Intel Foundry to provide 60,000 high-margin wafers to the U.S. government a year. Really?

[hist78]

I happened to see this comment on Intel Foundry by Northland Capital Markets analyst Gus Richard.

“As a foundry, Intel will likely have the [Department of Defense] and [National Security Agency] as customers and will be able to sell roughly 60,000 high-margin wafers to the U.S. government a year,”

'The thrill is gone' on AMD stock, but Intel is looking better: Analyst

Shares of Advanced Micro Devices Inc. have been red-hot since the start of the pandemic, but one analyst worries that momentum could stall due to a possible...

www.marketwatch.com

TSMC makes about 530 units of Apple A14 processors on a 12-inch wafer. Let's assume a 400 units/wafer as the calculation base, 60,000 wafers a year will lead to 24,000,000 processors a year. How can DoD, DOE, NASA, etc. use so many chips that need to be sourced from Intel Foundry every year repeatedly?

Here is the consideration:

1. Most government semiconductors usage are riding on the public available commercial products.

2. For those sensitive national security related application, there are several foundries other than Intel, such as TSMC and Globalfoundries, already work with US government for many years. It's technically impossible and tactically not safe to kick them out in favor Intel Foundry as the sole source.

3. Some numbers related to national security related applications:

~ US makes about 156 F-35 Lightning II a year.

Source: https://www.airforcemag.com/f-35-production-set-156-per-year-until-completion/#:~:text=F-35 Production Set at,Until Completion - Air Force Magazine&text=More than 300,000 parts from,Texas, Lockheed Martin production facility.

~ US will build about 13 major Navy ships in 2022

Source: https://www.forbes.com/sites/davida...get-a-whole-lot-of-new-ships/?sh=5691257d1b38

~ As of December 2018 there are 320 known military or dual-use satellites in the sky, half of which are owned by the US, followed by Russia, China and India.

Source: https://en.wikipedia.org/wiki/Military_satellite

My Thoughts:

National security related application for semiconductors can be profitable. But it doesn't mean they can generate a killing amount of revenue and profit for a super expensive leading-edge foundry. On the other hand the big money are made by those prime contractors and system integrators such as Lockheed Martin, Northrop Grumman, and Boeing. To assume Intel can make tons of money from the new Intel Foundry division due to US government's orders is very questionable.

 

[Daniel Nenni]

My guess is that Intel will use mature nodes for US military business. Intel 14nm is an excellent process. With all of the improvements made over the years I would put Intel 14nm on par with TSMC 10nm but with much more volume maturity.

Given the technical challenges Intel is having with leading edge nodes and limited throughput I don't see them pimping that capacity out anytime soon.

 

[Paul2]

The only real thing there is Xilinx FPGAs in radars, about which there is apparently a giant drama.

Even if they have 20 these FPGAs in a jet they can easily buy millions of then, and stash them for a century+ of F35 production for a fraction of price of redesigning the die from 40nm UMC process to 5nm TSMC one.

People who make policy decisions about "critical lack" of domestic IC manufacturing capability are clearly disinformed, or worse, being disinformated.

 

[Mike Bruzzone]

Under Docket 9341 consent order if Intel gets an xx federal government wafer contract AMD cannot be denied one and gets a split or unique equivalent offer. AMD can decline but AMD cannot be denied. They have to be treated equally. Mike Bruzzone, Camp Marketing

 

[hist78]

The only real thing there is Xilinx FPGAs in radars, about which there is apparently a giant drama.

Even if they have 20 these FPGAs in a jet they can easily buy millions of then, and stash them for a century+ of F35 production for a fraction of price of redesigning the die from 40nm UMC process to 5nm TSMC one.

People who make policy decisions about "critical lack" of domestic IC manufacturing capability are clearly disinformed, or worse, being disinformated.

I think the policy teams at the US Federal level probably are thinking in the similar direction as you have suggested.

If the national security related semiconductor products are existing products and being used by the defense contractors in assembling the weapon or something daily, it's too late to switch to another foundry now. The cost, time, and uncertainty make no sense at all. It's much easier and cheaper to pay Samsung or TSMC (if they are the current foundries) to build additional fabs here in US than move to someone else.

Remember DoD and DOE need real products, not some theories or roadmaps. If they don't follow this strategy seriously, a F35 stealth fighter or a supercomputer for making hypersonic missiles will show up 10 years late.

After spending lots of money in supporting domestic IDM and fabs, DoD and DOE know very well why some defense contractors use non domestic foundries at the first place. I think Intel will get more taxpayers' money regardless. But bringing more capable and relevant foundries to US is more practical and has a higher priority in those policy teams' minds.

That's probably why Intel feels mistreated and keeps doing PR campaigns.