Anyone budgeting for AI infrastructure in 2027 is planning around a constraint that Nvidia says will last until at least early 2028. On the US chipmaker’s August 26 earnings call, Chief Financial Officer Colette Kress said Nvidia expects “supply to remain a bottleneck” through at least the end of its fiscal year 2028, which closes in late January 2028. Its commitments, meanwhile, jumped from $119 billion to $279 billion in a single quarter, an increase the company’s CFO commentary calls “primarily related to the procurement of memory.”
Supply Stays Tight Until at Least Early 2028
Nvidia names its fiscal years for the calendar year in which they end, so fiscal 2028 runs from early 2027 to late January 2028. Kress called the roughly 70% revenue growth Nvidia expects for that year a preliminary expectation and a “supply-constrained outlook.” Customer forecasts point to growth of about double, she said, which leaves supply as the ceiling.
CEO Jensen Huang framed the gap this way: demand is “much greater than 70%,” he told analysts, but supply “allows us to confidently deliver 70%.” A Goldman Sachs analyst asked him to rank the most acute constraints, among them data center and power availability, DRAM and wafer foundry capacity. Huang did not pick one. His answer was that “our entire supply chain is challenged.”
What the $279 Billion Contains
The $279 billion is the supply and capacity line in a table of future commitments, as of July 26, that totals 366billion.Cloudserviceagreements(29 billion), data center leases not yet started (25billion),equityinvestments(25 billion) and capital expenditure ($8 billion) make up the rest. Nvidia schedules the supply line by fiscal year, and $267 billion of the $279 billion falls between the remainder of fiscal 2027 and fiscal 2029.
The memory attribution covers the $160 billion increase, not the full $279 billion. The commentary does not break either figure down by component, so the memory share of the total stays unknown. Kress said memory price increases have “exceeded our prior expectations” and are “headed even higher into next year.”
Kress tied the margin reset to those costs. “As a result, we are resetting expectations today,” she said, and guided third-quarter gross margin to 74%, down from 75.0% in the second quarter. She expects margins to bottom at 71% to 72% in the fourth quarter of fiscal 2027 before settling at 72% to 73% in fiscal 2028 “as executed price increases take effect in Q1.” The sequence suggests Nvidia expects to recover part of the memory bill through its own pricing.
Memory Makers Describe the Same Wall
Manish Bhatia, president and chief operating officer of US memory maker Micron, said on the company’s September 30 analyst call that more than 75% of its shipments for fiscal 2027 are committed, a year that runs roughly from September 2026 to August 2027. He added that Micron has no “line of sight” to when supply and demand will balance. On October 28, Park Jun-deok, vice president in charge of DRAM marketing at South Korea’s SK hynix, said virtually all of the company’s DRAM and high-bandwidth memory volume through 2029 is effectively sold out, as reported by South Korea’s Seoul Economic Daily.
Micron and Nvidia weigh the bottlenecks differently. Bhatia said DRAM is “the principal constraint versus logic or power to the data center,” while Huang answered that the whole supply chain is challenged. SK hynix posted a record third-quarter operating margin of 79.7%, which Chief Financial Officer Kim Woo-hyun attributed to “higher product prices and expanded sales centered on high-value products.” The implication is that memory makers profit from tight supply, so their outlooks are company statements, not neutral measurements.
What Memory Prices Do to Infrastructure Budgets
The US research firm Gartner forecast on October 21 that worldwide spending on data center systems will rise 73.8% to $879 billion in 2026 and 55.5% to $1.366 trillion in 2027. It said demand from data center systems has driven memory prices to more than double over the past 12 months. Gartner analyst John-David Lovelock expects “memflation” to persist through 2027, and said higher average selling prices “continue to support spending growth” while the pressure on affordability is expected to suppress end-user demand. The release does not separate how much of the growth is price and how much is volume.
Contract terms are already adjusting. On its March earnings call, the US server maker Hewlett Packard Enterprise disclosed that it had amended its quoting terms to include a right to reprice existing orders for commodity cost increases between quote and shipment. CEO Antonio Neri said HPE expects elevated memory and storage prices to persist well into 2027. HPE has also shortened its quote cycle and steers customers toward lower-memory configurations, Network World reported. For planners, an approved quote may not match the final invoice, at least with this vendor.
Startups and PC Buyers See the Squeeze Differently
Startups sit one step removed from Nvidia’s cloud financing arrangements. According to the CFO commentary, Nvidia has partnered with AI clouds to serve startups, enterprises and other customers, earning revenue on the upfront sale of its infrastructure and, if certain criteria are met, a share of the clouds’ revenue from their own customers. Kress told investors that Nvidia compute is “fully utilized across every cloud we serve.” The Wall Street Journal reported on August 27, citing people familiar with the matter, that Nvidia had paused some of these deals after employees raised concerns that the initiative could draw antitrust scrutiny. An Nvidia spokesperson said the model “is still in place and continues to evolve due to high demand.”
Consumer hardware shows the pressure too. Nvidia’s CFO commentary lists “slower consumer PC sales that were tempered by elevated memory and systems prices” as a drag on its edge computing revenue. The US research firm IDC’s preliminary data show third-quarter PC shipments fell 20.1% from a year earlier to 62.7 million units. IDC’s Jitesh Ubrani blamed part of the slump on a first-half pull-in, as vendors and channels loaded up on inventory early in the year to get ahead of price hikes. The US firm Jon Peddie Research offers a similar theory for graphics cards, whose second-quarter shipments rose 10% from the prior quarter to 12.5 million: consumers bought before prices climbed further.
Nvidia said on its August call that it will discuss third-quarter fiscal 2027 results on November 17. The commitments total, the gross margin outlook after memory price increases and the preliminary 70% growth expectation for fiscal 2028 will show whether the picture Kress described in August still holds. Until then, signals from Nvidia, Micron, SK hynix and Gartner suggest memory supply and price will keep shaping AI infrastructure budgets into 2028.
The post Nvidia’s $279 Billion Supply Commitment: Memory Drives the Jump as Constraints Last Into 2028 appeared first on DataFLOQ.
