엔비디아의 순환 논리: $5억 달러 베팅의 양날의 검 - AD HOC NEWS

The paradox at the heart of Nvidia’s current market position is almost too neat to be coincidence. The company that supplies the foundational tools for the artificial intelligence boom is now financing its own future customers, prompting critics to question whether the resulting demand is as durable as it appears.

This tension is most evident in the July 27 announcement that Nvidia would invest in Safe Superintelligence Inc. (SSI), the startup founded by former OpenAI chief scientist Ilya Sutskever. Bloomberg, citing sources familiar with the matter, valued the investment at approximately $5 billion. While neither party disclosed official financial terms, the strategic rationale is clear: SSI gains access to Nvidia’s upcoming Vera Rubin platform, which the company states will expand its computing capacity by an order of magnitude. In return, SSI becomes a chip customer—likely a substantial one.

The arrangement has drawn sharp scrutiny from a familiar quarter. Michael Burry’s Scion Asset Management has been steadily building a short position against Nvidia, while also adding bets against Micron and Caterpillar. Burry’s thesis is that a substantial portion of Nvidia’s current and projected demand does not originate from end users but is instead sustained by off-balance-sheet financing structures—a circular model in which future revenue depends on deals that ultimately channel capital back into Nvidia’s own products. Late last month, Burry publicly observed that five-year credit default swaps on Nvidia bonds were moving “parabolically,” a trend that coincided with reports the company had secured data center lease agreements worth up to $50 billion in Texas to support its “neocloud” partners. Bernstein’s Stacy Rasgon highlighted the same dynamic in early August.

While the concern is legitimate, it does not tell the whole story. Financing growth through circular arrangements shifts risk into the future—a familiar pattern in every technology boom. What distinguishes this cycle is the breadth of Nvidia’s partnerships, which extend far beyond any single transaction. Over recent weeks, Nvidia has forged a network of strategic alliances: an expanded partnership with South Korea’s SK Group spanning AI manufacturing facilities and memory technology; a joint initiative with NAVER and Brookfield to develop Korea’s national AI infrastructure; a research laboratory established with KAIST; and a national AI infrastructure program supported by the Japanese government and industry. This roster of state-backed and industrially anchored projects makes it difficult to dismiss the broader demand landscape as mere financial engineering.

Regardless of the structural debate, the market has been voting with its capital. Over a seven-day trading period, Nvidia shares rose 11.23 percent—reportedly the largest weekly gain in the company’s history, adding approximately $562 billion to its market capitalization. The stock closed Friday at €193.68, up 2.03 percent for the day, placing it just 4.36 percent below its May record high. Nvidia’s total market valuation now stands at €4,588.78 billion.

The catalyst for this rally was unmistakable: Elon Musk. SpaceX announced it would exclusively adopt Nvidia’s Vera Rubin architecture for data centers on Earth and in orbit, a commitment formalized under the “Starmind AI1” project—a satellite constellation engineered to process AI workloads in space. Given that Nvidia’s largest enterprise customers have spent years developing custom silicon to reduce reliance on the company, SpaceX’s exclusive partnership sends a contrasting message: at the highest tiers of computational demand, Nvidia’s architecture faces no serious competition.

Major cloud providers are reinforcing that sentiment with massive capital commitments. Alphabet has outlined 2026 capital expenditures between $195 billion and $205 billion, while Amazon is targeting $220 billion. A significant portion of that spending will flow directly into the type of infrastructure Nvidia provides.

Nvidia also signaled that it is not relying solely on hardware. On August 4, the company unveiled “Alpamayo 2 Super,” an AI model optimized for robotaxis that reportedly surpassed both Google’s Gemini 2.5 Pro and GPT-4o in key performance benchmarks. This vertical integration—delivering both chips and models from a single source—supports the analyst consensus price target of €261.99, which implies a 35.3 percent upside from Friday’s closing price.

Risks remain visible, even as the market currently chooses to overlook them. A severe shortage of High Bandwidth Memory (HBM) could compel Nvidia to scale back the HBM4E memory stack on its upcoming “Rubin Ultra” chip to accelerate production volumes and delivery timelines. While some investors interpret this as a potential margin pressure, the market has largely accepted it as a pragmatic trade-off when customers prioritize deployment speed over peak specifications.

China exposure introduces another layer of uncertainty. Media reports indicate that up to $5 billion in Chinese revenue could disappear following a single regulatory shift. Concurrently, corporate insiders have sold approximately $410 million in shares over the past three months while the stock traded at elevated valuations—a development that warrants monitoring, though not immediate alarm.

The next critical test arrives on August 26, when Nvidia releases its second-quarter fiscal 2027 earnings. Consensus estimates project year-over-year revenue growth of 96 percent. Currently, the market is pricing in the growth narrative rather than debating circular financing, and the structural case for Nvidia—rooted in broad, partially state-backed demand that is undeniably tangible—carries considerable weight. Nevertheless, the convergence of circular financing structures, lingering China-related risks, and notable insider selling should temper enthusiasm before investors mistake current momentum for a self-sustaining trend. With the relative strength index (RSI) at 63.5, the stock is nearing overbought conditions without fully entering them. At present, the world’s largest technology firms recognize no viable alternative to Nvidia’s roadmap. The enduring question is whether that reality holds once the memory constraints persist, rival architectures narrow the performance gap, and the circular financing arrangements mature.