Apple closed its fiscal year holding $14.5 billion of stock, two and a half times what it held a year earlier, and added $32.7 billion of purchase commitments in thirteen weeks [1]. Memory contract prices are forecast to rise about 20% this quarter because suppliers are steering capacity to AI servers [3]. The usual cost of holding stock is 20% to 30% a year, or 5% to 7.5% a quarter [7]. Put the two side by side and the stock already on the floor is the cheaper memory.

At a glance

  • The shift: Memory now reprices faster each quarter than stock costs to hold.
  • The number: 20% a quarter against 5% to 7.5%.
  • The risk: The hedge lasts about one quarter; the commitment outlives it.

The ladder ran out of room

Dynamic random-access memory (DRAM) is the chip that holds data while a machine runs. Its contract price is the quarterly rate large buyers negotiate, not a retail sticker [3]. That price rose about 80% in the second quarter, 40% in the third, and is forecast to rise about 20% in the fourth [3]. Those three rises compound to roughly three times the starting price, and I keep coming back to that, because it decides whether the hedge is worth carrying. The slowdown is not the relief it looks like.

Quarter DRAM contract price One quarter of holding cost
Second 2026 +80% 5% to 7.5%
Third 2026 +40% 5% to 7.5%
Fourth 2026 (forecast) +20% 5% to 7.5%

Suppliers are steering advanced capacity toward server memory and enterprise solid-state drives, where AI inference demand pays better [3]. Long-term agreements are beginning to cap how far prices can climb [3]. A buyer of ordinary components faces a market that is short and repricing upward.

Apple answered in the same quarter three ways. It leaned on inventory bought in the June quarter, raised prices by $100 on the iPhone 17 and on some MacBooks, and lifted its purchase commitments sharply [5]. Product gross margin, the share of sales left after the direct cost of the goods, finished the fiscal year at 39.3% against 36.8% a year earlier [1]. The first of those moves is the one an operations team can copy without renegotiating anything.

The rest of manufacturing is buying forward too. The Institute for Supply Management (ISM) puts customers’ inventories at 41.6, in “too low” territory, for a 25th consecutive month [4]. No industry reported customers’ inventories as too high in October [4]. Its prices index has also increased for 25 months, and memory has been listed as a material in short supply for ten [4].

By the numbers

  • $14.5 billion — stock at the year end: Two and a half times the $5.7 billion on the books a year earlier [1].
  • $89.7 billion — manufacturing purchase obligations: Committed component and factory purchases, mostly due within twelve months, equal to 123 days of direct costs [1].
  • $10.4 billion — the component share: Parts held for production rather than finished goods, against $2.1 billion a year earlier [1].

When the hedge stops paying

The hedge is already losing value. Apple guided gross margin, the share of sales left after the direct cost of goods, to 45% to 46% for the December quarter, down from 47.9% [5]. The chief financial officer expects memory costs to keep rising into next year [5]. Stock bought at the old price runs out, and the margin guide is the first sign of that.

What I am watching now is whether the obligations line keeps outrunning the inventory line. A buyer reading only the stock line sees a modest build. Reading both, Apple has committed close to four months of production to a market it expects to keep repricing [1].

Three lines answer the open question next quarter. First, whether obligations keep outrunning stock. Second, the customers’ inventories index, stuck at 41.6 for a year; a turn up would mean forward buying has overshot the shelves [4]. Third, the long-term agreements that TrendForce says are capping price rises [3]. A ceiling on the price rise is a ceiling on what the hedge earns.

The catch

The obligations are not money spent. They are commitments, payable mostly within twelve months, and Apple accrues cancellation fee reserves for orders it cannot use [1]. A demand shock turns them into either a penalty or inventory bought at the top of the market.

The arithmetic leans on a convention. Clear Spider reckons stock costs 20% to 30% of its value to hold for a year [7]. SourceDay gives the same range for what it calls holding costs [8]. Neither figure is audited, and each bundles money, space, insurance and obsolescence into one number. I find the comparison directionally right and the precise multiple soft.

Two limits should travel with the decision. The 20% forecast applies to memory, while the holding cost applies to everything in the warehouse. That makes this a spread on one input, not a reason to raise stock across the catalogue. And the hedge pays only if the units sell. Stock bought against a rising price is still stock, and a margin protected by carrying inventory is protected once. Apple’s year is not permission to build.

Related reading

Sources