Could Intel’s Foundry Be Worth $500 Billion?

[MKWVentures]

CapEx waste or capacity mismatches could explain why Intel’s revenue declined during a booming semiconductor cycle, even while the company claimed its own capacity was tight after spending $127.27 billion in CapEx.

But there’s another possibility (entirely legal): Intel may have been capitalizing expenses. By doing this, Intel can spread costs, costs that otherwise would have been fully expensed in the year they occurred, over eight years or more through depreciation or amortization. The benefit is that it boosts reported profit or reduces reported losses.

The drawback is that when the market needs Intel chips, Intel may not have enough effective, real capacity available. Capitalized expenses looks good on financial statements, but it doesn’t guarantee usable manufacturing capability for an IDM - Intel.

Intel capitalizes equipment related service contracts like everyone and building improvements. I am not sure of any "operating expenses" that are capitalized.

Just as an example: I would guess that If Intel did no new nodes in production (just R&D) and no new wafer volume expansion. the Capex budget would be 10-12B per year

I can tell you that Intel made some very problematic calls in 2024 and as a result has the wrong capacity in the wrong place at the wrong time

 

[XYang2023]

All my spreadsheets are forward looking. DZ specifically said IF Intel has to spend to ramp a customer for 14A it would HURT their chance for breakeven. Sometimes AI is not the best summary of what is actually happening LOL.

If there is tons of demand out there, they would have to start designing on Intel silicon today. Which means they would not have meaningful revenue until 2029? correct.

What is you estimate of Tesla demand as a percentage of total foundry TAM?

Please let me know what numbers would be a scenario for breakeven (other than a massive write off, which may be planned). What do you think will happen?

If the timeline delay is due to external demand for 14A, then that is a good problem to have. If so, IFS should be valued on a sum-of-the-parts basis, which appears to be reflected in analysts’ recent target prices.

CPU supply is constrained. Whatever Intel can deliver is being sold. Also LBT also said they pulled Coral Rapids forward.

Tesla could also help share the costs of foundry/process development

 

[XYang2023]

An alternative approach to address AI accelerator market:

https://twitter.com/x/status/2046306708012884119

 

[XYang2023]

https://twitter.com/x/status/2046236633960485359

Also, does this growth imply that Intel’s products are becoming more competitive and starting to regain some market share?

 

[MKWVentures]

Also, does this growth imply that Intel’s products are becoming more competitive and starting to regain some market share?

Is Intel regaining market share? we should know this week

Question: If IFS gets 3 large customers. total revenue = 5B/year. and IFS has losses of 2-3B per year through 2030. [hypothetically] .... Is that a good thing? should they continue?

 

[kevin01]

those are great questions. They were asked in a emotional manner at last earnings report by analysts

Our model based on some inputs from very wise people on this site and inputs from people who know the Intel fabs:

1) Intel 7 demand is vert high for both client and server. 85+ of Intels sales are Intel 7 or Older.
2) Intel took Intel 7 capacity down thinking people would move to 18A
3) 18A cannot be ramped full out and customers forced to 18A (or 3) due to financial constraints and negative margin impacts (Intel 7 has HIGHER margin than 3 or 18A).
4) People are not as interested in higher price points for Newest products from Intel. Intel continues to lose share so Intel has limitations.

The amount of wasted and stranded Capex by Intel from 2021-2024 is very high....

Hopefully they clear up some items this week.

I respect your opinions and appreciate your insights, but you and that analyst guy, Stacy, are the closest to a 'bean counter' type of person I know. No offense.

By your logic, Intel should always stay on Intel 7 and older, because those have high margins.

Fortunately, there are investors who have a bit longer time horizon.

 

[XYang2023]

Is Intel regaining market share? we should know this week

Question: If IFS gets 3 large customers. total revenue = 5B/year. and IFS has losses of 2-3B per year through 2030. [hypothetically] .... Is that a good thing? should they continue?

1. IDM/2.0: Margin stacking
2. LBT answered it in the video: 94% of advanced semi manufacturing in Taiwan. Hence, there is an embedded geo-risk call option.

 

[XYang2023]

https://twitter.com/x/status/2046380733158281283

 

[XYang2023]

I respect your opinions and appreciate your insights, but you and that analyst guy, Stacy, are the closest to a 'bean counter' type of person I know. No offense.

By your logic, Intel should always stay on Intel 7 and older, because those have high margins.

Fortunately, there are investors who have a bit longer time horizon.

Stacy’s previous analyses were not correct, as he discounted the importance of CPUs in those assessments.

When I was doing research last year, we found that LLMs without verification were less useful. This observation has also been reported in several other studies. Several times, I discussed this with my supervisor, arguing that most of this verification work is performed on the CPU. At the time, many people believed that Nvidia GPUs were the way, and the only way, even in academic environments. In any case, I just want to highlight that those analyses may not be accurate, as they contain inherent biases.