Nvidia told investors it expects to lock in far more components to support its data center roadmap after reporting second-quarter results. During the quarter, the company’s supply commitments rose to $279 billion over the next few years, up from $119 billion three months earlier, with management linking the increase largely to memory availability as Blackwell and Vera systems ramp.
In the same earnings update, Nvidia said its data center segment remained the primary driver of growth, underscoring that the limiting factor in expanding AI infrastructure is increasingly tied to specialized memory supply rather than only chip production capacity.
Key takeaways
- Price move: Nvidia’s shares were reacting to updated supply-commitment plans alongside its latest data center growth figures.
- Catalyst: Management disclosed that supply and capacity commitments will total $279 billion over the next few years, up from $119 billion earlier.
- What drove it: Nvidia attributed the increase primarily to memory requirements as Blackwell and Vera scale.
- Key implication: Investors are likely to focus on how higher memory costs flow through Nvidia’s cost structure and gross margin before any eventual benefit shows up in pricing.
What drove the jump in supply commitments
At Nvidia’s second-quarter earnings, CFO Colette Kress stated the company would be spending $279 billion on supply commitments over the coming years. She contrasted that with the $119 billion level cited just three months earlier, highlighting a rapid change in procurement plans.
According to Nvidia, the reason for the acceleration is tied primarily to memory. As generative AI workloads increase, Nvidia’s next-generation platforms—including Blackwell systems and Vera central processing units—require more high-bandwidth memory (HBM) and server DRAM, along with tighter packaging and system-level integration than earlier generations.
Nvidia’s management also indicated that memory is where most of its supply and capacity spend will be directed. The company’s budgeting suggests it is aiming to secure the most constrained part of the AI hardware supply chain well ahead of time rather than leaving demand to be matched by later spot purchases.
Data center momentum and the memory bottleneck
Data from Nvidia’s second-quarter results showed the data center business grew to $89 billion, up 117% year over year. The segment was also on track with a companywide outlook implying continued strength, including a guide of $108 billion for the following quarter.
Nvidia’s broader growth outlook referenced continued Blackwell shipments and scaling Vera production, alongside the integration of Vera CPUs with the company’s existing GPU architectures. Management also discussed Vera’s ramp toward full production and the pairing of components across the platform stack.
Industry investors have increasingly treated memory as the bottleneck in AI hardware build-outs. Nvidia’s procurement commitment effectively reframes memory from a background input into a critical gating factor for system availability—meaning supply constraints for HBM and DRAM can limit shipments even if GPU production capacity is available.
To put the commitment in context, Nvidia said it expects $92 billion of the total in the remainder of fiscal 2027, followed by $87 billion in fiscal 2028 and $88 billion in fiscal 2029. That schedule signals a multi-year attempt to reserve near-term supply rather than deferring procurement decisions.
How investors may interpret margins and costs
The supply build also has implications for profitability timing. When companies pre-commit to constrained components, some portion of the cost pressure can show up in gross margin before any corresponding revenue benefit is fully realized.
According to Nvidia, gross margin was 75% last quarter and is guided to 74% for the current quarter. Management also commented that an additional slide into the low 70% range is realistic by the end of fiscal 2027.
That guidance aligns with the logic that higher memory-driven input costs may weigh on profitability in the near term. At the same time, Nvidia’s message to the market is that securing supply is necessary to support the shipment ramp of Blackwell and Vera, potentially reducing the risk of lost deliveries later.
Why this could matter beyond Nvidia
Nvidia’s procurement approach implies a large, sustained demand profile for memory suppliers. The company’s management indicated it would focus most capacity-related spend on memory rather than distributing it evenly across a broader component mix.
Within that context, Nvidia’s commitment is likely to be read as a form of multi-year offtake for specialized memory components used in AI systems. The effect could be particularly significant for suppliers involved in HBM and qualified DRAM used in Nvidia’s platforms, where supply expansion typically takes time and requires substantial investment.
For investors, the key question is how much of Nvidia’s memory cost pressure shows up in margins versus how much stabilizes availability for data center customers. The procurement totals and the timing of spending are the clearest indicators of the direction of that trade-off.
What to watch next
Investors will likely monitor whether Nvidia’s gross margin trajectory holds to management’s guidance as memory costs flow through its cost base and as Blackwell and Vera shipment momentum continues. Upcoming company updates, including subsequent earnings reports and any changes to data center guidance, will be important for confirming whether supply commitments translate into sustained revenue growth.
Separately, market participants may also watch broader AI infrastructure themes that influence memory pricing and supply—particularly developments in the availability of HBM and DRAM—along with macro factors that can affect enterprise spending on AI data center expansion.
