How memory has become a spot market

Due to the AI demand chip manufacturers have introduced ‘post-settlement pricing’

Major chip manufacturers are boosting prices and shifting contracts to significantly shorter terms. In doing so, they are forcing even large hardware OEMs such as HP and Cisco to pass some of the some of the additional risk on to their partners and customers.

Contract validity periods are decreasing sometimes to two weeks or less, which is way shorter than the decision-making process in companies and public sector bodies can accommodate.

Meanwhile, retailers and service providers are having to decide how much of the price rises and risk they can realistically pass on to customers.

Criticism is being directed toward the OEMs for chasing profits in the memory shortage triggered by the AI boom. However, the OEMs are in a similar predicament to their reseller partners and have very limited choice or scope for manoeuvre. They too are part of a supply chain that originates with the chip manufacturers.

Shortages and price explosions as the new normal in the memory market

In recent weeks, the world's memory manufacturers, led by Samsung, SK Hynix and Micron, have turned the rules of the market upside down.

The first step was to partially decouple production volumes from overall demand and align them with the margins, which are highest in the datacentre sector.

As a result, datacentres now receive more high-speed memory (HBM) and server DRAM chips to feed the AI demand at the expense of areas including memory for consumer devices from smartphones to PCs, where prices have risen even more rapidly.

This drives up prices in all areas and would significantly increase the margins of chip manufacturers and other participants in the distribution chain - were it not for existing long-term contracts with OEMs. Until now, these contracts have usually run for one year giving all parties time to plan. Price adjustments were typically a maximum of plus or minus 10%.

From a buyer's market to a seller's market

Chip manufacturers are now removing these barriers and radically shifting their sales strategy to a short-term model in which contracts only run for a quarter or even a month. And that's not all. They are also reducing their validity periods to almost zero through "post-settlement pricing", i.e. price determination after the contract has been signed.

This allows chip manufacturers to benefit from fluctuations while passing on the risk, which they previously bore themselves, to hardware manufacturers, partners and customers.

Price stability and predictability can no longer be assumed. Even major customers such as Apple are now only being offered short contracts and must therefore expect significant price increases at short notice. Given the availability situation, they have little option other than to swallow the bitter pill.

This situation is forcing vendors such as HP and Cisco to adapt their contracts with partners and customers to fit with these new circumstances. A buyer's market, in which demand determined oversupply, has become a seller's market, in which scarce supply is the determining factor.

There is no end in sight. The AI boom is expected to continue according to most forecasts, while new production capacities and more efficient manufacturing technologies will only be able to ease the situation in one or two years. Even then, it is uncertain whether there will ever be a complete return to the old model with longer contract and price cycles.

While chip manufacturers have no problem finding buyers, distribution partners, especially at the lower levels, have to fight hard not to lose orders and customers. One of the biggest challenges is to calculate their own risk and incorporate this into their pricing, inventory and contract models.

This article was first published in German on Computing’s sister site CRN Deutschland