Nvidia is telling its largest customers to expect AI server prices to rise more than 15% on systems shipping in early 2027, a jump driven by a scramble for scarce memory chips rather than the GPUs themselves. The increase lands just three days before Nvidia reports quarterly earnings, adding a fresh variable to a stock that Wall Street has spent the past week describing as unusually cheap for a company at the center of the AI buildout.
Bull Theory flagged the report, noting the increases are driven by surging memory chip costs and expected to hit systems shipping in early 2027 that use Nvidia's flagship Vera Rubin and Grace Blackwell architectures. The Kobeissi Letter added that the hikes affect both chip families, with pricing ultimately varying by configuration. Bloomberg, which first reported the increases, said companies that build servers under contract for hyperscalers including Microsoft, Google and Oracle have already notified their own customers, with Blackwell-based NVL72 GB200 racks priced between $2.8 million and $3.4 million and early inquiries on Vera Rubin systems running $5 million to $7 million per unit.
Memory, Not Compute, Is the Bottleneck
The driver isn't a shortage of Nvidia's own silicon. DRAM and high-bandwidth memory from Samsung, SK Hynix and Micron have become one of the largest single line items in an AI server's bill of materials, and shortages across all three suppliers have pushed those costs up sharply through 2026. That dynamic means the price increases will ripple through every company building AI infrastructure on top of Nvidia hardware, not just Nvidia's own margins.
A Cheap Stock Heading Into a Big Print
Coin Bureau noted that the $5.28 trillion company is trading at its cheapest valuation in five years heading into Wednesday's earnings, after guiding $91 billion in revenue for the quarter excluding China data center sales. Independent data backs that framing: Nvidia trades around 31 times trailing earnings and 23 times forward earnings, multiples not seen in half a decade and roughly 41% below its own 10-year average, even as the company posted an all-time-high $81.6 billion in quarterly revenue last quarter.
Related: SpaceX Stock Drops 12.8% in Two Months as Share Unlocks Bite
Put together, the two data points describe a company whose pricing power over its own customers is intact, even as its stock multiple has compressed relative to its growth. Investors will be watching Wednesday's print for guidance that reconciles the two: whether the memory-driven cost pressure is being passed through cleanly enough to protect margins, and whether demand for Blackwell and Vera Rubin systems is strong enough to absorb list prices that are rising even as competition from open-weight and Chinese AI models intensifies.
Why Crypto Traders Are Watching Too
Nvidia's earnings have become a de facto macro event for risk assets broadly, crypto included, given how closely Bitcoin has traded alongside the Nasdaq for most of 2026. A stock trading at a five-year-low multiple heading into results, combined with confirmation that AI infrastructure costs are still climbing, sets up Wednesday's report as a swing factor for risk appetite well beyond semiconductor investors.
