Chips & Compute

Micron's AI Memory Is Nearly Sold Out Through 2027 — And That Is the Risk

Micron's Q4 revenue hit $54.2 billion with 87% gross margins as AI memory sold out through 2027 — yet Nvidia reportedly asked TSMC to explore 3D DRAM production.

Micron’s AI Memory Is Nearly Sold Out Through 2027 — That Could Be Its Biggest Risk
Micron’s AI Memory Is Nearly Sold Out Through 2027 — That Could Be Its Biggest RiskAI-generated
By Marcus Bennett4 min read

Why it matters

  • Micron's fiscal Q4 revenue reached $54.2 billion, nearly five times the $11.3 billion a year earlier, with 87% non-GAAP gross margin and $33.42 EPS.
  • More than 75% of Micron's fiscal 2027 shipments are committed under customer agreements, mostly at higher prices than 2026.
  • A November 6 report from The Information said Nvidia asked TSMC to explore advanced 3D DRAM production.
  • Micron invested $27.4 billion in capital expenditures in fiscal 2026; its Idaho fab starts wafer production no earlier than mid-2027.
  • Jefferies analysts suggested meaningful memory competition for Micron might not arrive until 2035.

Micron Technology's fiscal fourth-quarter revenue reached $54.2 billion, nearly five times the $11.3 billion the company generated a year earlier, and more than 75% of its fiscal 2027 memory shipments are already committed under customer agreements. The numbers, reported by 24/7 Wall St. on November 16, describe the most profitable stretch in the memory-chip maker's recent history. They also describe, according to the same analysis, the conditions for its next downturn.

Non-GAAP gross margins climbed to 87% in the quarter, with earnings of $33.42 per share. For the current quarter, Micron management expects approximately $61.5 billion in revenue and adjusted earnings of $38.15 per share. The driver is high-bandwidth memory (HBM), the stacked-memory technology that sits alongside AI processors and feeds them data. Without sufficient memory, even the most advanced graphics processing units cannot operate efficiently.

Why does sold-out production threaten Micron?

The structural story matters here. For decades, memory manufacturers cycled through oversupply, collapsing prices, and shrinking profits. AI data centers have inverted that dynamic: they consume memory faster than manufacturers can produce it, pushing prices higher and handing suppliers something they rarely enjoyed — the ability to dictate terms to customers.

Micron, listed on NASDAQ under the ticker MU, is among the biggest beneficiaries. Producing HBM also consumes manufacturing capacity that would otherwise supply conventional DRAM (dynamic random-access memory, the standard working memory of servers and PCs). Micron says HBM requires roughly three times as much wafer capacity per bit as ordinary memory, which helps explain why conventional DRAM prices are now approaching those of specialized AI memory. Most of Micron's calendar 2027 HBM supply is covered by contracts at higher prices than in 2026.

For investors, that constitutes extraordinary visibility into future revenue. It also constitutes the catch.

What did Nvidia reportedly ask TSMC to build?

The same shortages driving Micron's profitability are creating headaches for its largest customers. Nvidia (NASDAQ:NVDA) needs enormous quantities of memory for its AI accelerators, and dependence on a handful of suppliers gives those manufacturers pricing power while potentially constraining Nvidia's growth.

A November 6 report from The Information sharpened that tension. According to the publication, Nvidia asked Taiwan Semiconductor Manufacturing (NYSE:TSM) to explore producing advanced 3D DRAM, a technology that stacks memory cells vertically to increase capacity.

The foundry already manufactures Nvidia's AI processors and handles advanced chip packaging. Entering memory production would extend its influence into another critical link in the AI supply chain. The financial incentive is straightforward: Micron's 87% adjusted gross margin compares with Taiwan Semiconductor's approximately 66% gross margin in its latest reported quarter. As TSMC considers expanding beyond logic chips, Nvidia could eventually gain another supplier, while Micron would face a competitor with enormous manufacturing resources.

How soon could TSMC actually compete?

Two qualifications temper that scenario, and both come from the source analysis. TSMC has emphasized its continued focus on logic chips and has not announced a DRAM manufacturing program. Jefferies analysts suggested meaningful competition might not arrive until 2035.

The report's significance lies elsewhere. When customers become desperate enough to encourage new competitors, today's pricing power can eventually undermine itself. Nvidia's reported interest reveals the underlying problem rather than an imminent threat.

Has the memory cycle really disappeared?

Micron has seen the pattern before. Strong demand encourages manufacturers to expand production. Additional capacity eventually catches up with demand, prices retreat, and profits follow. The difference today is that AI infrastructure requires increasingly sophisticated memory, which makes new capacity expensive and technically difficult to build.

The capital numbers illustrate the scale of the commitment. Micron invested $27.4 billion in capital expenditures during fiscal 2026. Its Idaho fabrication plant is not expected to begin producing wafers until mid-2027, with meaningful output arriving later.

Those construction schedules cut both ways. They protect incumbent manufacturers — Micron, Samsung, and SK Hynix (NASDAQ:SKHY) — from rapid new entry. They also force companies to make multibillion-dollar investment decisions years before knowing what demand will look like.

What does the market already believe?

Micron's unusually low forward earnings multiple, as noted in the source analysis, suggests investors already question whether current profits are sustainable. The immediate threat, according to 24/7 Wall St., is not Taiwan Semiconductor entering memory production in the near term. It is the possibility that Micron, its established rivals, and eventually new entrants all expand capacity to capture profits that cannot remain this high indefinitely.

The bottom line from the analysis: Micron's memory shortage remains a powerful earnings catalyst, and its multiyear customer agreements protect it from an immediate downturn. But sold-out production should not be confused with permanent pricing power. Micron remains attractive for investors comfortable with semiconductor volatility, particularly given its earnings growth and low valuation. The warning sign to watch is whether capacity additions begin outpacing AI memory demand.

Nvidia, in the source's framing, helped make Micron's business more profitable than ever. It may also be helping create the competition that eventually brings those profits back to earth — whether that competition arrives in 2035, as Jefferies projects, or sooner.

Source: 247wallst.com

Share this article:

More from Marcus Bennett

Marcus Bennett

Show full bio

Senior reporter covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.

718 articles

Related articles

  1. Micron's 87% Margins Meet the Memory Cycle Again
  2. Micron's 87% Gross Margin Towers Over Intel's 44% in AI Memory Squeeze
  3. Micron's $14.63 Billion China Revenue Meets New U.S. Export Rules
  4. Micron Falls 7.62% as Nvidia Leadership Shock Hits HBM Orders
  5. Micron’s 6% Rebound Tests a $35.16 Billion Buyback Claim

« Previous articleNext article »