Intel CEO Chen Liwu Says Memory Makers Told Him Supply Tightness Won’t Ease Until 2028

Intel Chief Executive Officer Chen Liwu warned that shortages in memory chips within the computing industry are likely to persist for at least two more years. Speaking at a Cisco Systems conference on Tuesday, February 3, Chen reported that two major memory manufacturers informed him there would be no relief before 2028. He attributed the sustained demand pressure to the large‑scale expansion of artificial intelligence infrastructure, which has absorbed substantial memory capacity and reduced availability for conventional devices.

Chen emphasized that Nvidia, as a leading supplier of AI processors, will further elevate memory requirements with its Rubin platform and its next‑generation products, noting that artificial intelligence workloads “consume massive amounts of memory.” He also disclosed Intel’s intention to enter the GPU market and confirmed the company has recruited a chief GPU architect to lead that effort. Chen said the GPU initiative will be closely integrated with Intel’s data center chip organization and its foundry services. On the day of his remarks, Intel shares opened higher but later retreated.

Chen described the GPU hiring as a deliberate effort and welcomed the new architect to the team. Reports indicate that former Qualcomm executive Eric Demmers joined Intel last month and subsequently confirmed the move on LinkedIn. Chen said the GPU program is overseen by Kevork Kechichian, head of Intel’s data center chip business, and stressed that the project is being developed in close collaboration with customers to define their specific needs.

On Intel’s foundry business, Chen stated that several customers are engaging deeply with Intel’s wafer‑foundry operations. He previously indicated that customer interest centers on Intel’s 14A manufacturing node and that volume production could accelerate later in the year. Chen noted that customers must provide forecasts of product volumes and types so Intel can plan and allocate the necessary capacity.

Industry reports have noted that after Nvidia announced a USD 5 billion investment in Intel in September 2025, the two companies planned collaboration on the successor to the Rubin series, the Feynman‑architecture chips, with Intel responsible for advanced packaging for the GPU portion. Supply‑chain sources indicate that GPU compute cores will continue to be fabricated by TSMC, while I/O chips—which include memory controllers and handle inter‑chip connectivity—may be produced using Intel’s 18A process or the 14A process slated for mass production in 2028, depending on subsequent yield performance. Although I/O chips do not demand the same compute performance as GPU cores, they still require advanced process technology.