Micron has projected that the global memory chip shortage will worsen and persist until at least 2028, driven by high demand from artificial intelligence expansion and long-term supply commitments, forcing enterprise buyers to adapt strategic planning.
Micron has warned that the shortage of memory chips is likely to tighten further over the next two years, with demand already running so far ahead of available production that much of its future output is committed well in advance. The company says customer allocation discussions are now largely centred on 2028, underscoring how far ahead buyers must plan as the market remains constrained. Micron’s chief executive, Sanjay Mehrotra, said the firm does not yet have a clear point at which supply and demand will return to balance.
The pressure is being felt across both DRAM and NAND. Micron said DRAM pricing rose by a high-teens percentage in its fiscal fourth quarter, while NAND prices increased by about 30%. Market researchers at TrendForce expect further rises in the fourth quarter, and say some cloud operators and original equipment manufacturers are already responding by fitting lower-capacity memory modules into servers to manage costs. That reflects a broader scramble to secure supply rather than a short-term pricing spike.
A key driver is the build-out of artificial intelligence infrastructure. Reports from industry sources say high-bandwidth memory, or HBM, is especially scarce, because much of the industry’s manufacturing capacity is already allocated and new plants will take time to qualify and scale. Micron has also indicated that more than three-quarters of its fiscal 2027 supply is already booked, with multiyear agreements helping to lock in capacity long before chips are delivered. Even where new fabrication sites come on stream, the company does not expect them to ease conditions quickly.
For enterprise buyers, the implication is that upgrade plans will need to be more selective. Extending refresh cycles on routine desktops or file servers may help to defer spending, but delaying investment in core infrastructure or engineering systems can carry a heavier operational cost in lost productivity and reliability. Vendors are preparing for that environment: Lenovo says memory constraints are likely to persist at least until the end of 2027 and that it can pass through higher component costs if needed. Dell, HPE, HP, Cisco and Supermicro did not comment in the material reviewed, but the broader message from the sector is consistent: memory is becoming both more expensive and harder to secure.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
