Intel plans to establish its Foundry business as an independent subsidiary as it aims to cut costs and optimize its business.
The move, outlined in a letter from Intel CEO Pat Gelsinger, came after a day of news aimed at calming fears about the x86 chip giant's future.
Intel also announced an expanded partnership with Amazon Web Services for custom chip designs. Intel Foundry will produce an AI fabric chip for AWS on the Intel 18A architecture as well as a custom Xeon 6 chip for AWS. The two companies will work on other chip designs going forward.
That AWS partnership comes as Intel was awarded $3 billion in direct funding under the CHIPS and Science Act for the U.S. government's Secure Enclave program.
The takeaway from the AWS and CHIPS and Science Act award is that Intel Foundry has a pipeline. For Intel, the issue with Intel Foundry is that manufacturing requires a lot of cash. Intel is trying cut about $10 billion in costs amid competition from AMD and Nvidia.
- Intel's Q2 wipeout: Guidance cut, 15% layoffs, dividend suspended
- Intel Foundry had $7 billion operating loss in 2023
- Intel launches Gaudi 3 accelerator with availability in Q2
- Intel's AI everywhere strategy rides on AI PCs, edge, Xeon CPUs for model training, Gaudi3 in 2024
Gelsinger said that Intel is "more than halfway" through the process of cutting 15,000 jobs by the end of the year. Employees will be notified in mid-October. Intel is also cutting about two-thirds of its real estate footprint by the end of the year.
By making Intel Foundry a subsidiary, Intel stopped short of a complete spin-off, but noted that the structure "provides our external foundry customers and suppliers with clearer separation and independence from the rest of Intel," said Gelsinger. Intel Foundry will also be able to find independent sources of funding and capital structure.
As for Intel the chip design and product company, Gelsinger said the plan is to maximize its x86 architecture for AI across PCs, edge devices and data center. The focus will be on AI inference workloads.
