of estimates and bias in the error

It is widely accepted as a given among financial analysts that Apple makes a practice of lowballing its revenue and earnings estimates for the near future, and those estimates are routinely treated as a worst case baseline.

Apple has apparently also begun using best case estimates of sales in calculating what to tell component suppliers about how many units of their products it may be purchasing, and then waiting for the prices to come down before actually making those purchases. Taken together, these behaviors have been characterized as bullying.

You may recall that just a couple of years ago, Apple was implicated in an effort to monopolize the market for flash memory, via extended contracts for nearly all of the production capacity of the main suppliers, and that at about that same time the U.S. DOJ began an investigation into price fixing on the part of those suppliers. That investigation ended without any indictments, but it seems clear that the struggle over the price and availability of high density, non-volatile memory chips continues.

Gadget manufacturers (not just Apple) want the latest, highest density chips at commodity prices, but will substitute a larger number of lower density components in their designs if the total cost of production ends up being lower. Chip foundries need higher prices for the latest, highest density chips, to cover the costs involved in development and upgrading their production lines, and they need something above production costs generally to cover the costs of building new factories to meet the seemingly insatiable demand, which can go soft at any time with a downturn in the economy or the advent of some new technology that renders their own products obsolete, so amortization schedules must be relatively short.

The question is what is a reasonable profit and during what phase of a product's life-cycle should the foundries be able to expect to turn a profit. If they attempt to take the bulk of it early, they end up pressuring their customers to keep using older, lower density components. If they postpone profits until later in the life-cycle, they run the risk that their customers will move on to the next new thing before they've completely covered their development and production upgrade costs. Undoubtedly, it ends up being a balancing act.

Apple's recent approach, if understood for what it is, merely complicates the calculation. Gadget manufacturers are always competing to provide a better perceived value to their customers, while leaving some room for a profit of their own. One of the ways they go about this is to attempt to get a better deal than their competition on components used by all. That they are out to get the best deal they can while securing the availability of the components they need shouldn't surprise anyone, least of all the foundries.

Supplying estimates of future component purchases based on best case sales is just a way of improving the chances that the components will be available, if needed, should that best case turn out to be reality. That subsequent orders typically don't total to those estimates shouldn't surprise anyone.

It's something of a given in any business that a change in the per-unit price as a function of an increment in volume is always either negative or zero, never positive; maybe that needs to change. Maybe, in this situation, per-unit pricing needs to follow a model where it decreases from small lots to moderate volume, and then begins to increase again above moderate volume, combining the volume from all customers, with rebates based on the same formula if some customers end up canceling their orders because the price went too high, vaguely like the market for electrical power works, but operating over periods on the order of a week rather than fractions of an hour. This would provide gadget manufacturers with an incentive to spread their orders out evenly over the entire life-cycle of any component they made use of, and provide some systematic moderation on the substitution of newer, higher density components.

Such a system may require a disinterested, third-party broker.