When Nvidia filed its fiscal second-quarter 2027 results on August 26, disclosing record revenue of $96.2 billion, the clearest external confirmation of Micron Technology’s pricing leverage sat inside the same release. Chief Financial Officer Colette Kress stated that memory costs had exceeded Nvidia’s internal modeling, were still rising, and would compress gross margins from 75% in the second quarter to a trough of 71% to 72% by the fourth quarter of fiscal 2027 before a partial recovery in fiscal 2028 once Nvidia’s own price increases take hold. The data point that sharpens the Micron memory Nvidia premium pricing narrative is procurement: Nvidia’s supply commitments swelled to $279 billion, up from $119 billion the prior quarter, primarily to lock in memory at current rates.
That margin pressure does not vanish. It transfers. On a quarter guided to roughly $108 billion in revenue, each percentage point of gross margin represents about $1.1 billion, and only three companies on earth manufacture the high-bandwidth memory absorbing the cost: Samsung, SK Hynix, and Micron. Within two trading sessions of Nvidia’s report, all three suppliers gained between 3% and 5%, according to Motley Fool analyst Anders Bylund’s August 28 note. Yet Micron trades at approximately six times its fiscal 2027 consensus earnings, against a semiconductor industry forward P/E median of 29.18, an 80% gap.
Nvidia Earnings Reveal the Magnitude of Micron Memory Nvidia Premium Pricing
High-bandwidth memory now accounts for an estimated 30% to 40% of the total build cost of an AI accelerator, up from under 20% two generations ago. Every dollar Nvidia pays above its prior expectations for HBM is a dollar landing on the income statement of one of three suppliers. Kress’s framing of the situation is unusually direct. She confirmed that Nvidia has already negotiated price increases slated to take effect in the first quarter of fiscal 2028, meaning the margin compression is the interval between when Nvidia begins paying more for memory and when it begins charging more to its own customers.
Investors will focus on Nvidia’s November 17 report for confirmation that the trough holds at 71% to 72%. A stable range supports the timing interpretation. A deteriorating range would suggest memory manufacturers are capturing more than Nvidia budgeted, converting the fiscal 2028 recovery from a firm plan into a softer forecast. For Micron, Jensen Huang cleared the company alongside Samsung and SK Hynix in June 2026 to supply HBM4 for Nvidia’s Vera Rubin platform, locking Micron into the top tier of the AI memory roadmap and confirming technical parity with the two Korean incumbents that have historically dominated HBM market share.
Why the Valuation Discount Persists Despite HBM4 Qualification
Micron shares sit roughly 29% below their 52-week peak even as the company’s HBM capacity is fully booked for 2026. The analyst consensus price target stands at approximately $1,295 to $1,318, implying upside of 69% or more depending on the reference date. The discount reflects a specific and durable concern: the market is pricing in a repeat of the historical DRAM boom-bust pattern, in which memory manufacturers build capacity in parallel during an upturn, oversupply the market, and watch prices collapse. In the prior downturn, Micron’s gross margins fell from 45.2% in fiscal 2022 to approximately 2.7% in fiscal 2023. A forward P/E of six on projected earnings implies the market assigns meaningful probability that those projections will not materialize.
The bear case carries a recognizable name. Michael Burry, the investor who identified the 2008 housing crisis, has quietly assembled a short position against Micron, adding to it near $924 per share on the thesis that AI data-center oversupply could emerge by 2028. He has simultaneously placed shorts against Oracle and Nebius, framing the call as a broader AI infrastructure overcapacity argument rather than a company-specific one. Citi Research analyst Atif Malik, who maintains a Buy rating, trimmed his Micron price target from $1,400 to $1,150 in August 2026, citing potential deceleration in DRAM and NAND prices over the next four quarters. Goldman Sachs, by contrast, has revised its 2027 DRAM shortage forecast sharply upward to a 5.9% supply deficit and described the current market as the most severe memory shortage in 15 years.
What Separates This Cycle From Prior DRAM Busts
The bull case rests on three structural changes that the cyclical bear argument does not fully incorporate. First, the wafer economics of high-bandwidth memory differ from conventional DRAM. HBM stacks multiple DRAM dies vertically, connects them through microscopic copper channels called through-silicon vias, and co-packages the resulting stack directly beside a GPU on a silicon interposer. The manufacturing complexity raises the capital cost per gigabit and lengthens the qualification cycle, slowing the speed at which incremental supply reaches the market.
Second, customer concentration has shifted. The buyers of leading-edge HBM are a small set of AI accelerator designers, primarily Nvidia, AMD, and a handful of custom silicon programs at hyperscalers. That concentration gives memory suppliers visibility on multi-quarter demand that did not exist in the fragmented PC and smartphone DRAM market of prior cycles. Third, the packaging step that combines HBM with a logic die is now a binding constraint. CoWoS and similar interposer-based flows are gated by foundry capacity at Taiwan Semiconductor Manufacturing Company, meaning that even if DRAM wafer output expanded rapidly, finished HBM units could not ship without matched advanced packaging capacity.
Burry’s Bear Case Against the Micron Memory Nvidia Premium Pricing Thesis
Burry’s public filings point to a 2028 horizon, by which point he expects AI data-center buildouts to exceed end-demand for inference and training compute. If that view is correct, accelerator shipments slow, HBM orders contract, and the premium currently embedded in memory pricing collapses toward historical norms. The thesis gains credibility from the gross margin history. Micron’s fiscal 2023 trough of 2.7% remains a recent memory for institutional investors who carry cycle-aware position sizing into every upturn. Citi’s Malik reduction to $1,150 indicates that even bullish houses are hedging the timeline.
Goldman Sachs’s 5.9% deficit forecast cuts the other way, suggesting supply remains structurally tight into 2027 even if end-demand softens. The contradiction between the two research notes is itself the market’s central uncertainty. What Nvidia’s August 26 disclosure settled is that, at minimum, the next three quarters of memory pricing will run above prior expectations, and that all three qualified HBM4 suppliers, including Micron, will capture the spread. Whether that spread persists into fiscal 2029 is the open question that will determine whether six times forward earnings is a value trap or the entry point of the cycle for Micron memory Nvidia premium pricing exposure.
Source: Micron Still at Six Times Forward Earnings as Nvidia Pays Premium for Memory