Does IFS or IF has a single HVM External Customer Yet?
- Thread starter fansink
- Start date
Definitely there are some orders like Mediatek to use Intel 16. The question is how big of the volume? For the reference, IFS 2023Q4 revenue is 0.29 billion, in contrast, tsmc 2023Q4 revenue is 19.6 billion. Intel 3 does not win a big customer and Intel 18A seems the same from the CFO's interview with SeekingAlpha.
Intel CFO: "We probably won't win anybody's major volume in 18A. We'll win some smaller SKUs, and that's all we need, to be honest with you. That will be very significant to us, even though it seems maybe marginal in the marketplace, particularly if we can collect enough of these customers."
fansink
Active member
To clarify, does IFS or IF have any external customers at any node, at substantial “delivered” volumes?
Not pending orders with contingencies, like the $15B Microsoft ~Order.
MKWVentures
Moderator
@Maximus
Intel revenue is 300M per quarter including packaging and sales of masks and equipment. external wafer revenue is less than half this (Intel has disclosed this). So current revenue and shipments is trivial
Intel does not have 15B in orders from Microsoft. they claim to have lifetime value of ALL deals = 15B. 90% of those deals are plans and not committed take or pay contracts. We have details what Intel external revenue will be over the next 5+ years. Side note: Microsoft agreed to do one chip with Intel sometime in the future.... no details. Lifetime value is a term used for new customers on new businesses. it is not really meaningful as it is not a binding contract
Intel did receive some prepayment on foundry .... it is believed it is well under 1B.
the CFO stated that Intel will not be a primary provider for any company in the next couple years. They will try out Intel and see how it goes. They may be a backup plan
Foundry is a slow growth business and it will be for Intel if they execute well. We can track fab build outs and delays to understand more how this will work.
I do expect the UMC, Tower, and DoD/government work to be major volume in 2-5 years. Major is about 1 fab total (Say 50k wafers per rmonth)
We have some more details by process and by Fab over time
Let's look at the process Intel Foundry has to go through: First Intel has to prepare foundry quality PDKs which is very difficult given that TSMC sets the PDK quality standard after 30+ years experience. Then comes the IP and EDA dance to build an ecosystem. Next IFS has to win the customer's business (test chips etc...). Customers have to design and tape-out. Finally it gets to manufacturing, packaging, test etc... Then system level products are prototyped, built and go to market. We are talking years here for HVM, right?
In my opinion Intel 18A will be the "trust test" node for the big semiconductor companies, who by nature hope for the best but plan for the worst to reduce risk. If Intel delivers quality 18A wafers that meet product specs (no binning like Intel internal products) 14A will be the real HVM node, if delivered on time. Unfortunately TSMC and Samsung are not standing still so this is a huge competitive challenge.
All-in-all though I do believe Intel Foundry will be successful due to the need for leading edge multi source foundry offerings. The question in my mind is how successful? Will IFS dominate the "Not TSMC Business"? Yes, definitely. Will IFS win big competitive business from TSMC? Certainly possible but it will not be at N2 / 18A, that design win competition is over, absolutely.
The big problem I see is trust and that comes down to setting the proper expectations and delivering accordingly. I saw a lot of eye rolling at the Intel Foundry event from semiconductor insiders. Hopefully IFS can deliver and not get hooked on selling futures (BSPD and High-NA EUV). It is hard selling "the thing" when you are always selling "the next big thing". Just my opinion of course.
hist78
Well-known member
"I do expect the UMC, Tower, and DoD/government work to be major volume in 2-5 years. Major is about 1 fab total (Say 50k wafers per rmonth)"
The DoD/Government semiconductor need (in terms of volume) is small and very often they are also high cost, high maintenance, and/or low profit margin. The reason is very simple: the volume of DoD weapon system production is limited and typically it spreads out to 10, 20 years or longer.
For example:
F35 Fighter Jet annual full-rate production: 150 ~ 160
Air Force new T-7A Red Hawk training jets, total order placed: 351
US Navy new Constellation class frigates: total 20 ships planned and annual production rate is 1 to 2.
Total number of US satellites (both government and civilian) launched in 2023: 2234
The upcoming next generation US Stealth Bomber, B-21 Raiders, total program planned acquisition: 100
The list can go on and on. Even assume they all use one type of chips, the volume is still very small.
This is why DoD declined to commit to Department of Commerce's proposal to have Intel build a semiconductor enclave in Ohio for national defense and security related chips production. It is not practical at all.
Last edited:
MKWVentures
Moderator
You are assuming DoD is just bombers and weapons. It is not. my notes on Intel foundry plans hint at that. But as I said, DoD/govt+Tower+UMC is one fab of volume in 2-5 years. This is true
The chips inside the weapons you show are several generation behind leading edge (N3). DoD does want ability to use leading edge, they do not have that ability today.
Please correct me if I'm wrong or out of date, but military/space chips often need radiation hardening. It is one reason why some older technologies are still used alot for mil spec and aero. SoI is one such tech. There are ways in system design to overcome some radiation effects, like triple voting schemes and etc.
Are there leading edge nodes coming out with rad hard versions?
Are there leading edge nodes coming out with rad hard versions?
hist78
Well-known member
Chips manufactured by TSMC have been used on the NASA's Perseverance Rover that went to Mars. I don't know if they were made using mature or leading edge nodes. But they definitely had been hardened by TSMC and/or its partners.
hist78
Well-known member
No, I don't assume DoD only needs chips for weapon system. If we dig into those high volume or higher volume semiconductors DoD needs, I don't see much chance that Intel can persuade DoD that Intel deserve an exclusive deal or an extraordinary amount of additional subsidy.
Such as: (not a complete list)
1. Chips for desktop and laptop: Apple, AMD, Intel, and Qualcomm (soon) are covering DoD's needs. DoD has no intention to limit itself to Intel only and has no ability to force Apple, AMD, or Qualcomm to use Intel as their foundry.
2. Chips for mobile (not the notebooks), IoT, and tablets: This is Qualcomm, Apple, Samsung, Broadcom, and MediaTek's strength. In many areas Intel doesn't have a comparable product at all.
3. Chips for servers, Cloud services, and high performance computing (including AI):
AMD, Nvidia, Intel, Google, Amazon, Microsoft, Cerebras Systems, Ampere Computing and Broadcom are DoD's suppliers. Intel is an important one in this group of DoD suppliers but Intel is not the only one. Again, in several areas Intel doesn't have a comparable product at all.
4. Networking, communication, sensors, analog, automotive, powers, industrial:
AMD, Cisco, Intel, Nvidia, Broadcom, Marvel, ADI, Texas Instruments, NXP, Infineon, Wolfspeed, and Microchips are strong players along with foundry service provider GlobalFoundries and TSMC. Once again, in several areas Intel doesn't have a comparable product at all.
5.Memory:
Micron. Intel has no memory product.
DoD wants to have a more robust and diversified semiconductor supply chain. To bring more fab capability onto US soil, no matter it's from Micron, TI, GF, Intel, TSMC, Samsung, or SK Hynix. Unless there is no other option, DoD doesn't want to limit itself to Intel just because Intel like to be the only one.
BTW, Tower Semi and UMC are both mature nodes companies. What do they do with Intel's upcoming leading edge fabs? Tower and UMC can bring volume to old and fully depreciated Intel fabs, not the leading edge Intel fabs that are getting Chips Act subsidies.
Although people might speculate that UMC and Tower Semi can act as resellers to promote Intel's leading edge fabs and consequently bring more volume to Intel. This is a wishful thinking to me.
A fabless company who dares to use leading edge nodes today is either big in size or rich in capital or both. Why can't such company talk to Intel Foundry directly? Do they really need a middleman?
Last edited: