Inside the murky new AI chip economy - FT中文網
登錄×
電子郵件/用戶名
密碼
記住我
請輸入郵箱和密碼進行綁定操作:
請輸入手機號碼,透過簡訊驗證(目前僅支援中國大陸地區的手機號):
請您閱讀我們的用戶註冊協議私隱權保護政策,點擊下方按鈕即視爲您接受。
輝達

Inside the murky new AI chip economy

The growth of loans secured by Nvidia』s GPUs is a note for caution

Financial history is littered with weird and wonderful examples of collateral. In the 19th century, for instance, Peru used its future earnings from guano — a substance made from bat, bird and seal droppings — to secure loans for large-scale projects. The pooey mixture was an effective fertiliser, and readily available in the nearby Chincha Islands. Today, securities are thankfully less pungent, though not necessarily less toxic. Dodgy mortgage-backed securities helped trigger the 2008 financial crisis. What, then, to make of the latest financial innovation: collateralised artificial intelligence chips?

The Financial Times has reported that Wall Street’s largest financial institutions had loaned more than $11bn to “neocloud” groups, backed by their possession of Nvidia’s AI chips. These companies include names such as CoreWeave, Crusoe and Lambda, and provide cloud computing services to tech businesses building AI products. They have acquired tens of thousands of Nvidia’s graphics processing units (GPUs) through partnerships with the chipmaker. And with capital expenditure on data centres surging, in the rush to develop AI models, the company’s chips have become a precious commodity.

Euphoria over new technologies often goes hand in hand with financial innovation, which also reinforces it. Two centuries ago, during the railway boom in America and Britain, some railroad companies secured loans to lay more tracks, backed in part by their existing routes. Neoclouds are emulating them today. They provide data storage infrastructure for AI developers via power-purchase agreements. The loans they obtain from the likes of Blackstone, Pimco, Carlyle and BlackRock, secured by Nvidia chips, then allow them to buy more chips. In the event of a default, the lenders would acquire their chips and leasing contracts.

The rapid growth of a new debt market in a still nascent industry requires a note for caution. First, chips are unlikely to hold their collateral value over the long-term. Although GPU demand remains high, supply has risen as hardware reserves have been resold and could rise even further when leasing contracts expire. New chips developed by Nvidia, or its wannabe competitors, which include Microsoft, Google and Amazon, could also undermine the value of existing collateral.

Second, the deals may stretch valuations in the sector. The precise details of the arrangements between Nvidia and the neoclouds are unclear. But the chipmaker is itself an investor in some of the start-ups, which are in turn among its largest customers. Armed with Nvidia chips to secure loans, the cloud providers can then use the capital to buy more chips from Nvidia. This dynamic could inflate Nvidia’s earnings, and means the neocloud groups risk becoming highly leveraged, too. Third, the tie-ups with cloud providers could allow Nvidia to maintain the dominance of its chips, which adds to market concentration risks.

The chips-for-security trend is still young, and based on current lending volumes Wall Street’s largest financiers are perhaps not too concerned about their exposure just yet. But the development does shine a light on some risky lending, circular financing and competition dynamics that are propping up the AI boom. Investors ought to be wary of the potential pitfalls. Nvidia may be wise to draw clearer lines between its commercial and venture interests, which would support market transparency.

Financial innovation is often positive, and done well, it can channel capital to growth-enhancing projects. But as billions of dollars continue to flow into AI infrastructure, the pressure on developers to generate revenue is mounting. If risky and opaque financial engineering continues to feed the frenzy, prices risk moving further from reality. In that case, the deeper and wider any pain will be should there be a correction.

版權聲明:本文版權歸FT中文網所有,未經允許任何單位或個人不得轉載,複製或以任何其他方式使用本文全部或部分,侵權必究。

對話Otter.ai的梁松:我們可以從會議和對話中獲取有價值的數據

這家會議轉錄新創公司的聯合創辦人認爲,我們甚至可以用虛擬形象代替自己進行工作互動。

蕭茲迎來自己的「拜登時刻」

德國總理受到黨內壓力,要求其效仿美國總統拜登退出競選。

歐盟極右翼黨團在氣候和高層任命問題上獲得更多支援

歐洲議會中右翼議員正越來越多地與極右翼聯手瓦解該集團的綠色議程,並推動更嚴格的移民限制措施。

毛利人對紐西蘭後阿德恩時代的民粹主義轉向感到憤怒

盧克森的保守黨政府推翻了前總理的許多進步政策。

Lex專欄:輝達令人炫目的成長與每個人都息息相關

這家晶片巨擘的盈利對美國股票投資者來說是一件大事,這不僅僅是因爲其3.6兆美元的市值。

歐洲比以往任何時候都更需要企業成長冠軍

歐洲正在急切地尋找企業成長冠軍,FT-Statista按長期收入成長對歐洲企業進行的首次排名展示了這方面的可能性。
設置字型大小×
最小
較小
默認
較大
最大
分享×