# Nvidia's Suspicious "Round-Trip" Transactions Source: https://www.youtube.com/watch?v=ahN7JfeOhEw Recap page: https://rapidrecap.app/video/ahN7JfeOhEw Generated: 2025-10-12T13:43:50.16+00:00 --- ## Quick Overview Nvidia's massive revenue growth is highly suspicious because the company engages in complex "round-trip" financial transactions, such as the $100 billion investment into OpenAI, which mimics accounting manipulations used during the dot-com bubble by companies like AOL and Global Crossing to inflate demand artificially, suggesting the current AI boom's revenue recognition is not supported by true end-user spending. **Key Points:** - Nvidia's data center revenue skyrocketed since late 2023, resulting in $23 billion in AI-related GPU sales beyond its $3 billion quarterly baseline, driven primarily by hyperscalers purchasing hardware for AI startups. - The top 20 AI startups generate an estimated annual run rate of only $15 billion in revenue, which is minuscule compared to the $116 billion in data center GPUs Nvidia sold in the first three quarters of 2025. - OpenAI, the largest AI startup, expects to generate $12.7 billion in revenue in 2025 but anticipates losing money, with internal projections not seeing positive free cash flow until 2029, and even their most expensive Pro tier loses money. - Nvidia engaged in deals like guaranteeing Coreweave excess capacity purchases up to $6.3 billion if demand falls short, and paying $1.5 billion to Lambda Labs to lease back GPUs it previously sold them, demonstrating self-stimulation of demand. - The $100 billion investment into OpenAI, where the money flows from Nvidia to OpenAI and then back to Nvidia via GPU purchases, is cited as the biggest roundtrip deal, functionally moving money in circles to inflate revenue figures. - The current AI boom mirrors the dot-com bubble, where infrastructure providers like Global Crossing inflated revenue through capacity swaps, though Nvidia's essential infrastructure role provides a stronger parallel to broadband companies than to failed dot-coms like Pets.com. **Context:** The video analyzes the unprecedented surge in Nvidia's valuation, reaching $4.6 trillion, driven by the AI investment theme following the release of ChatGPT in November 2022, as its GPUs are critical for training large language models. The analysis focuses on whether the massive capital expenditures by hyperscalers like Microsoft and Google, and the subsequent revenue recognized by Nvidia, are sustainable given the current low revenue generation and high losses reported by the end-user AI startups, drawing direct parallels to historical financial bubbles. ## Detailed Analysis Nvidia's success hinges on continued massive capital expenditures from hyperscalers, yet end-user demand from AI startups is weak; the top 20 AI startups combined only generate about $15 billion in annual revenue, while Nvidia sold nearly $180 billion in data center GPUs in 2025, with the deficit being covered by venture capital, which accounts for a vast percentage of AI startup spending on compute resources. The core concern is that Nvidia is using its own capital to artificially stimulate this demand through "round-trip" transactions; for instance, the announced $100 billion investment in OpenAI will result in OpenAI immediately spending that cash on Nvidia GPUs, creating circular revenue that does not reflect organic end-user adoption. This behavior is compared to AOL inflating revenue in the late 1990s and, more accurately, to broadband infrastructure companies like Global Crossing using capacity swaps to recognize revenue far exceeding actual consumer demand before collapsing under debt. The analysis concludes that while AI infrastructure is vital, the current valuation relies on a financial structure where Nvidia is essentially paying its customers to buy its product, raising serious questions about the long-term sustainability once VC funding slows. ### AI Ecosystem Financial Disparity - AI startups generate ~$15B annual run rate revenue while Nvidia is on track to sell ~$160B in data center GPUs in 2025 - Hyperscaler capex surged by ~$190B over two years for AI data centers - AI startups are losing money, even OpenAI expects losses until 2029. ### Nvidia's Demand Stimulation Tactics - Nvidia invested up to $100B into OpenAI contingent on GPU deployment - Nvidia guaranteed Coreweave excess capacity purchase up to $6.3B if demand falls short, derisking Coreweave's borrowing - Nvidia paid $1.5B to Lambda Labs to lease back GPUs previously sold to them. ### Dot-Com Bubble Parallels - The current boom is reminiscent of the dot-com bubble where investor hype led to malinvestment - AOL used round-trip transactions when revenue growth slowed - Global Crossing inflated revenue via capacity swaps with other broadband companies before bankruptcy in 2003. ### Valuation Concerns - Nvidia holds a P/E ratio of 54, expecting massive future growth - OpenAI's $500B valuation in October 2025 corresponds to a 39x price-to-sales multiple despite ongoing losses - The $100B OpenAI investment requires hundreds of billions in future revenue for OpenAI to generate a positive ROI.