AI is buying up memory faster than it can be produced: Micron’s forecast

The global memory shortage may not only drag on for several years, but could also worsen due to the rapid expansion of AI data centres. Micron’s Chief Executive, Sanjay Mehrotra, stated that More than 75% of the company’s planned production volumes for 2027 have already been contracted, whilst negotiations with major clients are increasingly focusing on deliveries in 2028 and even much later.

The memory shortage is not going to end any time soon

According to Sanjaya Mehrotra, memory manufacturers are currently struggling to keep pace with the growth in demand by expanding their production capacity. The situation regarding future contracts is particularly telling. More than three-quarters of Micron’s production capacity for 2027 has already been sold or reserved, whilst an increasing number of negotiations with major clients are already focusing on 2028.

Some customers are seeking supply guarantees even for the period beyond 2030. This is unusual for the memory industry, where the market has historically experienced sharp cycles of shortages and overproduction.

The main reason is the boom in artificial intelligence

The development of generative AI has become the main factor putting pressure on the market. Companies are building huge data centres with thousands of AI accelerators, and each such system requires significant amounts of RAM and storage. There is particularly high demand for high-speed memory for AI accelerators. Manufacturers are being forced to divert an increasing amount of production capacity towards the highest-margin server solutions. As a result, traditional segments – personal computers, smartphones, servers and SSDs – are also feeling the pressure.

Micron will increase production, but this may not be enough

Micron forecasts that over the next two years its shipments DRAM and NAND are set to rise by approximately 20%. DRAM is used primarily as random-access memory, whilst NAND forms the basis of SSDs, flash drives and other permanent data storage systems.

At first glance, the 20% increase appears significant. However, Micron expects demand to grow even faster. That is why even the commissioning of additional production capacity does not guarantee that the market will return to equilibrium.

New factories will not solve the problem straight away

The problem is that semiconductor production cannot be scaled up in a matter of months. It can take years from the start of construction of a factory to mass production. In addition to the production buildings themselves, expensive equipment is required, as well as the fine-tuning of manufacturing processes and a gradual increase in the yield of usable chips. Therefore, even plants that are already planned will not be able to instantly bridge the current gap between supply and demand.

What does this mean for shoppers?

A prolonged shortage does not necessarily mean that RAM or SSDs will suddenly disappear from the shops. However, the situation creates conditions under which manufacturers have far less incentive to lower prices.

If demand from the AI industry continues to grow at the current rate, the limited supply of DRAM and NAND could keep prices high for components used in servers, computers and other devices. Of particular importance will be how manufacturers allocate available production capacity between profitable AI solutions and the mass consumer market.

There is no sign of a balance yet

Mehrotra has effectively admitted that Micron is currently unable to specify when production will finally catch up with demand. And this is probably the most important signal for the market. We are no longer talking about a short-term shortage of specific types of chips. Major customers are trying to secure memory supplies for years to come, whilst manufacturers are simultaneously investing billions in new factories.

If the current The AI boom will continue, 2027 and even 2028 may be characterised by a very tight memory market.