News AMD 1Q19 Earnings Report

[Phynaz]

I'm surprised how long it is taking AMD to make some serious profits, considering how much better their CPU lineup is now compared to the Bulldozer days.

But now you have the consoles dragging them down. Lisa Su said the semi-custom business is going to struggle until 2020, and it's probably second half 2020.

 

[moinmoin]

But now you have the consoles dragging them down. Lisa Su said the semi-custom business is going to struggle until 2020, and it's probably second half 2020.

No, the decrease in semi-custom business actually increases the gross margin as that's below corporate average for the consoles. The product mix decides the gross margin (in the call they later talk about which businesses are above and below corporate average, e.g. consumer graphics are below average).

"As you know very well we have businesses that are higher than corporate average, gross margin and lower than corporate average and the mix of the businesses, semi-custom that we just talked about lower than corporate average is driving the 41% guide for Q2."

https://seekingalpha.com/article/42...-2019-results-earnings-call-transcript?page=7

About console business in general:

"Joe Moore
Great. Thank you. I'm wondering if you can talk about gross margin in the consul segment, obviously, that's been pretty low historically. You also have customers funding for R&D but as you look to the next generation of consoles, do you see opportunity to improve the gross margin in that space?

Lisa Su
Yeah, Joe. I think it's a bit early to talk about margins for the next generation. As you stated, the gross margins of the console business are below corporate average. The operating margins are quite good because the customers are paying the engineering expenses for it. But I think the gross margins are below corporate average. We would expect though and I think you would expect us, as the company continues to grow, the percentage of the company that is semi-custom is lower than it has historically been."

https://seekingalpha.com/article/42...-2019-results-earnings-call-transcript?page=8

 

[PotatoWithEarsOnSide]

He was responding directly to the profits themselves, not profit margins. What you say is correct, but you're both on differing wavelengths.

 

[moinmoin]

He was responding directly to the profits themselves, not profit margins. What you say is correct, but you're both on differing wavelengths.

Eh, that's an inane point then considering the consoles are in their 7th year now. Nobody in their right mind expects console hardware sales (even less profit from it) to remain flat or increase at that point in time.

 

[krumme]

Eh, that's an inane point then considering the consoles are in their 7th year now. Nobody in their right mind expects console hardware sales (even less profit from it) to remain flat or increase at that point in time.

It was a response to comment from Chadboga not understanding profit is so low vs bd days now zen is here.

It's because people still think consoles is bad business.

I mean that's exactly why I wrote my comment about it. Margins is not equal to profit. Ask Amazon.

Look at your own quotes. Analysts is obsessed with margins and cant get their fat head out that understanding. Beats me how they are still so obsessed with that. One number of 1000 of interest yet get 5% of the time and 20% attention?

 

[krumme]

I'm surprised how long it is taking AMD to make some serious profits, considering how much better their CPU lineup is now compared to the Bulldozer days.

If you look at amd entire history is one long road of loss. Next to no profit some years and only a handful with some serious profit.

We have been over the consoles selling far less now taking away profit. Same to gpu derived crypto sales. It more or less explains it all.

That said my hunch is Lisa is used to sandbag a bit. I dont think zen server catched up so fast as they hoped. And zen apu first gen with its disappointing battery usage didn't have good market penetration.

Now the apu big disadvantage is fixed. Zen2 looks like a blast for servers so I think next year will be real profitable. Regardless if gpu still tank.

Remember also the wsa. If amd have such a relatively strong portfolio in 2024 when it ends it will give them a solid boost. That's 5 years from now but in amd history of more or less loss all the time - little profit often mostly lipstick - then that time will come too.

 

[rbk123]

But now you have the consoles dragging them down. Lisa Su said the semi-custom business is going to struggle until 2020, and it's probably second half 2020.

He's just trolling you guys. The AMD lovers are all saying how great things are looking. The Intel lovers are going to say "I'm surprised they're not doing better."

 

[Abwx]

AMD marketshare in DTs/Laptops/Servers was 15.8/12.1/3.2% in Q4 2018, up from 12/6.9/0.8% in Q4 2017, progress is decent in Mobile but is somewhat underwhelming in DTs, as for Servers that s a mixed bag, it does depend a lot of adoption by a few OEMs.

https://www.computerbase.de/2019-05/dell-amd-epyc-server-verdreifacht/

 

[Yotsugi]

but is somewhat underwhelming in DTs

No iGPUs means slower, way slower OEM box adoption.

as for Servers that s a mixed bag, it does depend a lot of adoption by a few OEMs.

Naples is a weaksauce offering versus Rome, so don't expect H1'19 the prop the server share up.

 

[Shivansps]

AMD marketshare in DTs/Laptops/Servers was 15.8/12.1/3.2% in Q4 2018, up from 12/6.9/0.8% in Q4 2017, progress is decent in Mobile but is somewhat underwhelming in DTs, as for Servers that s a mixed bag, it does depend a lot of adoption by a few OEMs.

https://www.computerbase.de/2019-05/dell-amd-epyc-server-verdreifacht/

I you want my opinion, at least what happens im my country is that the 2200G and 2400G are driving AMD forward... 2700 sales are nearly nothing vs 8700/9900, the 2600 is doing a bit better vs the 8400/9400F but it is still bad.

The 200GE is underwhelming, i hoped it would do much better on the low end, but no, the A8-9600 was far better in sales.

AMD needs to step up their game at high end, im sure the 3000 series will do that and figure out that Vega 3 at low end is not enoght.

 

[amd6502]

No iGPUs means slower, way slower OEM box adoption.

Naples is a weaksauce offering versus Rome, so don't expect H1'19 the prop the server share up.

yes and no. i agree with you 50%

AMD needs to step up their game at high end, im sure the 3000 series will do that and figure out that Vega 3 at low end is not enoght.

It's quite good enough, but some more variety in the low end (~50%+ more shaders) would be better. BR is a better low end product than RR.

 

[Yotsugi]

yes and no. i agree with you 50%

Naples is an absolute, total weaksauce of an offering versus Rome.
But, being frank, everything is a weak offering versus Rome.

 

[moinmoin]

AMD marketshare in DTs/Laptops/Servers was 15.8/12.1/3.2% in Q4 2018, up from 12/6.9/0.8% in Q4 2017, progress is decent in Mobile but is somewhat underwhelming in DTs, as for Servers that s a mixed bag, it does depend a lot of adoption by a few OEMs.

https://www.computerbase.de/2019-05/dell-amd-epyc-server-verdreifacht/

It would be interesting to know how much of the respective markets go through OEMs. I can imagine AMD's market share to look much better in the DIY and B2B markets, and those being the majority of the overall market share increases so far. This obviously excludes laptops, so the increase there is the promising.

The fact that OEMs are slowly moving currently helps Intel, their steady minor updates satisfy the existing channels, most of which aren't eager to apply any big changes to their currently well oiled businesses. Now with Zen 2 arriving, it may be the best time for OEMs to position themselves as "special" ahead of all generic "lazy" competition by offering and supporting AMD derived systems that aren't possible with current Intel chips. Though this again will take time on AMD's side as well at least for the laptop market, considering 7nm APUs arrive last.

In the call Su mentioned going directly for big companies for server sales, skipping OEMs:
"I would just say that the Enterprise tends to move a bit more slowly than cloud. There are longer qualification cycles because their qualifications cycles on both the OEM side as well as the end customer side. We are continuing to build out our direct sales force as it relates to facing the Fortune 1000 customers and CIOs in that area."
https://seekingalpha.com/article/42...-2019-results-earnings-call-transcript?page=9

 

[Yotsugi]

In the call Su mentioned going directly for big companies for server sales, skipping OEMs:
"I would just say that the Enterprise tends to move a bit more slowly than cloud. There are longer qualification cycles because their qualifications cycles on both the OEM side as well as the end customer side. We are continuing to build out our direct sales force as it relates to facing the Fortune 1000 customers and CIOs in that area."

Rome is a prime hyperscaler chew, so AMD going directly to them is very expected.

 

[Phynaz]

Eh, that's an inane point then considering the consoles are in their 7th year now. Nobody in their right mind expects console hardware sales (even less profit from it) to remain flat or increase at that point in time.

So inane that the CEO had to make a point of it

 

[Phynaz]

He's just trolling you guys. The AMD lovers are all saying how great things are looking. The Intel lovers are going to say "I'm surprised they're not doing better."

What are the rational people saying?

 

[maddie]

So inane that the CEO had to make a point of it

Don't you know by now that most 'analysts' don't actually analyze from basic principles?

They're almost always 'after the fact' individuals. I guess the best ones are busy getting rich for themselves by seeing trends, etc, whereas, the rest jump on later. These fora allow some discussion into future developments and I'm know that some here have made monetary gains.

In other words, sometimes you have to hold the slower one hands and guide them into the light. We would have seen much more preemptive behavior with regards to competitors stocks if predictive work is being done, instead of only waiting for press releases and oracle like statements.

Apologies to any analysts reading this, but by this fact, you're not in the group I'm accusing.

 

[NostaSeronx]

I think AMD is going to get hit hard by the Annual Fixed Revenue Floor.

Yeah, they can do whatever they want beyond 7nm. However, pretty much the complete Original WSA is back on for everything above and not equal to 7nm.

 

[moinmoin]

I think AMD is going to get hit hard by the Annual Fixed Revenue Floor.

Yeah, they can do whatever they want beyond 7nm. However, pretty much the complete Original WSA is back on for everything above and not equal to 7nm.

Any reason you think they'll "get hit hard" by that? Thanks to the IOC all AMD CPU products still contain parts by GloFo. Thanks to guaranteed availability for PRO (3 years from launch) and embedded (10 years from launch) chips (I'm sure server chips also have some extended availability range, but can't find that right now) this will be true well beyond the current WSA.

 

[NostaSeronx]

Any reason you think they'll "get hit hard" by that?

The exact nature is GlobalFoundries is going to be de-ramping 14HP/14LPP/14LPP+/12LP/12LP+ soon, with a complete halt of those lines by third quarter 2020.

The hard hit is how will AMD sustain the revenue going to GlobalFoundries. If there isn't going to be wafers for those nodes.

Q4-2019 => ~75% of the wafer starts
Q1-2020 => ~50% of the wafer starts
Q2-2020 => ~25% of the wafer starts
Q3-2020 => ~0% of the wafer starts

TSMC won the FinFET fight, hence no customers to sustain the nodes after IBM/AMD leave. Production of course dips after the next product is introduced. (Malta contributes the most to GlobalFoundries net loss.)

TSMC has lower prices, higher capacity, more incentives, etc. There is nothing stopping AMD from making a 7nm I/O chip at TSMC. AMD only needs to produce MPUs/GPUs/Chipsets per Orignal WSA agreement for anything above the 7nm node.

Hence the hard hit, the Malta nodes are suspect and high-risk.
-> Go TSMC pay 7nm.
-> Port 14LPP products to Samsung.
-> Stay GlobalFoundries port to another process.

All of it costs money, only one doesn't cause delta percentage take-pay.

 

[Hitman928]

The exact nature is GlobalFoundries is going to be de-ramping 14HP/14LPP/14LPP+/12LP/12LP+ soon, with a complete halt of those lines by third quarter 2020.

The hard hit is how will AMD sustain the revenue going to GlobalFoundries. If there isn't going to be wafers for those nodes.

Q4-2019 => ~75% of the wafer starts
Q1-2020 => ~50% of the wafer starts
Q2-2020 => ~25% of the wafer starts
Q3-2020 => ~0% of the wafer starts

TSMC won the FinFET fight, hence no customers to sustain the nodes after IBM/AMD leave. Production of course dips after the next product is introduced. (Malta contributes the most to GlobalFoundries net loss.)

TSMC has lower prices, higher capacity, more incentives, etc. There is nothing stopping AMD from making a 7nm I/O chip at TSMC. AMD only needs to produce MPUs/GPUs/Chipsets per Orignal WSA agreement for anything above the 7nm node.

Hence the hard hit, the Malta nodes are suspect and high-risk.
-> Go TSMC pay 7nm.
-> Port 14LPP products to Samsung.
-> Stay GlobalFoundries port to another process.

All of it costs money, only one doesn't cause delta percentage take-pay.

You really think GF will stop making any finfet nodes barely a year after Zen 2 is released which has embedded and server products reliant on them being made? Pretty sure if GF tried that, AMD would have a get out of WSA free card. They'd be driving away their biggest customer. The only way I could see this happening is if AMD agreed with them to move to an FDX node for the IO chip but I doubt AMD does this without some incentive from GF and I doubt it happens less than a year after Zen 2 is released.

 

[DrMrLordX]

GF will do whatever it takes to keep that I/O die business. If that means supporting "old" nodes then so be it.

 

[maddie]

The exact nature is GlobalFoundries is going to be de-ramping 14HP/14LPP/14LPP+/12LP/12LP+ soon, with a complete halt of those lines by third quarter 2020.

The hard hit is how will AMD sustain the revenue going to GlobalFoundries. If there isn't going to be wafers for those nodes.

Q4-2019 => ~75% of the wafer starts
Q1-2020 => ~50% of the wafer starts
Q2-2020 => ~25% of the wafer starts
Q3-2020 => ~0% of the wafer starts

TSMC won the FinFET fight, hence no customers to sustain the nodes after IBM/AMD leave. Production of course dips after the next product is introduced. (Malta contributes the most to GlobalFoundries net loss.)

TSMC has lower prices, higher capacity, more incentives, etc. There is nothing stopping AMD from making a 7nm I/O chip at TSMC. AMD only needs to produce MPUs/GPUs/Chipsets per Orignal WSA agreement for anything above the 7nm node.

Hence the hard hit, the Malta nodes are suspect and high-risk.
-> Go TSMC pay 7nm.
-> Port 14LPP products to Samsung.
-> Stay GlobalFoundries port to another process.

All of it costs money, only one doesn't cause delta percentage take-pay.

Why halt the lines?
Is GloFlo losing money on 14/12 nm?
AFAIK most companies have fabs with products on old nodes still being produced.
What happens to the GloFlo old fab?
A massive investment to upgrade or are they mothballing fabs and closing up shop?

 

[coercitiv]

What happens to the GloFlo old fab?
A massive investment to upgrade or are they mothballing fabs and closing up shop?

This is what happens.

They sell, they retool, they do everything that makes sense financially. Some fabs live longer that we imagine. The discussion above is borderline comical and people should know better that take the cat's omniscient whispers at face value.

 

[maddie]

This is what happens.

They sell, they retool, they do everything that makes sense financially. Some fabs live longer that we imagine. The discussion above is borderline comical and people should know better that take the cat's omniscient whispers at face value.

It was meant to be farcical, but I guess I missed by a mile.