Intel’s upcoming 18A process has reportedly failed tests with chipmaker Broadcom.
Citing three sources with knowledge of the matter, Reuters reported that after receiving silicon wafers back from Intel that had been through its advanced 18A manufacturing process, Broadcom was unsatisfied with the results and has concluded the process is not yet viable for high-volume production.
In early August, Intel announced that its first two 18A products were out of the fab, had powered on, and were on track to start production in 2025.
Intel 18A is one of the final nodes from the company’s '5 Nodes in 4 Years' roadmap and is the first to include the company’s PowerVia power delivery system, RibbonFET gate-all-around transistor architecture, and leverage Intel's open system foundry model.
The company released its 18A Process Design Kit (PDK) 1.0, allowing electronic design automation (EDA) and intellectual property (IP) partners to update their tools and design flows, and enable customers to begin their final production designs.
Reuters said it has not been able to establish whether Broadcom had walked away from a manufacturing deal with Intel as a result.
In a statement to the news outlet, Broadcom said it is "evaluating the product and service offerings of Intel Foundry and have not concluded that evaluation,” while Intel declined to comment on “specific customer conversations."
Intel shelves 20A process node for consumer processors
Elsewhere this week, Intel has announced it will no longer be using its own 20A process node for its upcoming Arrow Lake consumer processors.
In a blog post announcing the decision, Ben Sell, VP of technology development at Intel, wrote that due to the company’s 'sooner-than-anticipated success' with its more advanced 18A node, the decision has been made to “shift engineering resources from Intel 20A earlier than expected.”
He added: “The Arrow Lake processor family will be built primarily using external partners and packaged by Intel Foundry.” Although it has not been formally announced, it’s been speculated that the external partner referenced is likely to be TSMC.
Referencing comments made by Intel’s CFO Dave Zinsser at Citi Global TMT Conference, Tom’s Hardware reported that the decision to skip over the 20A node will save the company half a billion dollars.
Intel is currently undertaking a $10 billion cost-cutting plan, resulting from the $1.6bn net loss Intel posted for Q2 2024. The plan involves eliminating more than 15,000 jobs from its global workforce, in addition to reducing R&D and marketing spend by billions between now and 2026, cutting capex by 20 percent, and reducing the non-variable cost of goods sold by roughly $1 billion in 2025.
Sell did not reference the report from Reuters in his blog post and said Intel remains on track to launch 18A in 2025.
