Summary
Intel is still one of the biggest players in the CPU (and now GPU) segment, but it has struggled to keep up with the new times. As it turns out, it’s struggling with yields for its newest technologies and silicon—though if you’ve followed Intel for a while, you know this is not exactly surprising news.
Intel’s next-generation production process, known as 18A, is struggling with low-quality yields, according to sources briefed on the company’s internal data. This year’s Panther Lake chips are supposed to be based on 18A, which, if Intel successfully mass-produces and launches to store shelves, would mean Intel would finally get ahead of TSMC, which has yet to mass-produce any chips with its upcoming 2nm process. As of late last year, the yield—the percentage of usable chips from a silicon wafer—for Panther Lake was reportedly as low as 5%. While that figure has improved to around 10% by this summer, it remains drastically below the levels required for profitable mass production. Historically, Intel has aimed for yields above 50% before ramping up production to avoid damaging profit margins, with true profitability for a complex chip not being achieved until yields reach the 70% to 80% range.
The Panther Lake chips produced using 18A reportedly have a defect density approximately three times higher than what is considered acceptable for initiating high-volume manufacturing. Without a dramatic improvement by the chip’s planned launch, Intel may be forced to sell its new chips at a significantly reduced profit margin or even at a loss. Intel executives have pushed back against the grim assessment. In a July interview, Chief Financial Officer David Zinsner disputed the specific yield figures, stating that “yields are better than that,” though he declined to provide a precise number.
As of the reasons as to why this is happening, it is apparently having to do with Intel focusing so much on regaining the performance edge that it used multiple unproven technologies simultaneously. This includes a next-gen transistor design as well as a new method for delivering power to the chip. The company has publicly warned that it could exit leading-edge manufacturing altogether if its subsequent process, 14A, fails to secure external foundry customers.
This is not exactly new for Intel. The company has struggled to keep up for years with companies like TSMC when it comes to making its silicon smaller, faster, and more efficient. As an older example, Intel’s desktop 11th gen chips back in 2021 continued to use the company’s 14nm process node, even as competitors such as AMD were already using 10nm and were in the process of moving to 7nm. It has sought to keep up and surpass competitors over these past years, and if it manages to do so this year, it will likely be after a great struggle for the company—and perhaps at a cost, too.
