Semiwiki on LinkedIn Semiwiki on Facebook Semiwiki on Twitter Register / Log in 
Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/hidden-inside-amds-financials.9015/
        )

    [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] => 2021770
            [XFI] => 1050270
        )

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

Hidden Inside AMD's Financials...

Daniel Nenni

Daniel Nenni

Admin
Staff member
As I have mentioned before, AMD is very loose with their financial reports and non GAAP numbers. And as the saying goes "where there is smoke there is fire". Here is another example:

In particular, Senior Analyst Hunter Gray found an unusual item in Advanced Micro Devices’ (AMD: $14/share) 10-K.

Like many firms, Advanced Micro Devices has non-operating expenses hidden in operating earnings. Non-operating expenses are unusual charges that don’t appear on the income statement because they are bundled in other line items. Without careful footnotes research, investors would never know that these non-recurring expenses distort GAAP numbers by lowering operating earnings.

In 2016, AMD took a $340 million (8% of reported revenue) charge related to a warrant agreement they entered into with West Coast Hitech L.P. This adjustment removes that non-operating expense and reduces AMD’s pre-tax operating loss by $340 million.

After all adjustments, AMD’s operating profit (NOPAT) in 2016 was -$124 million compared to GAAP net income of -$497 million. Without this adjustment one cannot get the true recurring profits of AMD’s business.

https://www.newconstructs.com/filing-season-finds-thursday-february-23/
 
Last edited:
As great as it is to see more competition in the x86 space, I think people are a bit too excited about AMD's prospects. For the time being, sales channels appear to be very limited and no major OEM has announced a Ryzen PC, although we are led to believe this will change in a couple of months. Second, Ryzen CPUs are only available on desktop class machines, notebook processors won't arrive until at least Q3. This means Ryzen will only appeal to a limited enthusiast market for now. By the time AMD is selling notebook PCs through major PC OEMs it will be up against Cannonlake, and AMD may not be as competitive. Also lost in the current hype around AMD is the fact that AMD is planning to skip 10nm entirely, and it might be 2-3 years before they see a node shrink. If that's the case, while AMD has a competitive part this year, they could fall back behind next year.
 
"[FONT=&quot]In 4Q16, AMD made its seventh-generation mobile APU (application processing unit), Bristol Ridge, available. It secured orders from the world’s top two PC makers, Hewlett-Packard ([/FONT]HPQ[FONT=Helvetica Neue, Arial, sans-serif]) and Lenovo, for their business computers."[/FONT]
[FONT=Helvetica Neue, Arial, sans-serif]Actually, AMD was already not doing too bad in the laptop market.
With the next generation RavenRidge APUs, things can only get better: they had already the best iGPU on the market, and the gap will even increase thanks to the new polaris/vega graphic architecture. Let´s add to that a very nice Ryzen CPU cluster, and cannolake could really be DOA in terms of performance per dollar (per watt).
Let´s be realistic, Intel circus is way too big and expensive to survive a lower margin market. They will badly feel the pain.
And let´s not forget servers. Since Naples is gonna be an upscale configuration of Ryzen, the performance per watt (per dollar) is going to be extremely good too.
Your move Intel.[/FONT]
 
I'm planning on building a desktop system with my Nephew this summer and it will have an AMD CPU. I don't think you will see a better chip value anytime soon. Unless of course he chokes on his finals then it will be an Intel CPU. :p

It is interesting to note that the majority of the Intel 14nm chips come from Fab 32 in Chandler AZ (updated from 22nm) which is a fully depreciated fab so wafers couldn't be cheaper. AMD, on the other hand, is on GF Fab 8 in Upstate NY, a greenfield fab, so wafers are costed at a premium. AMD is also running asset light and gave up their packaging operations in 2015. Given the pricing I have seen thus far, the first AMD chips are certainly lost leaders so this is going to be another rough year financially for AMD.
 
Last edited:
Daniel, about the foundry wafer cost, you are a bit pessimistic. Let´s stick to the numbers.
A 14nm LPP wafer costs definitely less than 10k$ to AMD. Now if we talk about Ryzen CPUs, considering the chip size and the node average yield, AMD is getting at least 200 shippable CPUs per wafer. Both wafer cost and yield are a worst case scenario.
And even in this case, at an average selling price of 250$ per piece, AMD can get ~50K per wafer. But let´s say 40k$. It is a 30k$ margin per wafer. From here they have to pay of course for the R&D, tape out costs and all the rest.
Now this last part I´m pretty sure is way higher at Intel than AMD. They have over 100k employees, AMD is way more flexible.
I truly believe that if Intel is forced to lower the margins on CPUs, AMD makes more profit than Intel.
 
This is certainly not my area of expertise but I have had this discussion with experts and I will have more at SPIE next week. The estimates I have seen today put the 14nm wafer price to AMD at about $7k whereas Intel wafer price is about $3k. The AMD die price is about $5 versus Intel's $2.50 since AMD does get more die per wafer. Packaging is a much more complicated formula that I am still working on but my guess today is that Intel has the advantage there as well. Maybe not 2x like the die but probably close to that since they own packaging and ship many more chips (economies of scale). And remember, AMD is just now getting chips to market so there is a manufacturing learning curve that has to be factored in whereas Intel 14nm has been shipping for 2+ years.

I should also note that Intel 14nm had one of the most expensive learning curves in the history of Intel (my opinion) but that is not factored in.






astilo said:
Daniel, about the foundry wafer cost, you are a bit pessimistic. Let´s stick to the numbers.
A 14nm LPP wafer costs definitely less than 10k$ to AMD. Now if we talk about Ryzen CPUs, considering the chip size and the node average yield, AMD is getting at least 200 shippable CPUs per wafer. Both wafer cost and yield are a worst case scenario.
And even in this case, at an average selling price of 250$ per piece, AMD can get ~50K per wafer. But let´s say 40k$. It is a 30k$ margin per wafer. From here they have to pay of course for the R&D, tape out costs and all the rest.
Now this last part I´m pretty sure is way higher at Intel than AMD. They have over 100k employees, AMD is way more flexible.
I truly believe that if Intel is forced to lower the margins on CPUs, AMD makes more profit than Intel.
Click to expand...
 
Daniel thanks for the Cloud details, very interesting. The wafer price figures are also very accurate, but dice prices cannot be right (for both companies).
Even with 100% yield, Ryzen could have at most 300 dice per wafer based on its size, so it is going to cost more than 23$ per piece definitely. Taking also a realistic yield into account, it would most likely be anything between 30 and 40$.
There is still plenty of margin to be profitable there.
 
Astilo, are you sure about your die-count?

If I calculate right, a 300mm wafer has about 70 000 mm² area. If the die size is about 100 mm² (guessed, for an 8 core part), this would give you about 700 parts, if you ignore clipped parts and yield. With this I would still expect way over 300 dies? For a 4 core part, there is a die size of 44 mm² mentioned in the internet, resulting in a much higher number, of course.
 
My 2 cents: AMD chips are being sold for hundreds of dollars, rather than 10s. That has to mean profitable economics. Most 14nm chips are mobile parts being sold for $25-50 per chip. With AMD processors selling for $200-300 each, even with a considerable die size, it's likely to be a pretty substantial gravy train, with considerable positive cash flow. Whether that cash flow hits the bottom line, could be a different story. The warrants in the 10K are pretty weird and make me wonder what deals AMD had to make to get Ryzen to market.
Intel Custom Foundry wafer costs are a lot higher than $10K per wafer.
 
Where did you get the 100mm2 figure? It is almost twice as big.
Yes, 44mm2 is the size of just 4 Ryzen core, then you have 16mm2 for 8Mb L3, 6mm2 for the L2, etc etc. All must be multiplied by 2 for the 8 core variant. Below is a die shot published by AMD:

View attachment 19235
 
Daniel Nenni said:
I can assure you Google also evaluated an AMD board. Let's see what Amazon does. Microsoft cloud will also be Intel.

By Urs Hölzle, Senior Vice President, Google Cloud Infrastructure
I’m excited to announce that Google Cloud Platform (GCP) is the first cloud provider to offer the next generation Intel Xeon processor, codenamed Skylake.

Google Cloud Platform Blog: Google Cloud Platform is the first cloud provider to offer Intel Skylake
Click to expand...

This is certainly a nice endorsement, but by no means does it indicate anything exclusive.
 
You must log in or register to reply here.
Back
Top