Intel is redirecting some manufacturing capacity intended for client processors towards data-centre chips, as server demand outstrips supply. The shift shows AI infrastructure pressure reaching beyond accelerators to the CPUs and packaging needed to build data-centre systems.
Intel AI Watch analysis
What happened
Intel is moving some Oregon resources from its Intel 18A process, intended for Panther Lake client processors, to Intel 3 production for data-centre chips. The additional capacity is expected to enter production in the second half of 2027. Intel is also expanding its Fab 24 complex in Ireland to produce more Xeons, although that capacity is not expected to come online until the end of 2027 unless the work can be accelerated.
The company is increasing advanced packaging capacity in New Mexico and Malaysia, and expects about 40% more substrate capacity to become available in the second half of 2027. Substrates are another constraint on producing its multi-chip processors. Tom’s Hardware reports that Intel expects supply constraints to remain particularly severe in the first half of 2027.
Why it matters
AI data centres need more than specialised accelerators. Intel expects demand for data-centre CPUs to remain strong into 2027 and 2028, including for inference and agentic workloads. Its reported response involves shifting existing capacity as well as expanding factories and packaging operations, but much of that new capacity will take time to arrive.
That leaves a near-term squeeze in view: Intel is trying to serve rising server demand while the capacity and materials needed to increase supply are still catching up. The company’s third-quarter Data Center and AI revenue was $7.3 billion, up nearly 80% year on year, according to the report.
Our read
The interesting part is the trade-off. Intel is prioritising server chips where demand is stronger, while moving some resources away from a client product line. It is a reminder that AI infrastructure can compete with ordinary PCs for manufacturing capacity, even when the chips involved are not the headline-grabbing accelerators.
For customers, the practical question is how quickly the extra capacity comes online. Factory plans make a useful signal, but they are not extra processors on shelves today.
What to watch
- Whether Intel’s Oregon, Ireland and packaging expansions meet their stated timelines.
- Whether the expected substrate capacity arrives in the second half of 2027.
- How Intel balances client and data-centre production as demand and supply change.
Discussion spark: When chip capacity is tight, should manufacturers prioritise AI data-centre demand over consumer processors, or protect supply for both?
Sources and evidence
Independent WittyWires tracker for public updates about Intel AI. Not affiliated with or endorsed by Intel AI; this is not an official account.