Bottom line: Intel executives like to tell a good story about the company’s ability to innovate and do it fast enough to take on the competition, but it’s hard to take it seriously when actual products are always on the horizon. However, the chip giant says it is making significant progress on advanced process technologies that will leave the lab in 2024 at the earliest.
Industry observers are focusing their eyes on Intel’s 3nm-based technologies, which are expected to debut in consumer CPUs next year. In the meantime, the company seems to be hard at work on more advanced process nodes like the Intel 18A and 20A that are central to its future as a semiconductor giant.
According to a Taiwanese news outlet, Intel’s Foundry Services division has completed the tape-out process for the two manufacturing technologies. The publication mentions the senior vice president and chairman of Intel China, who explained that the first chips based on an Intel 18A process can be made in a trial run sometime in the second half of 2024. However, the mass production of commercial products based on it are not There are no plans to start until 2025.
Both Intel 20A and Intel 18A are based on something called gate-all-around FET transistors (GAAFET), which is a common theme for all foundries that develop process nodes where the transistor gate pitch is smaller than 3 nm. Intel’s version of this is called RibbonFET and represents a major change in design since the introduction of FinFET in 2011.
Another advantage of Intel’s new technology is the backside power delivery (called PowerVia). At least in theory, this should allow for higher logic densities, higher clock speeds, and lower power leakage – leading to more energy-efficient designs that are expected to outperform of those made by companies such as Samsung Foundry or TSMC.
Also read: How did TSMC get so good?
It’s a big bet that will help Intel’s Foundry Services division secure big chip-making contracts in the coming years while making its products more competitive with Arm and AMD-based designs. It is also a risky and expensive transition because it requires adding more steps to the manufacturing process and using more materials and equipment compared to previous nodes.
Time, however, was not on Intel’s side. The company is trying to ensure that it will be the first to use the bleeding edge ASML Twinscan EXE scanner with 0.55 NA optics for the Intel 18A node, but this will cause delays that will not reach it. As a result, the company chose to rely on existing EUV machines to bring the process to market faster.
Whether this strategy will work out as planned is anyone’s guess, but suffice it to say that analysts are not as optimistic as Intel CEO Pat Gelsinger. The general sentiment is that the company is in a bad place, and not only because of low consumer demand, the rise of homegrown silicon, or the constant reduction of costs throughout the organization. Intel’s Foundry Services ambitions are a long-term dream that is not expected to materialize before the end of this decade.