Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/intel-change-depreciation-years-to-some-equipment.17383/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Intel change depreciation years to some equipment

VCT

Active member
"Before turning to Q1 guidance, let me take a moment to discuss an accounting change that will impact our results beginning in the first quarter. Effective January 2023, we increased the estimated useful life of certain production machinery and equipment from five years to eight years. This change better reflects the demonstrated economic value of our machinery and equipment over time and is more aligned with the business model changes inherent to our IDM 2.0 strategy."

The actual cost difference in compare to TSMC will be larger than people realize.
 
The actual cost difference in compare to TSMC will be larger than people realize.
How so? Until now the model was 5 years, this doesn't mean that 2022's cost was being hidden behind this different depreciation model (my shot in the dark guess is that a large number of the 10nm tools are either done or almost done depreciating). Long term this also doesn't change structural cost at all. All this means is that in the short term their nodes are cheaper then before (enabling internal and external customers to get on the bus sooner), in the medium term they are more expensive, and in the long term they are the same cost. To my understanding, all they have done is flatten their depreciation curve. This might also have the effect of encouraging DCAI and foundry customers to get on new nodes faster and improve the rate of yield learning. This isn't some "200 IQ play to hide their wafer cost from investors and look more competitive then they are". To me this just seems like a logical play given the current business realities and future business model intel is gunning for.
 
Last edited:
How so? Until now the model was 5 years, this doesn't mean that 2022's cost was being hidden behind this different depreciation model (my shot in the dark guess is that a large number of the 10nm tools are either done or almost done depreciating). Long term this also doesn't change structural cost at all. All this means is that in the short term their nodes are cheaper then before (enabling internal and external customers to get on the bus sooner), in the medium term they are more expensive, and in the long term they are the same cost. To my understanding, all they have done is flatten their depreciation curve. This might also have the effect of encouraging DCAI and foundry customers to get on new nodes faster and improve the rate of yield learning. This isn't some "200 IQ play to hide their wafer cost from investors and look more competitive then they are". To me this just seems like a logical play given the current business realities and future business model intel is gunning for.
2023 Q1 outlook margin is 34.1% which reflects new depreciation years.
 
Yes but my point was that this isn't hiding one of intel's many current issues, because the current status quo of intel's higher wafer cost is already visible within the current earnings results. TSMC could do the same if they wanted to, but they probably don't want that extra cost at the middle end of their node lifetime. Removing depreciation ASAP and selling the node at similar prices at massive profits is what enables TSMC to have such high margins. Their leading edge customers would also probably still buy the same number of wafers regardless of the depreciation model they used. In short for TSMC 5 years probably still makes sense. For intel I think this makes a ton of sense because 14nm tooling has long since deprecated by now, and much of the tooling for i4 and beyond hasn't started deprecation yet. By the time that this hurts wafer costs either intel has either sunk or swam. Additionally with the main margin/volume leader being CPUs getting those as cheap as possible makes sense for intel's business.
 
"Before turning to Q1 guidance, let me take a moment to discuss an accounting change that will impact our results beginning in the first quarter. Effective January 2023, we increased the estimated useful life of certain production machinery and equipment from five years to eight years. This change better reflects the demonstrated economic value of our machinery and equipment over time and is more aligned with the business model changes inherent to our IDM 2.0 strategy."

The actual cost difference in compare to TSMC will be larger than people realize.

Intel is trying hard to reduce their true cost on paper in order to get by in short term. Typically a highly profitable and fast growing company prefers to have higher depreciation expenses and faster depreciation schedule. It will help them to reduce tax liability and invest more to expand their capacity. They are not worried about the potential to incur higher tax due to the decreasing depreciation expenses once the accelerated depreciation finished. It's because they keep investing more to capture the market opportunities and it can keep the depression expense high anyway.

That's why governments around the globe like to use accelerated depreciation incentives to encourage companies to invest.

Intel's decision to prolong the depreciation schedule from 5 years to 8 years might indicate that:

1. They desperately want to improve the profit margin on paper. Or in the worst situation turn the true loss to profit on paper.

2. They believe their profitability is very limited in the future and consequently tax liability (based on profit) is limited. Reduced depreciation expenses won't bring up tax liability too much.

3. To make both Intel CPUs and IFS service more competitive because the lower depreciation expenses and lower cost in short term.

The problem is that Intel's fast shrinking revenue and shrinking profits are far more worse than this accounting manipulation (or "maneuver" in a more polite way) can help.
 
Last edited:
Intel already foresee big margin crash in the future due to rising equipment cost so they do some adjustment on accounting to prevent getting fired.
 
They went from 4 to 5 years a couple of years ago, now 5 to 8. So they have effectively gone from 4 to 8 years. This could have been justified based on the longevity of the 14nm node, and implicitly assumes that all future nodes will have a similar longevity. Is this a reasonable assumption going forward? Especially with the 4 nodes in 5 years plan? Personally I don't think so, but I'd like to hear other's thoughts on the matter.
 
They went from 4 to 5 years a couple of years ago, now 5 to 8. So they have effectively gone from 4 to 8 years. This could have been justified based on the longevity of the 14nm node, and implicitly assumes that all future nodes will have a similar longevity. Is this a reasonable assumption going forward? Especially with the 4 nodes in 5 years plan? Personally I don't think so, but I'd like to hear other's thoughts on the matter.
Intel has said they want to run nodes in perpetuity like TSMC does. Although I think you misunderstand what 5 nodes means. 1 was i7 (aka 10nm ESF). 2 is intel 4 (the old 7nm). 3 is intel 3 (7nm+ as well as well as finishing all of the bits that weren't in i4). 4 was 20A (former 5nm). 5 is 18A, but as Dan has said it is a similar situation to i3. Tool wise there should be little to no differences between the tick nodes and the tock nodes. Back when intel announced that GNR was moved from i4 to i3, I think they said something about how i4 was fwrd compatible with i3 (not really a shock but good to know I guess). My guess is once the tock node is ready production will likely be seamlessly transferred to these objectively better processes versions. Think of it like 14 or 10nm, where the new version didn't suddenly mean you needed a new production line, it was just an improved version. Same deal with N5 thru N4P or N4x.
 
They went from 4 to 5 years a couple of years ago, now 5 to 8. So they have effectively gone from 4 to 8 years. This could have been justified based on the longevity of the 14nm node, and implicitly assumes that all future nodes will have a similar longevity. Is this a reasonable assumption going forward? Especially with the 4 nodes in 5 years plan? Personally I don't think so, but I'd like to hear other's thoughts on the matter.
Most companies like the accelerated depreciation schedule. It can help a company to reduce the taxable net profit, lower the tax liability, and accumulate more cash faster for reinvestment. Intel is doing the opposite way for the purpose of increasing the gross profit margin on paper. I can't think about other real benefit.
 
Last edited:
Most companies like the accelerated depreciation schedule. It can help a company to reduce the net profit, lower the tax liability, and accumulate more cash faster for reinvestment. Intel is doing the opposite way for the purpose of increasing the gross profit margin on paper. I can't think about other real benefit.
It also allows for faster ramp. Let's say in year 0 an intel 18A wafer costs $100 and let's say $40 is dep. if they want to sell to intel with a 10% margin and an external customer for a 35% margin then they have to sell the wafer for $110 and $135 respectively. However if they can flatten their depreciation curve, that lowers the cost from dep from say $40 to maybe $25 (I don't have 5 and 8 year MAACRs tables on me but just roll with me here). Then they can sell those wafers at $94 and $115 for the same margin. You can of course also sell these wafers at the same price for a larger margin. Either way the lower prices driving demand up or greater margins allowing for more investment back into IFS will allow IFS to scale up faster then they otherwise could over the course of this decade. Of course under this new scheme at year 6-8 you are still paying for deprecation and can't offer as large of a savings to your customers or raise your margin as you much as might like, but lower processing costs and better yields would at least help soften that blow.

One final benefit I can think of is when this is combined with the change to wafer agreements for internal customers, it might encourage DCAI to move to new nodes more aggressively. This will allow that BU to be more competitive than if they had to wait longer before moving to the new node when it finally meets their margin targets.
 
IFS can not sell to Intel at 10% margin and an external customer for a 35% margin. It will just prove to all that IFS is not independent.
 
IFS can not sell to Intel at 10% margin and an external customer for a 35% margin. It will just prove to all that IFS is not independent.
Before I go any further, those margin number were just random numbers no reason to get upset about them. I have no clue what everybody's deal at the various foundries are, and even then not every customer will get the same deal. See for example TSMC; where members of the inner circle get prefered wafer pricing and first dibs on signing wafer agreements.

With that out of the way, what does it matter? If I am mediatek buying my wafers for smart TVs all that matters to me is what is the PPAC for intel 16 vs 28nm and 16/14nm, as well as how much capacity will they give me? What intel charges intel is not my concern. I wouldn't be surprised if when they go to Samsung, Samsung LSI pays a 0% margin. Either way much better than the alternative of the non internal foundry model, where intel can kick me to the bottom of the list and only ever pay at cost.

Also define independent? They have separate management and P&Ls, but I don't know if I would call them independent anymore than I would say Radeon is independent from AMD.
 
Before I go any further, those margin number were just random numbers no reason to get upset about them. I have no clue what everybody's deal at the various foundries are, and even then not every customer will get the same deal. See for example TSMC; where members of the inner circle get prefered wafer pricing and first dibs on signing wafer agreements.

With that out of the way, what does it matter? If I am mediatek buying my wafers for smart TVs all that matters to me is what is the PPAC for intel 16 vs 28nm and 16/14nm, as well as how much capacity will they give me? What intel charges intel is not my concern. I wouldn't be surprised if when they go to Samsung, Samsung LSI pays a 0% margin. Either way much better than the alternative of the non internal foundry model, where intel can kick me to the bottom of the list and only ever pay at cost.

Also define independent? They have separate management and P&Ls, but I don't know if I would call them independent anymore than I would say Radeon is independent from AMD.
On the other hand, if you are the leader of IFS, you want to do well and be profitable. The other department's result is not your priority. Why would you do Intel a favor by giving them a cheaper cost than may later cost you your job?
 
On the other hand, if you are the leader of IFS, you want to do well and be profitable. The other department's result is not your priority. Why would you do Intel a favor by giving them a cheaper cost than may later cost you your job?
Because it is one company, and success for one is success for all? If design fails then the whole firm fails, and at least for the foreseeable future IFS is not in the position to just "spin off" or otherwise have design side implode.

Either way they all still report to the same CEO and BOD. If they tell you to lower design's margin you lower their margin. There is also just the plain fact that intel is buying in bulk. For example if Apple wanted to come to intel exclusively and fill up AZ and Ireland by themselves, then I don't see why intel wouldn't give them one hell of a sweetheart deal. But if intel remains IFS's biggest customer by a wide margin then why shouldn't they get the best deal if for no other reason than they are buying in bulk/ensuring the fab is kept full.
 
Back
Top