Nvidia’s earnings beat expectations: quarterly revenue of $9.62 billion, doubling year over year. Guidance for next quarter is $108 billion, ahead of estimates, and the stock rose 6% in pre-market trading. But the real signal isn’t in the income statement—it’s in the balance sheet. Supplier purchase commitment letters jumped from $119 billion to $279 billion in a single quarter, more than doubling. This isn’t cost; it’s a bet—Nvidia locks in TSMC’s CoWoS capacity and HBM memory capacity ahead of time using long-term contracts and deposits.
Huang Renxun said demand is far more than the 70% growth guidance, but what’s holding it back is supply. This $279 billion is the quantified realization of that statement. Who benefits? TSMC and HBM manufacturers (SK Hynix and Micron) gain certainty on capacity expansion; the logic for scaling the secondary server supply chain also becomes smoother—provided they squeeze into Nvidia’s approved supplier list.
Who is hurt? AMD and cloud companies running in-house chips: the difficulty and cost of securing HBM and packaging capacity will rise. For second-tier AI startups, it will be harder to rent even ad hoc capacity; Nvidia is already providing financial support to customers with weaker credit—while locking capacity and simultaneously creating its own demand.
If we assume 70% growth, Nvidia’s fiscal 2028 revenue would reach $673 billion, surpassing Apple and Alphabet to become the second-largest U.S. tech company, behind only Amazon. Nvidia isn’t just selling chips—it’s reallocating the entire AI hardware supply chain’s capacity.
Next, watch TSMC’s monthly revenue and the ramp-up pace of CoWoS to see whether they match the timing of purchase commitment fulfillment, along with Nvidia’s inventory turnover days. If inventory growth far outpaces revenue growth, it suggests the cost of locking capacity is starting to eat into profits.
The biggest takeaway from this earnings report isn’t how strong demand is, but that Nvidia transfers demand uncertainty to its suppliers and welds supply certainty into its own hands.
Huang Renxun said demand is far more than the 70% growth guidance, but what’s holding it back is supply. This $279 billion is the quantified realization of that statement. Who benefits? TSMC and HBM manufacturers (SK Hynix and Micron) gain certainty on capacity expansion; the logic for scaling the secondary server supply chain also becomes smoother—provided they squeeze into Nvidia’s approved supplier list.
Who is hurt? AMD and cloud companies running in-house chips: the difficulty and cost of securing HBM and packaging capacity will rise. For second-tier AI startups, it will be harder to rent even ad hoc capacity; Nvidia is already providing financial support to customers with weaker credit—while locking capacity and simultaneously creating its own demand.
If we assume 70% growth, Nvidia’s fiscal 2028 revenue would reach $673 billion, surpassing Apple and Alphabet to become the second-largest U.S. tech company, behind only Amazon. Nvidia isn’t just selling chips—it’s reallocating the entire AI hardware supply chain’s capacity.
Next, watch TSMC’s monthly revenue and the ramp-up pace of CoWoS to see whether they match the timing of purchase commitment fulfillment, along with Nvidia’s inventory turnover days. If inventory growth far outpaces revenue growth, it suggests the cost of locking capacity is starting to eat into profits.
The biggest takeaway from this earnings report isn’t how strong demand is, but that Nvidia transfers demand uncertainty to its suppliers and welds supply certainty into its own hands.