Nvidia's $1.5B Prepayment: An Option, Not an Investment

I noticed an interesting detail: The payment method for this $1.5 billion collaboration between NVIDIA and Amkor is not equity injection, nor joint venture factory construction, but prepayment. From a financial perspective, this structural choice reveals more information than the amount itself.

Let's break down the financial substance of this deal. Prepayments are legally classified as debt claims and usually recorded as "other non-current assets" or "prepaid accounts" in accounting. For NVIDIA, this $1.5 billion won't directly increase fixed assets, nor will it be depreciated or amortized; instead, it will gradually convert into packaging service procurement costs as Amkor delivers capacity over the coming years. This means NVIDIA's capital expenditure intensity does not rise, and free cash flow won't deteriorate significantly due to this money—for someone like me who focuses on cash flow health, this is good news.

But more critical is the return mechanism of this prepayment. After receiving this money, Amkor will expand advanced packaging capacity on US soil, and what does NVIDIA get? From the agreement wording, it's "priority capacity assurance" and "joint development of next-generation technology." This is essentially a call option—NVIDIA uses $1.5 billion to lock in priority usage rights for future US domestic advanced packaging capacity, with the exercise price being the subsequent packaging service fees.

Why is this option needed? Because the packaging stage of AI chips is becoming a bottleneck for the entire supply chain. Technologies like CoWoS and 3D packaging are currently highly concentrated in the hands of a few foundries like TSMC and Samsung, with most capacity located in Asia. Geopolitical risk premiums are being repriced by the market, and for NVIDIA, the world's highest-market-cap semiconductor company, supply chain singularity is the biggest risk exposure in valuation models. If any change occurs in the Taiwan Strait situation, NVIDIA's chip delivery capability could instantly drop to zero, and the capital markets would directly cut over 30% of the valuation premium.

Now let's look at NVIDIA's response strategy: It didn't build its own packaging plant—that's too asset-heavy, and ROE would drop significantly. Nor did it choose to add prepayments to TSMC, because TSMC's capacity is already snapped up by global customers, and the gap in US domestic advanced packaging capacity remains. As the world's second-largest OSAT (Outsourced Semiconductor Assembly and Test) provider, Amkor already has a presence in Arizona, USA. This collaboration is essentially replicating a CoWoS-level packaging line on US soil.

From a financial perspective, the impact of this transaction on NVIDIA's balance sheet is very limited. $1.5 billion is less than 5% of NVIDIA's cash reserves of over $30 billion, and this money will likely be paid over several years, with annual cash outflows of about $300-500 million. In contrast, NVIDIA's single-quarter free cash flow exceeds $10 billion, so it doesn't affect buyback and dividend plans at all.

But for Amkor, the significance of this prepayment is completely different. Amkor's capital expenditures have been relatively conservative in recent years, with about $1 billion in capex in 2023, and cash flow wasn't abundant either. Getting NVIDIA's prepayment is equivalent to obtaining a source of zero-cost long-term funds, allowing them to start expansion early without taking bank loans or issuing bonds. At the same time, NVIDIA's endorsement will give Amkor more bargaining power in subsequent financing and customer acquisition.

However, I noticed a risk point: The recovery issue of the prepayment. If Amkor's capacity construction progress lags behind expectations, or if there are deviations in the technical roadmap, who bears the impairment risk of this prepayment? From the agreement structure, this should be an arrangement that is "refundable" or "deductible against future service fees," but specific clauses haven't been disclosed. If NVIDIA doesn't have sufficient legal safeguards, once Amkor defaults, the $1.5 billion could turn into bad debt. Of course, with NVIDIA's legal team, this risk should have already been hedged.

Finally, revealing my core viewpoint: This $1.5 billion is essentially not an investment, but insurance purchased by NVIDIA for supply chain risks. The premium for the insurance is the cost of capital for the prepayment, and the coverage amount is priority usage rights for US domestic capacity in the coming years. In the valuation logic of capital markets, a stable supply chain can support a higher P/E ratio because uncertainty is reduced. If NVIDIA avoids even one capacity interruption thanks to this insurance, the $1.5 billion is worth the ticket price.

From a financial perspective, the elegance of this transaction lies in allowing NVIDIA to gain strategic buffer without increasing asset burdens or affecting return on capital.

Original link: https://www.ithome.com/0/980/882.htm