Seven Videos, One Thesis: Micron's Memory Shortage Is Writing the 2027 Price

I watched seven videos and walked through seven different doors into the same room: one computes the 2027 price, one reads the all-time-high chart, one drops a Micron note inside an Nvidia news roundup, one maps the entire memory stack, one distills a single signal from five earnings reports, one explains the physics of the chip, and one views the stage through jobs and rates. The shared sentence: the memory shortage stretches beyond 2027, and Micron stands in the middle of it. My primary source is the 2027 prediction video; the other six serve as verification, context, and commentary around it.

Parkev Tatevosian's prediction starts from Wall Street consensus: Micron's earnings per share were $1.23 in the fiscal year ending August 2024; expectations stand at $73 for the current fiscal year (August 2026), $155 for 2027, and $171 for 2028. The video also notes the company cleared an 80% operating margin last quarter, the highest among the companies he follows. This ladder is the ground everything rests on; if the ladder breaks, every price above it breaks too.

The price math is built on multiple scenarios: at $171 of 2028 EPS with a 6x forward price-earnings multiple, the target is $1,002, roughly flat from today. At 7x it becomes $1,197 (about 20% upside); at 9x, $1,539 (about 50% upside). If the multiple compresses to 4x, the price falls below $684, a drop of more than 30%. The video's base case is the 7-9x range, a $1,200-1,500 band with a $1,350 midpoint implying 34% upside from the current price.

Two strong stories sit at the ends of the forecast. On the upside, the hyperscalers (Nvidia, Meta, Microsoft, Alphabet, Amazon) are each said to spend at least $10 billion extra this fiscal year because of higher memory prices, while signing long-term agreements; if Micron sheds the cyclical label and earns an S&P 500-average multiple, even $2,000-3,000 gets mentioned, though the video itself calls that unlikely. On the downside sits the fear of a capital-expenditure cycle peak: should customers one day say they are cutting spend and protecting the balance sheet, the price could fall even below $684.

The load-bearing pillar of the prediction is 16 long-term customer agreements. Per the video, they stretch to 2030 and carry sales visibility into the 2030s; the company is already signing the replacement for today's memory four years out, and the replacement of that eight years out. If this mechanism works, investors will price Micron as a predictable company rather than a cyclical one, triggering a rerating. The web side confirms the frame: across 16 strategic agreements, 14 carry minimum-price commitments worth about $100 billion, generally multi-year take-or-pay structures covering 20% of DRAM volume and roughly a third of NAND volume, plus about $22 billion in deposits and financial commitments.

The Sacco Financial camp builds the same thesis off the chart: the stock sat near $1,030 at recording time against an old high of $1,255, leaving 222 dollars, or 21.5%, of road ahead, with the first wall at $1,040. The video adds the honest caveat: if these are the final months of another memory boom, $1,255 gets much harder to justify. So the high thesis is conditional, and the condition is a lasting shortage. The date that tests it is set: the September 30 report.

The quarterly math behind the high path runs as follows: Micron earned $25.11 per share on an adjusted basis last quarter; the coming quarter is guided at $31, with revenue guided from $41.5 billion to $50 billion, an $8.5 billion single-quarter jump and roughly 23% earnings growth. $31 a quarter annualizes to $124, and the $1,030 price is 8.3 times that pace. Reaching $1,255 without any multiple expansion would require a $151 annual pace; but there is an easier road: $35 a quarter ($140 annualized) at a 9x multiple equals $1,260. So 13% earnings growth plus modest multiple expansion draws a concrete path to the old high, and the company guides to an 86% gross margin on top.

The physics behind the math is the video's strongest section: the memory next to AI accelerators (HBM) consumes more manufacturing capacity per bit than conventional DRAM. Every shift toward HBM takes capacity from somewhere, and a fab cannot be ordered today with chips coming out in three months; the shortage persists a while even as billions are spent. The video also recalls memory history: shortage, price surge, fortunes, capacity race, eventual glut. Two exhibits are offered against the cycle repeating: the company's still-tight supply outlook and customers committing future spend today.

The web-side core numbers verify the videos. Per the SEC filing, Micron posted fiscal Q3 revenue of $41.46 billion (up 345.7% year over year, 17.6% above the $35.25 billion consensus), adjusted EPS of $25.11 (23.8% above the $20.28 consensus, a seventh straight beat), and 84.6% GAAP gross margin. Q4 guidance is the highest in company history: $50 billion plus or minus $1 billion of revenue, $31 plus or minus $1 of EPS, and roughly 86% margin. Data-center revenue passed $25 billion (over a $100 billion annualized run rate), and data-center SSD revenue more than doubled sequentially past $5 billion. Fiscal 2026 capex runs about $27 billion with $10 billion in Q4 alone; capital returns are slated to increase after December 9, 2026, the second anniversary of the CHIPS agreements.

FinVid's news roundup seats Micron inside the Nvidia context: nearly 10% one-month momentum after Nvidia's report, $2 from the old high; on one side the MAG7's return to the stage plus September seasonality, on the other Thinking Machines' reported $5-6 billion round at a $40 billion-plus valuation. The roundup's boldest line belongs to Jensen Huang: the suggestion of Nvidia as a $3 trillion-revenue company in the near future and $3-4 trillion of annual AI infrastructure spend by decade's end. Its bubble debate is worth logging too: GPUs sitting dark for lack of demand are likened to dark fiber in the dotcom bubble, with one difference, today's demand is plugged in.

The roundup's Micron segment carries concrete items: plans to add up to 60,000 wafers a month of HBM capacity by year-end (consistent with the press-reported 100,000 target and last year's 40-50,000 band; the company trails SK Hynix and Samsung clearly in HBM capacity share), most of it shifting to 12-high HBM4 ramping in Nvidia's Vera Rubin, CHIPS-related buyback restrictions starting to ease in December, and Micron's own September 30 report. The roundup also notes Korea-U.S. semiconductor investment talks via Reuters. These are video-specific claims; they fit the thesis but sit on my independent verification list.

The ZipTrader video widens the frame into a full map of the memory stack: $100 turning into $818 in a year, a top-pick call at $363 back in January, DRAM contract prices up in the low-60s percent in a single quarter, HBM sold out through 2027 at all three makers, and roughly $150 billion of fab spending this year that only comes online in late 2027 into 2028. It sketches a company that closed last quarter at $41.5 billion revenue (up 346%) near 85% margin and guides $50 billion for the quarter reporting September 30, and it hands HBM leadership to SK Hynix at 50-55% share; that share matches Counterpoint's measured 50% for SK Hynix in Q2 exactly.

The map's NAND leg has two names: SanDisk and Kioxia. Per the video, SanDisk closed last fiscal year at $20.2 billion revenue, up 175%; at its August investor day it guided double-digit growth through 2030 at an 80% gross margin, with all of 2026 production sold out on multi-year purchase agreements. Japan's Kioxia is up 380% in Tokyo; the two share factories in Japan, so the NAND shortage sits under one roof too. This explains why Micron's NAND leg, a third of contracted volume, matters.

The equipment leg exposes HBM's hidden cost: vertical stacking demands roughly three times the equipment per gigabyte of memory, because every layer is etched, deposited, and inspected separately. The video's lineup: Lam Research (fiscal revenue $23.2 billion, up 26%; EPS up 39%; a $8.1 billion September-quarter guide about $1 billion above the Street; plus a $10 billion upward revision to the industry's annual spending forecast), Applied Materials (HBM-specific revenue from near zero to $1.5 billion in a year, targeting $3 billion), KLA (net income up 163% on 18% revenue growth, mix shifting to high-margin HBM work), and Teradyne on the test side. In the video's phrase, every HBM stack pays a tax to the test and packaging companies. The same video lists $5, $6, and $29 small caps as untalked-about beneficiaries; it devotes its second half to a non-memory small health-care name, which stays outside this article.

The supply-chain read claims five different companies (Credo, Dell, Broadcom, Snowflake, HP) told investors the same story within 48 hours: AI now demands the fast movement of enormous data, not just compute. The video's Credo figures (a $535 million guide band, near-115% growth) check out on the web: the company reported a record $479 million quarter (up 114.7% yearly) on September 1 with a $525-535 million Q2 guide, and its OmniConnect line plays directly into the inference memory bottleneck. For Broadcom the video cites about $16.7 billion of semiconductor revenue and roughly $34.8 billion of next-quarter total revenue guidance (versus a $35 billion Street bar, which is how the video explains the after-hours pressure), for Snowflake about $1.5 billion of product and $1.55 billion of total revenue with $9 billion of remaining obligations, and for HP multi-year supply agreements plus a message that tight conditions persist.

The same video's conceptual frame is the shift to the inference era: as companies run AI applications continuously, memory needs diverge from training; more compute and networking means more memory bandwidth. The striking estimate it relays is that high-performance memory could represent about 75% of an AI server's hardware bill. The precision is debatable, but the direction matches what Dell and HP describe: both speak the language of memory-driven challenges and agreements.

The Semicon Taiwan act carries a shared message from the big three makers (Samsung, SK Hynix, Micron): classic 2.5D HBM architectures with interposers may be nearing performance and power-efficiency limits. Per the video, Micron is exploring tightly coupled DRAM architectures targeting more than 10 times HBM bandwidth in certain future designs at lower energy per bit; the video adds its own caveat that this is an architectural quest, not a product claim. Samsung moving the same way with its ZHBM concept shows the search is industry-wide. The goal has a name: breaking the memory wall.

The Denver narration descends into the chip's physics: picture HBM as a skyscraper of memory floors, DRAM dies stacked with through-silicon connections and a base logic die, seated next to the accelerator in one package. The engineering exam is mechanical: packaging tolerance, heat dissipation, layer yield. Per the video, Micron's filings openly list a 3D-stacked HBM portfolio, with HBM3E qualification completed on high-performance accelerator platforms. The conclusion is as crisp as the investment thesis: Micron's HBM share is indexed directly to its own packaging yield; if yield slips, server makers face hardware shortages, because substitute conventional modules mean surrendering bandwidth.

The share picture sharpens on the web side and puts the videos' scattered figures in place. Per Counterpoint's Q2 measurement, DRAM shares read Samsung 38%, SK Hynix 25%, Micron 24%, CXMT 10%; in HBM, SK Hynix leads at 50%, Samsung jumped from 21% to 33% in a quarter, and Micron slipped from 21% to 18%. Per TrendForce, Q2 industry DRAM revenue rose 59.5% to $154.73 billion; Micron's conventional DRAM revenue grew 65.5% to $36 billion for a 23.3% share. Conventional contract-price growth cools to 13-18% in Q3; the September bulletin revised the PC DRAM increase upward while noting HBM talks are stalled on unfinalized specs with prices still pointed up.

Micron's hand looks strong on margins and HBM4. Operating margins across the three makers: Micron 80.4%, SK Hynix 76.3%, Samsung's device-solutions division 70.0%. Micron is the only supplier to disclose hard HBM4 revenue, over $1 billion shipped, with a 12-high ramp running twice as fast as HBM3E and HBM4E due in volume production in 2027. Industry estimates see HBM4 12-high reaching toward 50% of Micron's HBM output by year-end, and the company has guided the press to a 100,000-wafers-a-month HBM capacity target. On qualification, expectations of Samsung more than tripling HBM4 sales in Q3 plus SK Hynix's mass-production lead keep this a three-way race.

The seventh video watches from the macro desk and puts Micron against the rates backdrop: August payrolls of 162,000, a 4.37% two-year yield, a 5.24% thirty-year; the market priced roughly 60% odds of a September hike. That day chip stocks (SOX up 3.52%) split hard from software (down 2.23%), with Micron up 6.10% on the hardware side. The video's key distinction: payroll counts and spending decisions are different things; Lululemon's 4% drop is a reminder that demand can cool while headlines run hot.

This macro frame cuts both ways for Micron. Hot payrolls plus stubborn long-end rates pressure growth multiples; the financial press already prices that with the 10-year at 4.83% and oil above $90. But Micron's forward multiple around 6-8x already carries a deep cyclical discount: an $975.26 close on September 11, a $1,513 average analyst target, 44 buy ratings against 4 holds and zero sells. The macro wind is harsh but priced in; what would break the thesis is not rates but the end of the shortage.

Every road leads to the September 30 report, and all videos converge there: the quarter's enormity is already known, and what matters is the next guide. The independent three-item checklist reads: the fiscal 2027 capex plan, whether customer deposits arrive without weakening terms, and the contract-price outlook. Those three are the living part of the thesis; the revenue and profit prints are the dead part, already-priced history.

Closing the picture into one frame: a $684 bear case (4x), $1,002 neutral (6x, flat from here), a $1,200-1,500 base (7-9x, midpoint 34% up), an ATH path at $1,255-1,260 ($35 a quarter at 9x), and a $1,513 analyst average. The common denominator condenses to three items: a shortage running past 2027, visibility locked by 16 agreements, and the HBM4 ramp. My verdict: supply physics, not price, carries this thesis; while the physics holds, targets are debatable and the thesis lives.