money could be hit.

Following the IMF's warning about the cascading effects of troubled AI investments, U.S. Treasury Secretary Scott Bessent also weighed in, but his assessment of the risks to the financial system differed. At a private investor event in New York on the 21st, Bessent said of the AI investment boom, “I don't see a systemic problem.” Referring to Pets.com, which went public just before the dot-com bubble peaked in March 2000, he also said, “This is not Pets.com.” His point was that the AI boom differs from the dot-com bubble, when money poured into companies with little substance. He cited the investment capabilities and business track records of executives leading major technology companies as the basis for his assessment. He explained that executives who had built some of the world's most profitable companies believed in AI's potential while keeping most of their wealth invested in their own companies. He also expressed expectations that AI would boost productivity and contribute to U.S. economic growth.

Annual Revenue of USD 6 Trillion Needed if Equipment Lifespans Are Halved

However, for the technology companies' record of success emphasized by Bessent to be repeated in AI investment, earnings must grow enough to cover already committed capital spending. A Columbia University study submitted to a Brookings Institution academic conference last month estimated U.S. AI infrastructure investment in 2025–2032 at USD 10.3 trillion (approximately USD 10,623 billion). Assuming operating cash flow equals 50% of revenue, the study calculated that annual revenue would need to reach USD 3.7 trillion (approximately USD 3.82e+3 billion) in 2032 to achieve a 10% unlevered return on investment. Reaching that level would require both enough demand to fill the expanded capacity and growth in usage-fee revenue. The study noted that investment is exceeding companies' ability to finance it internally, adding debt burdens to businesses whose profitability has not yet been adequately proven.

There is also a risk that equipment will lose its competitive edge before investments can be recouped. In the same study, shortening the economic lifespan of information technology equipment from 6 years to 3 years increased the annual revenue needed to achieve the same return to USD 6 trillion (approximately USD 6.18e+3 billion). Additional costs would arise if the emergence of new semiconductors and cooling technologies required existing facilities to be replaced or retrofitted. The shorter the equipment replacement cycle, the less time there is to recoup investments. The study noted that principal and interest payments on borrowing used to fund equipment remain fixed, so a combination of slowing demand, operational delays, and declining collateral values could magnify investor losses and disrupt debt recovery.

Price Competition From Chinese AI Companies Poses Another Risk

The low-cost offensive by Chinese AI companies is another major variable. On the 20th, DeepSeek released DeepInfer, the inference engine used to generate responses from its models, as open source. Designed to reduce unnecessary computation and semiconductor idle time, it supports the latest AI chips from both Nvidia and Huawei. The focus is on lowering response-generation costs by handling more requests with the same equipment. DeepSeek also adopted the Apache 2.0 license, which permits commercial use, opening the way for outside companies to modify the code and incorporate it into their own services. If technologies that reduce operating costs spread and lead to lower service prices, U.S. AI companies seeking to recoup massive capital investments through usage fees could face greater pressure on their earnings.

The financial resources supporting DeepSeek's low-cost offensive are also growing. DeepSeek is reportedly finalizing a USD 15 billion (approximately USD 15.5 billion) funding round, having secured major investors including Chinese cloud services provider Tencent and battery manufacturer CATL. With more than 15 participants, including Meituan and Geely Auto, the amount raised has doubled from the original target. The company is also expanding the computing infrastructure into which it will deploy the funds. DeepSeek is increasing investment in computing infrastructure within China and working with Huawei to expand model-training workloads handled by Chinese-made chips. It is pursuing U.S. AI companies by simultaneously releasing technology that lowers inference costs, raising substantial funding, and expanding its training infrastructure.