Intel plans data center GPUs, hires Qualcomm veteran

Intel CEO Lip-Bu Tan announced plans to develop GPUs targeting data center markets, where Nvidia currently leads.

He confirmed the hiring of Qualcomm’s Eric Demmers as chief GPU architect, reporting to Intel’s data center chip chief.

Intel aims to work with customers to define their needs before ramping up volume manufacturing of its 14A process technology later this year.

Tan also said several customers are engaging with Intel Foundry for chip manufacturing capacity.

He noted that Huawei has hired about 100 top designers despite US restrictions on certain chip design tools.

This suggests China could advance in chip design without Western software, as Intel pushes its broader strategy to compete in the data center chip market.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

Intel’s 14A plan is a high-stakes bet on uncommitted customers

Intel CEO said the company is “going big time into 14A.” That reverses earlier remarks that 14A could be paused or dropped without a large outside customer 1.
Intel said two possible outside customers are testing 14A chips. It expects firm supplier choices to start in the second half of 2026, then run into the first half of 2027 2.
Intel is holding off on 14A production capacity for foundry clients. It plans CapEx (capital expenditures) only after firm outside commitments 2.
Reports say Huawei has built an electronic design automation (EDA) platform that can design chips down to 14nm. The sources do not confirm how the tooling compares with leading Western suites 3.

The chip war is becoming a software and ecosystem battle

Intel’s data center GPU push is led by newly hired chief GPU architect Eric Demmers. It enters a market where software tuning or ecosystem pull can matter as much as hardware, with US restrictions shaping the backdrop 4.
Huawei is using open-source software such as Flex:ai to pool and orchestrate processors (coordinates how multiple chips share work). Huawei says this raises AI chip utilisation by 30% on average 4.
China is also building self-sufficient chip design tools. These efforts could form a parallel technology stack (a full set of chips, software, and developer tools that work together) that depends less on Western hardware or tooling 4.
Over time, the semiconductor market could split into separate camps. Competition may hinge on smoother developer workflows, better efficiency, or wider access.