With the unveiling of Intel’s “18A” (18 angstrom) chip designs, some feel the industry is changing rapidly. Others previously frustrated by the slow pace of domestic development say it’s about time.
The big picture: for years, the Taiwanese firm TSMC has been the go-to vendor for all AI chips. It’s the world’s foundry, and despite massive investments stateside, no one has been able to catch up.
Now, analysts are suddenly eyeing Intel as a potential challenger. There’s a lot online about investing – columnists asking whether it’s better to drop funds in Intel stock, or TSMC. That has to do with how quickly Intel can make itself into a pure play alternative.
The Volume Game
Research into actual production metrics shows that Intel produced between 3-5 million chips in the last year, against TSMC numbers of around 17 million. The year before, Intel’s yield was about the same. Estimates for 2027 are around 4-6 million chips. So it will be a while before Intel has the volume to emerge as a real competitor.
Meanwhile, industry people point out that demand is spiraling up quicker than TSMC’s capacity to max fabs. So there will probably be a lot of room for two big chip vendors.
18A Technology: GAA Design
Experts point out that one of the feathers in Intel’s cap with the 18A chip is called GAA or gate all around transistors. This type of build, putting gate structure all of the way around the transistor, is meant to improve performance while reducing power consumption.
TSMC uses GAA, too, notably, in its N2 design, but it’s late to the party, having stayed with prior gating engineering, believing that GAA was not yet necessary.
Backside Power
Another big technology built into the 18A is called PowerVia, which is described as a backside power delivery system.
It sounds like something a million grandpas might lampoon, with the suggestion of a “swift kick to the backside,” but in reality, in chip application, it means that the power structure is located on the back of the chip, leaving the front free for other stuff.
Here’s a bit of reporting from Intel’s own newsroom, albeit an article without a named author:
“Intel’s backside power solution is called PowerVia, and two new papers to be published at the 2023 VLSI Symposium show that Intel devised a process to manufacture it, test it and demonstrate positive performance results. … For the first time, chipmaking is going two-sided. Here’s how it works: Transistors are built first, as before, with the interconnect layers added next. Now the fun part: flip over the wafer and ‘polish everything off’ … to expose the bottom layer to which the wires for power will be connected.”
Word is that TSMC is putting its own similar system called SPR into N2 as well.
The Vendor Side
In thinking about all of this, it’s never lost on me that a Taiwanese man, Jensen Huang by name, moved to the U.S., specifically to Kentucky and Ohio, and started the world’s other big chip company, Nvidia, in a Denny’s restaurant.
I feel like there’s something significant about that.
And to be sure, while TSMC dominates foundry, Nvidia dominates the engineered chip side, providing some of the previous members of FAANG with the processing power they need to run their own high-tech operations. But now, there’s news that Apple, for one, will use Intel, and that has driven Intel stock up a lot. In fact, if you look at stock value year over year, the equity stayed around $20 per share until fall, and $50 until winter of last year, and really only spiked in March. It’s now up over $120, meaning something like a 500% gain in a few months. So it’s understandable that swing traders are chasing the dragon. But on another level, it’s what Intel’s growth symbolizes that is important: two big foundry firms for the global market, instead of one.
Stay tuned.