Skip to main content

The AI boom runs on shortages, and one forecast says the test comes in 2029

Seen from a walkway above, a technician in a full white cleanroom suit and mask loads an amber wafer carrier onto the tray of a tall processing tool. Below and beyond, a semiconductor fab floor runs away into the distance, lined with machines and an overhead track carrying more amber carriers, lit in blue and gold.
01

Micron charged about sixty percent more per bit while shipments barely moved.

02

Oracle's 2026 backlog of $638 billion converts slower than it spends and borrows.

03

The forecast puts peak risk at 2028-2029, because shortages clear slower than demand.

SiliconANGLE · Breaking Analysis by Dave Vellante · Aug 2026

Morning Byte · weekly digest

One email a week. The stories that mattered.

What changed

SiliconANGLE analyst Dave Vellante published a forecast in August for when the AI spending boom breaks, and his anchor is Micron's latest quarter. The data show Micron's price per bit rose in the low-60s percent range last quarter, up from a low-single-digit gain in shipments, which means buyers paid far more for barely more product.

An AI data center is a chain of gates. A chip is useless without memory, memory is useless until it is packaged, and none of it earns a dollar until the building has power. Whichever link is scarcest sets the pace and holds prices up, because buyers have nowhere else to shop.

Price per bit+low 60s%
Bits shipped+low single digits
What grew at Micron last quarter: price, not product.Source: Micron fiscal Q3 2026 via SiliconANGLE
Free cash flow, fiscal 2025+$26BFree cash flow, fiscal 2026−$24B
Oracle's cash swing while it builds ahead of demand.Source: Oracle fiscal 2026 results via SiliconANGLE

Why it matters

The risk is not that AI stops working; the risk is that the shortages end before the money shows up. Oracle is the clearest case. Its filings show $638 billion of booked contracts as of fiscal 2026, more than any peer, which means the promise dwarfs the payments. The company expects about 12 percent of that backlog to turn into revenue within the next year, a slower pace than the headline suggests, which means it borrows now and collects later. Its free cash flow swung to roughly negative $24 billion in fiscal 2026 from about positive $26 billion the year before, which means it now borrows to build what customers promised to rent.

For a working reader, that lands in three places. Memory scarcity is part of why phones, laptops and RAM upgrades cost more this year than last, and chip or data-center holdings in a retirement account are priced partly on the shortage lasting. Taken together, the pattern suggests scarcity is doing most of the work in reported revenue, not demand.

What to watch

The signal is public and simple. Healthy looks like memory prices easing while shipments climb, because real demand replaced the premium. Trouble looks like prices falling and shipments stalling together.

Vellante says rotating shortages could stretch the boom through 2027 or later, with 2028-2029 the highest-risk window rather than a sure date, which means the test sits years out. If power delays mount, the break could slip into the next decade; until shortages actually clear, nobody gets to test real demand. Whether energized capacity converts into paid work fast enough is not yet known.

Not signed in yet — hit Post and we'll finish it together

The AI boom runs on shortages, and one forecast says the test comes in 2029 | Morning Byte