Kuo: iPhone 18 Models Will Feature 2nm Chips

[MacRumors]

Apple supply chain analyst Ming-Chi Kuo today reiterated that the A20 chip in iPhone 18 models will be manufactured with TSMC's 2nm process.

Kuo said TSMC's trial production of 2nm chips has now reached yields well above the 60-70% range. The yield refers to the percentage of functional chips that can be obtained per silicon wafer, which is essentially a large, circular disc of chips.

Kuo first said that the A20 chip would be 2nm around six months ago, and another analyst Jeff Pu said likewise earlier this week.

An earlier rumor saying the A20 chip would remain 3nm was retracted.

This is ultimately good news, as the A20 chip being 2nm instead of 3nm means it will have more substantial performance and power efficiency improvements over the A19 chip for iPhone 17 models. Kuo, Pu, and others have all said that the A19 chip will be manufactured with TSMC's third-generation 3nm process, called N3P.

The move from a 3nm to 2nm process allows for more transistors in each chip, which helps to boost performance. Specifically, reports indicate that A20 chips should be up to 15% faster and up to 30% more power efficient than A19 chips.

An overview of current and expected chips:

• A17 Pro chip: 3nm (TSMC's first-generation 3nm process N3B)
• A18 and A18 Pro chips: 3nm (TSMC's second-generation 3nm process N3E)
• A19 and A19 Pro chips: 3nm (TSMC's third-generation 3nm process N3P)
• A20 and A20 Pro chips: 2nm (TSMC's first-generation 2nm process N2)

Keep in mind that these nanometer sizes are simply TSMC marketing terms, rather than actual measurements.

iPhone 18 models are still a year and a half away from launching.

Article Link: Kuo: iPhone 18 Models Will Feature 2nm Chips

 

[Justin Cymbal]

This is ultimately good news, as the A20 chip being 2nm instead of 3nm means it will have more substantial performance and power efficiency improvements over the A19 chip for iPhone 17 models

Definitely looking forward to these power efficiency improvements

 

[Rodney Williams]

I don't really care about all this talk about the potential performance of the A20 chip. Just fix the software of all these hiccups and bugs that seems to be plaguing the Apple products for past few years of iOS, macOS and iPadOS. Just fix the damn software. Period. What good is the chip if the software is garbage?

 

[MayaUser]

Of course, this will be an 100% thing

 

[Joe Rossignol]

Of course, this will be an 100% thing

Sometimes the obvious news still needs to be reported on, and then someone will make a comment saying it is obvious, and then life carries on.

 

[thefrost]

It will come out at some point surely! 🤣

 

[MayaUser]

Sometimes the obvious news still needs to be reported on, and then someone will make a comment saying it is obvious, and then life carries on.

Agree, but here, there were some rumours that this will not be the case and it will be on N3P/3nm...and that was strange since Apple usually lock their new TSMC architecture for an entire year

 

[vegetassj4]

Moars Law: the observation that the number of bugs in iOS is exponentially and inversely proportionate to the nm process of Apple's chips

 

[Joe Rossignol]

Agree, but here, there were some rumours that this will not be the case and it will be on N3P/3nm...and that was strange since Apple usually lock their new TSMC architecture for an entire year

Correct, because investment firm GF Securities put out a research note stating the following: "Apple’s A20 chip in 2026 will remain on the N3P node (leveraging CoW packaging to enhance AI capabilities) rather than migrating to N2, contrary to earlier rumors."

So we relayed the information accordingly.

However, I later spoke with GF Securities' lead Apple analyst Jeff Pu, who said to disregard this information about the A20 chip remaining N3P. That information was shared by another analyst at the firm, and Pu follows Apple more closely and confirmed that A20 will indeed be N2.

 

[azentropy]

Been an upgraded process quite a few times in a row now. I always wonder where the break even is with the cost involved in upgrading to the new process vs the cost savings itself of using the new process.

 

[Arm64MyMac]

“TSMC marketing terms, rather than actual measurements.“

If 2nm doesn’t mean 2nm and 3nm doesn’t mean 3nm, what’s the point of these terms? How can anyone then derive any power efficiency from this?

 

[rickdollar]

Not actual measurements???
"How much do I owe you?"
"One hundred dollars. But not really one hundred dollars."

 

[cdsapplefan]

That’s it, I’m skipping the 17 Pro and getting the 2nm 18 Pro. 📱 😉

 

[MacCurry]

Yield rates of 3nm wafers is at least 50%, and actual rejection rates are higher. At 2nm, both TSMC & Samsung Semiconductor are having issues for total production rates. Samsung Semiconductor has slowed down its development of their Taylor, TX facility mostly because of this. TSMC has Apple in the bag, and can afford a production rate of less than 20%. Others can't. I think we're at the end of the line now.

 

[Reason077]

reports indicate that A20 chips should be up to 15% faster and up to 30% more power efficient than A19 chips.

In other news, reports indicate that iOS 20 will contain at least 30% more battery-draining AI code compared to iOS 19.

 

[Howard2k]

“TSMC marketing terms, rather than actual measurements.“

If 2nm doesn’t mean 2nm and 3nm doesn’t mean 3nm, what’s the point of these terms? How can anyone then derive any power efficiency from this?

Because people understand the difference between 2 and 3 and can make a logical correlation. But saying that they went from FinFET to GAA is much more difficult for most people to understand and compare. There is a size reduction, but the measurements aren't exact, but it makes a much simpler point of reference and comparison.

 

[t0rqx]

Cancelled my iPhone 15 preorder 2 years ago.

Cancelled my iPhone 16 preorder last year.

Cancelling my iPhone 17 this year.

 

[Populus]

Maybe this is a similar case to what happened with the early N3B process with low yield and not that widespread implementation, and it is better to wait for the second gen 2nm process, likely the A21 and M7. We’ll see…

 

[Populus]

Yield rates of 3nm wafers is at least 50%, and actual rejection rates are higher. At 2nm, both TSMC & Samsung Semiconductor are having issues for total production rates. Samsung Semiconductor has slowed down its development of their Taylor, TX facility mostly because of this. TSMC has Apple in the bag, and can afford a production rate of less than 20%. Others can't. I think we're at the end of the line now.

Then, I guess the smartest move is to wait for the N3P chips, such as the M5, and stay there for the foreseeable future.

 

[BorisDG]

Kuo heard from me and now he changed his previous statement. 🤣🤣

Edit: Nvm was Jeff Pu.

 

[Mrkevinfinnerty]

Just in time for Apple Intelligence.

 

[jon9091]

Let me know when they hit 0nm chips.

 

[mlayer]

TSMC (and Apple as its #1 customer) are moving so fast that I wonder if the Arizona fabs will ever be leading edge like their counterparts in Taiwan. TSMC is still managing the ramp-up and hiring, which shows here in the PHX metro. Still, these are big leaps and even bigger investments to support them. Fab 21, the first fab in PHX, produces the 4 nm node.

 

[nt5672]

With the same buggy half designed half completed software. When will apple wake up.

 

[TechWhisperer]

I’m shocked. Who saw this coming.