Nvidia CFO Says Demand Points to 140% Growth: Supply Delivers Only Half – Tech Times

The most important disclosure from Nvidia’s record-setting earnings call on August 26, 2026, was not the $96.2 billion in quarterly revenue or the $108 billion guidance for Q3. It was a sentence from CFO Colette Kress that enterprise procurement teams and data center planners need to read before they finalize their 2027 infrastructure budgets: customer forecasts point to doubling next year, and Nvidia’s own guidance holds at 70% because the supply chain cannot keep up. That gap — between what buyers want and what memory manufacturers can deliver — is what makes AI server prices rise 15% or more on systems shipping in early 2027 and what will keep them elevated for at least 16 more months.
The constraint at issue is high-bandwidth memory — the ultra-fast, stacked chip architecture that feeds data to Nvidia’s AI accelerators — along with the advanced semiconductor packaging process required to assemble those chips alongside GPU dies. Together they form a dual-bottleneck structure that no single capital investment decision can quickly dissolve, and Nvidia’s own supply commitment figures confirm it: the company’s supply commitments more than doubled in a single quarter, from $119 billion at the end of Q1 to $279 billion at the end of Q2 of fiscal 2027, primarily driven by memory procurement for the Vera Rubin platform.
Nvidia’s second fiscal quarter of 2027, ended July 26, 2026, produced results that would look extraordinary from almost any company in any industry. Total revenue reached $96.2 billion, up 18% from the previous quarter and up 106% year-over-year — a performance reflected in record quarterly revenue and earnings that beat Wall Street consensus by approximately $4 billion. Data center revenue hit $89 billion, up 117% from a year prior. Non-GAAP earnings per diluted share of $2.22 beat the $2.09 analyst consensus. For the third fiscal quarter, the company provided a Q3 guidance of $108 billion in revenue — what will be its first quarter above the $100 billion threshold, guided with no China data center compute revenue assumed.
Kress acknowledged what many investors were already watching in the gross margin line. “Many of you have expressed concerns regarding our gross margins, as component costs have risen significantly,” she said. “We are experiencing extreme pricing conditions in memory. The magnitude of the price increase has memory pricing exceeded prior expectations and are headed even higher into next year.” The consequence is a guided decline across three consecutive quarters: from 75% in Q2, down to approximately 74% in Q3, to a gross margin trough of 71-72% in Q4 of fiscal 2027, before a recovery to 72% to 73% as Nvidia’s own price increases on its products take effect in fiscal 2028.
Kress did not frame this as a cost shock without a compensating benefit. She argued that tighter memory supply as symptom of the same demand surge driving Nvidia’s own growth differs fundamentally from a pure cost hit: “Unlike a component that simply raises our cost with no offset benefit, tighter memory supply is a symptom of the same demand surge that is driving our own growth,” she said. CEO Jensen Huang was more direct: unconstrained demand would produce growth “a lot higher” than the 70% Nvidia is guiding. Nvidia shares rose approximately 6% in after-hours trading following the report.
Read more: AWS Adds 2 Million More NVIDIA GPUs: Prior 1 Million Commitment Ran Out Early
To understand why Nvidia cannot simply produce more output in response to higher demand, it helps to understand what makes high-bandwidth memory different from every other component in an AI server — and why that difference creates a ceiling that is structural, not temporary.
Standard server DRAM delivers data at roughly 50 to 100 gigabytes per second per memory module. HBM3E, the type used in Nvidia’s current Blackwell generation, HBM3E delivers 1.2 terabytes per second of bandwidth per stack by stacking multiple DRAM dies vertically and connecting them through microscopic copper pathways called through-silicon vias (TSVs), then bonding that stack directly alongside the GPU on a silicon interposer. The result is dramatically higher bandwidth — but at a manufacturing cost that compounds the supply constraint in two independent ways.
The first bottleneck is wafer area. Producing one gigabyte of HBM consumes roughly three to four times the wafer area per gigabyte of HBM as standard DRAM requires. Every fab wafer reallocated from commodity DDR5 production to HBM production therefore removes three to four times as many conventional memory bits from the supply pool as it adds in HBM bits. HBM already accounts for roughly HBM drives 30% DRAM revenue despite representing only about 8% of total memory output in bits — and that ratio is still widening as Vera Rubin’s demand ramps.
The second bottleneck is TSMC’s CoWoS (Chip-on-Wafer-on-Substrate) advanced packaging capacity. Producing HBM is not enough — each stack must be bonded to the GPU die on a silicon interposer using CoWoS, a specialized packaging technology that TSMC operates at a limited number of dedicated facilities. TSMC’s CoWoS lines are estimated to be running at TSMC CoWoS capacity at maximum utilization in 2026, with Nvidia consuming roughly 60% of available capacity for its data center accelerators. Even if Samsung and SK Hynix were to produce significantly more HBM4 dies tomorrow, TSMC’s CoWoS lines could not package and deliver them into AI servers at a faster rate. These two constraints — HBM die production and CoWoS packaging — compound each other, which is why supply remains structurally limited regardless of how aggressively the memory makers invest.
The Vera Rubin architecture intensifies this pressure further. Each Vera Rubin GPU carries 288 gigabytes of HBM4 — a 50% increase over the 192 gigabytes of HBM3E on the Blackwell B200. Vera Rubin Ultra, the follow-on platform, targets the Vera Rubin Ultra HBM4E target spec of up to one terabyte of HBM4E per GPU. With Vera Rubin already in full production and ramping across CoreWeave, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure, the demand signal into the memory supply chain is not merely strong — it is accelerating into each successive GPU generation that carries more memory per chip than the last.
Nvidia’s balance sheet shows an inventory balance of $32 billion at Q2’s close, as the company pre-positioned Vera Rubin components at scale — a figure that itself reflects the memory procurement required for the ramp. That pre-positioning itself tightens near-term availability for other buyers in the broader memory ecosystem.
The explicit 16-month timeline disclosed by Kress — fiscal 2028 growth capped at 70% by supply, with the constraint expected to persist through the fiscal year ending January 2028 — sent a clear pricing signal through the global semiconductor supply chain. The primary beneficiaries are Samsung Electronics and SK Hynix, which together with US-based Micron Technology control effectively all global DRAM production.
All three suppliers are now qualified and shipping HBM4 for Vera Rubin — a milestone that Huang confirmed all three HBM4 suppliers were shipping publicly in June 2026 at Seoul’s Gimpo airport. SK Hynix had an early lead, beginning mass-production shipments of 12-layer HBM4 at the end of June 2026, ahead of an originally planned September ramp, while Samsung — which had spent two years recovering from HBM3E qualification delays — achieved record HBM4 shipments at competitive speed specifications. The supply warnings from both companies’ leadership predate the Nvidia earnings call and now appear strikingly accurate: Samsung’s memory division warned in April 2026 that significant shortages were expected through at least 2027, and SK Hynix CEO Kwak Noh-jung has stated that 2027 will be the worst year for the shortage, with the shortage forecast beyond 2030 — demand is expected to outpace supply capacity even after that date, the CEO said.
The scale of what Nvidia’s confirmation means for the memory market is captured in Gartner’s most recent semiconductor forecast, published on August 24, 2026. Gartner projects DRAM revenue forecast to increase 246.6% in 2026, with total memory expected to see memory revenue to surpass $1 trillion by 2027 — surpassing non-memory chip revenue for the first time in the semiconductor industry’s history. Gartner also projects NAND flash revenue to grow 371.9% in 2026 as AI infrastructure build-outs consume enterprise SSD capacity for storage-intensive training and inference workloads. “AI infrastructure has fundamentally changed the dynamics of the memory market,” said Gartner Director Analyst Shrish Pant.
Nvidia’s supply commitments — the $279 billion figure representing locked-in procurement across the next several years — effectively give Samsung and SK Hynix exceptional pricing visibility. For a memory manufacturer, a rare, explicit, multi-quarter forward commitment from the world’s most important AI hardware buyer confirming that the demand signal will not soften before 2028 is the closest thing to guaranteed revenue the sector has ever seen.
Read more: SK Hynix Soars 13% on Nasdaq Debut: HBM Architecture Makes It Irreplaceable in AI Chips
The memory squeeze is not simply an abstract constraint on Nvidia’s growth rate. It is a concrete cost event for every organization buying AI infrastructure for 2027.
Bloomberg reported, and multiple outlets including Tom’s Hardware and The Next Web confirmed, that some of Nvidia’s largest customers and the server companies that build systems for major data center operators — including Microsoft, Google, and Oracle — have been notified that server prices rise more than 15% for Vera Rubin and Grace Blackwell systems shipping in early 2027 in many configurations. The exact magnitude of each increase depends on the chip generation and memory configuration, with systems carrying maximum memory — typically the configurations used for the largest AI models — facing the sharpest increases.
For organizations planning large-scale AI infrastructure procurement, the practical implication is straightforward: systems that fit 2026 budget models do not have 2027 delivery prices. A 15% increase on AI server systems that already run into the hundreds of thousands of dollars each translates into material additional CAPEX for any facility sized around current hardware economics. Enterprise buyers whose deployment windows extend into early 2027 should treat that price increase as a planning input, not a contingency.
Nvidia has been explicit that it intends to protect gross profit dollars by passing memory cost increases through to customers rather than absorbing them at the margin line. The company’s estimated revenue opportunity per gigawatt of capacity has expanded from roughly $18 billion during the Hopper generation to $25 billion with Blackwell, and is estimated at approximately $40 billion with Vera Rubin when accounting for GPUs, CPUs, networking, systems, and software. Each generation that expands the per-rack sale bundle gives Nvidia more pricing flexibility even as individual component costs rise.
Huang used the earnings call to frame the demand environment in terms that go beyond a single company’s outlook. In Huang’s AI inflection point quote from the earnings release, he said: “AI has reached its inflection point. This time last year, one lab alone was driving the buildout. Today, we have a golden age of new AI labs and startups, multiple frontier labs scaling in parallel, a thriving open-model ecosystem and physical AI coming online — with strong momentum across the U.S. and around the world.”
Global AI investment surged to where AI venture capital exceeded $400 billion in the first half of 2026 alone — surpassing the $265 billion raised in all of 2025 — with roughly 70% of that capital spent on compute, according to Kress’s disclosures on the earnings call. Nearly 20 companies now exceed $1 billion in annualized run rate revenue from AI products, up from 13 at the end of last year.
The counterintuitive dimension of Nvidia’s guidance is that a supply-constrained ceiling is a qualitatively different problem from a demand ceiling — and markets treated it as such. Nvidia shares rising 6% after the report reflected investor recognition that the company has a backlog problem, not a saturation problem. Supply constraints, however frustrating for buyers, are eventually resolved by capital investment in manufacturing capacity. Demand weakness is not.
Still, the constraint carries a real cost for the broader AI development ecosystem. Amazon Web Services announced that it AWS committed additional two million GPUs in 2027 and 2028 — a commitment timed alongside the earnings call that brings AWS’s total publicly disclosed GPU count to more than 3 million units over a three-year window, the largest such commitment by any single cloud provider. AWS is absorbing some of the largest absolute dollar increases in memory-inclusive server pricing in the industry’s history, and the economics of every frontier AI lab, neocloud, and enterprise AI factory built around Nvidia hardware in 2027 are being shaped by the same supply ceiling Kress named openly on the call.
For memory makers, Nvidia’s candor is exceptional forward visibility in an industry not known for it. For customers, it is a call to act on procurement decisions before early 2027 pricing takes full effect — or to plan for a higher cost basis that, on current evidence, will not resolve until at least the end of fiscal 2028.
The increase is a direct pass-through of soaring memory costs — specifically high-bandwidth memory (HBM4), the stacked chip that feeds data to Nvidia’s Vera Rubin and Grace Blackwell GPUs. HBM requires approximately three to four times the silicon wafer area to produce compared to standard DRAM, and TSMC’s advanced packaging lines, which are needed to attach HBM stacks to GPU dies, are operating at effective capacity. Those two independent constraints — HBM die production and CoWoS packaging capacity — are structural bottlenecks that cannot be resolved quickly even with significant capital investment. Nvidia CFO Colette Kress confirmed the severity of the squeeze directly, and Kress confirmed memory price expectations exceeded their prior forecasts, stating prices are “headed even higher into next year,” and the company is passing those costs to customers on systems shipping in early 2027.
Nvidia’s own supply-constrained guidance of approximately 70% growth for fiscal 2028 — the year ending January 2028 — implies at least 16 months of elevated constraint from the current date. SK Hynix’s CEO has publicly stated that 2027 will be the worst year for the shortage and that demand is expected to outpace supply capacity beyond 2030. The Gartner August 2026 semiconductor forecast projects supply-demand conditions remaining tight even as additional capacity comes online in 2027, because AI infrastructure deployments continue to consume more memory per system with each GPU generation. For enterprise buyers: budget for sustained elevated memory pricing through at least 2027 and into 2028.
CFO Colette Kress stated on the August 26 earnings call that customer forecasts “point to our growth doubling next year” — implying that unconstrained demand would produce roughly 140% revenue growth for fiscal 2028. Nvidia’s official guidance of approximately 70% growth is lower specifically because memory supply cannot support the full demand at current production rates. CEO Jensen Huang independently confirmed that unconstrained demand would produce growth “a lot higher” than 70%. This is an unusual and significant disclosure: Nvidia is explicitly telling the market that its growth ceiling is set by supply, not by customer willingness to pay — a distinction that has meaningful implications for how enterprise buyers should think about AI infrastructure availability through 2027, as Kress disclosed supply-constrained 70% forecast directly on the earnings call.
Enterprise organizations planning AI infrastructure procurement for 2027 should consider three actions before early 2027 shipment windows open: (1) Lock in purchase orders and contract pricing for Vera Rubin and Grace Blackwell systems now, at current prices, where contractual terms allow — the 15% increase applies to systems shipping in early 2027, not necessarily to contracts signed before that date; (2) Add a 15% or greater memory cost buffer to any 2027 AI infrastructure CAPEX model that has not yet been finalized; (3) Evaluate whether workloads support alternative architectures — including inference-specific ASICs such as OpenAI’s Jalapeño for high-volume serving traffic — that use less HBM per unit of inference work delivered, as that efficiency advantage compounds directly in a supply-constrained market. Organizations with flexibility on deployment timing may benefit from reassessing 2026 vs. 2027 procurement windows. Reporting on AI server price hike practical implications confirms these dynamics are already shaping procurement decisions across enterprise and hyperscaler buyers.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.

source