Over the past several years, Intel (INTC +0.30%) has fallen behind Taiwan Semiconductor Manufacturing Company (TSM 0.60%) and Samsung in the "process race" to create smaller and more advanced chips.
Intel's R&D and manufacturing issues with the 14nm, 10nm, and 7nm nodes resulted in chip delays and shortages, enabling its rival Advanced Micro Devices (AMD 1.71%) -- which outsourced is chip production to TSMC's superior foundries -- to expand its market share in the PC and server CPU markets.
Last year, many analysts speculated that Intel would also need to go "fabless" and outsource its manufacturing to TSMC to catch up. But earlier this year Intel's new CEO Pat Gelsinger shot down those rumors and doubled down on the chipmaker's first-party foundries with fresh investments.
Image source: Getty Images.
At its "Intel Accelerated' event on July 26, Gelsinger unveiled a brand new technological roadmap and declared that Intel could catch up to TSMC and Samsung by 2024, then reclaim the process lead by 2025. Those are ambitious goals, but should TSMC -- which remains ahead of Samsung as the world's most advanced chipmaker -- be concerned at all?
How Intel plans to catch up to TSMC
Intel believes it can catch up to TSMC with three main strategies. First, it will boost its capex from $14.3 billion in 2019 to $19 billion to $20 billion this year, with most of that spending allocated toward expanding its domestic foundries and buying ASML's (ASML 0.90%) high-end EUV (extreme ultraviolet) machines, which are essential for manufacturing the world's smallest chips.
Second, Intel will rename its upcoming nodes to better reflect their transistor density and overall performance compared to TSMC's nodes. For example, Intel is renaming its 10+ node its "new" 7nm node, its old 7nm node its "new" 4nm node, and its old 5nm node as the "new" 2nm node. Intel can make this shift because measurements of node sizes aren't defined by set industry standards.
Lastly, Intel could consider buying smaller foundries to accelerate its expansion. It's already reportedly mulling a takeover of AMD's former chipmaking unit GlobalFoundries, the world's fourth-largest chip foundry, which could significantly expand its third-party foundry services.
How TSMC plans to stay ahead of Intel
This is how Intel's revised roadmap measures up to TSMC's plans.
|
Year |
TSMC |
Intel |
|---|---|---|
|
2021 |
5nm (mass production started in 2020), 3nm (test production) |
7nm (will be launched in late 2021) |
|
2022 |
4nm, 3nm (mass production for both nodes) |
4nm (first products will launch in early 2023) |
|
2023 |
2nm (test production) |
3nm |
|
2024 |
Unknown |
2nm (20A) |
|
2025 |
Unknown |
1.8nm (18A) |
Source: TSMC and Intel, industry websites.
Intel's chips are denser than TSMC's, so it routinely claims its 10nm chips are comparable to TSMC's 7nm chips, its 7nm chips are comparable to TSMC's 5nm chips, and so forth.
However, TSMC claims its 3nm (N3) chips -- which have been in development since 2019 -- will be up to 70% denser than its 5nm chips, while consuming up to 30% less power and running up to 15% faster. Even Intel, in an odd twist, will reportedly task TSMC with manufacturing its upcoming 3nm CPUs in 2023 before it launches its own 3nm CPUs.
TSMC will also likely outspend Intel to maintain its lead. It intends to boost its capex from $17.2 billion in 2020 to roughly $30 billion this year, then collectively spend approximately $100 billion over the next three years. It also recently teased a potential breakthrough in the development of 1nm chips.
Intel might gain some support from government subsidies in the U.S. and Europe, but TSMC has also secured U.S. government subsidies for its new plant in Arizona -- a decision Gelsinger loudly protested. In an Intel-sponsored article on Politico, Gelsinger pointed out that TSMC would still produce its most advanced chips in Taiwan instead of the U.S., and that "foreign chipmakers vying for U.S. subsidies will keep their valuable intellectual property on their own shores."
Taiwan's own government will also likely subsidize TSMC's expansion to keep it ahead of Intel, since it needs TSMC to serve as the country's most valuable bargaining chip in its relationship with the U.S., as well as its most effective deterrent to a Chinese invasion.
Why TSMC shouldn't be too concerned ... yet
Intel is making bold promises about challenging TSMC and helping the U.S. become a leading manufacturer of chips again. But it probably can't outspend TSMC without big government subsidies -- and some of those subsidies are already flowing to TSMC as part of the escalating tech war between the U.S. and China. TSMC is also ASML's largest customer, so Intel could struggle to secure enough high-end EUV machines to achieve its lofty expansion goals.
Intel's rocky track record of delays and shortages will also likely prevent TSMC's top customers -- which include Apple (AAPL 3.40%), AMD, and Qualcomm (QCOM 2.22%) -- from switching chip manufacturers. Therefore, TSMC's investors shouldn't be worried about Intel's ambitious plans yet -- but they should still keep an eye on its progress over the next few years.
