Intel Delays 7nm Products, May Lean More On Foundries

The delay means Intel’s first 7nm client CPU won’t arrive until at least 2022 while its 7nm data center CPU won’t come until 2023, and it’s looking at turning to external chip foundries if more manufacturing issues persist. ‘The 7nm push isn’t a positive announcement as many products were dependent on it,’ an analyst says.

While Intel continues to report strong sales for its PC and data center business, the chipmaker‘s plan to return to a quicker cadence of new architectures is already starting to fall off track. The move may prompt the company to lean more on external chip foundries.

The Santa Clara, Calif.-based company said Thursday that its 7-nanometer products have been delayed by six months due to slower-than-expected progress with the company‘s next-generation manufacturing process, which is 12 months behind.

[Related: Why Google Cloud Built Confidential VMs With AMD EPYC]

Intel‘s stock price was down by as much as 10 percent in after-hours trading.

On the company‘s second-quarter earnings call , Intel CEO Bob Swan said the delay, which was caused by an issue with the 7nm process’ yield, means the company now expects its first 7nm product, a client CPU, to arrive in late 2022 or early 2023.

“We have identified a defect mode in our 7nm process that resulted in yield degradation. We’ve root-caused the issue and believe there are no fundamental roadblocks,” he said.

However, Intel is investing in contingency plans to “hedge against further schedule uncertainty,” according to Swan, by leveraging advances in design methodologies like die disaggregation and advanced packaging, which the company is using in its new Lakefield hybrid processors.

Swan said Intel‘s ability to compete won’t be affected by missteps in its process technology development, which the company will continue to invest in. However, he said, the company is also looking at using external chip foundries to a greater extent to stay on track.

“We will continue to invest in our future process technology roadmap, but we will be pragmatic and objective in deploying the process technology that delivers the most predictability and performance for our customers, whether that be on our process, external foundry process, or a combination of both,” Swan said. “Our advanced packaging technologies combined with our disaggregated architecture, gives us tremendous flexibility to use the process technology that best serves our customers.”

Intel is already using external and internal process technologies in tandem with its chip packaging technologies for the company‘s previously announced Ponte Vecchio data center GPU, which Swan said will now be released in late 2021 or early 2022. In other GPU news, Swan said the company “successfully powered on a petaflop-scale GPU with high bandwidth memory” using its advanced embedded multi-die interconnect bridge 2D packaging technology.

If Intel were to increasingly rely on foundries for future products, Swan said, there would be economic implications that would raise questions about how Intel can continue to deliver “attractive” average selling prices and keep them in line with its costs. Going in that direction would mean that Intel would spend more on 10nm production in 2022 and less on 7nm, he added.

“That‘s the kind of the optionality that we’ve tried to build in as we evaluate the future of Moore’s law,” he said in response to an analyst question.

Outside of process manufacturing improvements, Swan said the company intends to maintain an annual cadence of “significant product improvements,” which will include products targeted for holiday 2022. The company‘s first 7nm data center CPU, on the other hand, is expected in the first half of 2023.

While Intel reported delays for its 7nm products, the company said its transition to 10nm products this year is “accelerating,” citing increasing volumes of product and high demand, with Tiger Lake laptop CPUs launching soon and initial Ice Lake server CPU shipments due at the end of the year.

“We now expect to increase our 10nm-based product shipments for the year by more than 20 percent versus our January expectations,” Swan said.

Intel‘s second 10nm server processor lineup, Sapphire Rapids, is expected to start sampling with customers in the second half of this year and its 10nm Alder Lake desktop CPUs are targeted for a launch in the second half of 2021, according to new disclosures.

However, Intel‘s 10nm products, which debuted with Ice Lake laptop processors last year, are only arriving after several years of delays as the company relies on an advanced version of its 14nm process that debuted in 2014 for recent server and desktop CPU launches.

Swan has previously said that Intel wants to "recapture process leadership going forward" by returning to a two to two-and-a-half year cadence for its 7nm and 5nm manufacturing processes — closer to the timing Intel previously had for nodes before its multi-year 10nm delay.

Meanwhile, Intel‘s chief rival, AMD, is now onto its fourth lineup of 7nm processors with the new Ryzen 4000 desktop APUs, which are arriving roughly a year after the company debuted its Ryzen 3000 desktop CPUs and EPYC Rome server CPUs and a few months after its Ryzen 4000 laptop CPUs.

Patrick Moorhead, president and principal analyst at Moor Insights and Strategy, said the company has demonstrated financial resilience through its prolonged transition from 14nm to 10nm.

“The 7nm push isn‘t a positive announcement as many products were dependent on it. Knowing Intel, it always has backups for its backups and I am sure we will be hearing about enhancements to 10nm to increase its competitiveness,” he said in a message. “The company has done well financially, particularly well in the data center, notebook and commercial product lines, on 14nm /10nm when the rest of the world was on TSMC 10nm/7nm.”

For Intel‘s second quarter, the company reported a revenue of $19.7 billion, a 20 percent increase over the same period last year and $1.15 billion above Wall Street’s expectations.

The chief drivers of Intel‘s double-digit growth were the Data Center Group, which grew 43 percent to $7.1 billion on the strength of high demand for cloud providers, communications service providers and enterprises, and the Nonvolatile Solutions Group, which grew 76 percent to $1.7 billion.

The company‘s PC business, the Client Computing Group, also continued to fare well, growing 7 percent to $9.5 billion, thanks to continued demand of laptops driven by pandemic-induced work-from-home and remote learning dynamics. The Programmable Solutions Group grew 2 percent to $501 million.

That growth offset the IoT Group declining 32 percent to $670 million due to slowdown in pandemic-impacted verticals and Chinese companies blocked by the U.S. government buying products in the country. Mobileye revenue was down 27 percent to $146 million due to lower automotive production.

“Amid a very challenging environment, cloud and network infrastructure and PC capabilities have been vital in allowing businesses and people to continue to work, learn, stay connected, and provide critical goods and services,” Swan said.