Key Takeaways
Since Intel announced plans to expand its semiconductor foundry business in 2021, the firm has plowed tens of billions of dollars into developing new manufacturing processes and facilities. Dubbed Integrated Device Manufacturing (IDM) 2.0, the initiative has been a significant drag on Intel’s bottom line. But there are signs that the strategy is starting to pay off.
In a major win for the chipmaker, Intel and Amazon Web Services (AWS) have signed a multi-year, multi-billion-dollar agreement that will see Intel produce custom AI chips for AWS data centers.
As part of the latest agreement, Intel will produce an AI fabric chip for AWS using its most advanced 18A process node. It will also build on an existing contract for 3nm node chips, designing a new Xeon 6 processor specifically for Amazon data centers.
While AWS deploys a range of different processors in its data centers, its in-house offering is increasingly capable. It has helped the firm reduce its reliance on third-party suppliers in recent years.
Amazon is Intel’s second major client for the new 18A process, which forms a key component of its long-term foundry strategy. Earlier this year, Microsoft CEO Satya Nadella revealed that the company is using Intel’s manufacturing capacity to install new chips in its data centers next year.
With Microsoft and AWS now signed on as customers, Intel’s 18A process is making the inroads IDM 2.0 is all about.
With a focus on high transistor density and improved power efficiency, the 18A process positions Intel to compete with the 5nm and 4nm processes used by Samsung and the Taiwan Semiconductor Manufacturing Company (TSMC).
As Intel ramps up production of the new semiconductors, other chip makers will be paying close attention. For fabless manufacturers like Nvidia and AMD, the 18A process could offer a compelling alternative to existing solutions.
However, Samsung and TSMC aren’t resting on their laurels either. Both foundries have started manufacturing more advanced 3nm chips and expect to launch the first 2nm chips next year.
As they scale production and competition in the space heats up, Intel will need to pick up more large clients if it is to achieve its goal of usurping Samsung as the second-largest chip foundry in the world by 2030,
While the AWS deal is a major win for Intel’s foundry business, the unit continues to weigh on the financial performance of its parent company.
According to CNBC, it cost the chipmaker roughly $25 billion in each of the past two years. According to Intel’s own estimates , the unit is projected to continue losing money until around 2030, with losses expected to peak this year.
Efforts to absorb the costs through restructuring have only been partially successful, and the stock market has punished Intel for the massive expenditure incurred by IDM2.0. In a year in which other chipmakers have seen their market valuation soar, Intel stock has declined more than 55%.
Responding to market forces, CEO Pat Gelsinger informed employees that Intel plans to spin its foundry unit off as an independent subsidiary.
He said the subsidiary model will provide customers and suppliers with “clearer separation and independence” from the rest of Intel. Meanwhile, it will allow the company to “evaluate independent sources of funding” that aren’t possible under the current structure.
“It’s time to shift from a period of accelerated investment to a more normalized cadence of node development and a more flexible and efficient capital plan,” he emphasized.