The AMD Q2 Results Thread

Arun

Unknown.
Moderator
Legend
Q2
Revenue: $1378M
Gross Margins Non-GAAP: 34%
Operating Income GAAP: -$457M
Operating Income Non-GAAP: -$332M
Interest Income+Expenses: -$80M
CPU+MCP Revenue & Income: $1,098M / -$258M
GPU Revenue & Income: $195M / -$50M
CE Revenue & Income: $85M / -$22M
Net Income: -$600M
ATI acquisition-related charges: -$78M
Severance charges: -$16M
Shareholders' equity: -$733M

Q1
Revenue: $1233M
Gross Margins Non-GAAP: 31%
Operating Income GAAP: -$504M
Operating Income Non-GAAP: -$363M
Interest Income+Expenses: -$62M
CPU+MCP Revenue & Income: $918M / -$321M
GPU Revenue & Income: $197M / -$35M
CE Revenue & Income: $118M / -$4M
Net Incomne: -$611M
ATI acquisition-related charges: -$113M
Severance charges: $0M
Shareholders' equity: -$580M

Source: http://amd.edgarpro.com/redirect_fr...rtBy=receivedate&AD=D&startrec=1&res=25&pdf=0

Editing over time to add more numbers and notes...
Note: Certainly better than I expected CPU-wise, and by quite a bit, nice! Still not that nice in absolute terms, but it's a start. Former-ATI is doing a fair bit worse than I had expected, however.

Quick Partial Notes
- Chipsets down 13%, down for Intel and up slightly for AMD.
- Handheld sales substantially, compensated slightly by DTV.
- $800M cash generating opportunities in the future via 200mm tools, Spansion, and more.
- CapEx for the year will be further reduced to $1.8B, compared to $2B previous guidance.
- Q3 revenue expected to be up with seasonality, operating expenses down slightly.
- Slowing down Fab38 ramping further based on Fab36 doing so well.
- Desktop CPU ASPs down by 10%, but up in servers.
- Expecting margins to be up (slightly, iirc) in 3Q07.
- Very happy about 65nm yields, for Barcelona too.
- Also very happy about the asset-light strategy, but will not disclose it yet for competitive reasons.
- 50% long-term gross margin goal was for pre-asset-light. Will not discuss expectations post that.
- Had a $30M inventory write-off that hurt margins slightly.
- Increasing mix of laptops/servers might help margins.
 
Margins have increased, which was against expectations. So that's good.

On the other hand, after a self-described "perfect storm" of a quarter, their bottom line improvement vs last quarter was slightly less than 2% from a loss of $611M to a loss of $600M.

And I'm getting more and more jaundiced about "one time acquistion related charges" when the damn things keep showing up in serious doses quarter after quarter. There's another $94M of that (with a bit of severance thrown in) in this report. Taking the "one time acquistion related charges" since the acquistion closed into the neighborhood of $750M. This for a merger that was described rapturously for its synergies, lack of overlap, etc. It's starting to get a little smelly to me what's really getting stuffed into these "acquistion related" one-time charges.

Quite frankly, my finger is starting to edge a bit closer to the panic button. I'm still not getting a sense of urgency from this executive team and it seems to me some urgency is required.
 
No its bad, it means they are charging more for stuff...

No, since they're obviously *not* charging more for stuff (what with the price war and all), it means their costs have come down.

Which makes sense as their yields improve at 65nm. But... I think the gross margin numbers are cold comfort. The hardcore crowd that served as soldiers for AMD in advancing them on the desktop and indirectly in the minds of corporate IT geeks has already gone, or is going, to Intel in the face of what looks to be an inadequate response to Penryn, let alone the standing challenge from Conroe. Longtime AMD folk like myself are further alienated by knowing that after 'arriving,' AMD did nothing with their ~2 years of advantage but sit on it and settle for marginal architectural improvements while they knew Intel was throwing money at a NetBurst replacement.

Ironically the server space has now become their refuge, aided by the inroads they have already made and HyperTransport. But as we see, said refuge does not make this otherwise lossy company profitable, and Intel will casually launch assault after assault on it.
 
Last edited by a moderator:
Yep, after the perfect storm they do seem to have steadied the ship somewhat but they do say they will have to do some more bailing out over the months ahead. Seems like choppy waters will continue for some time to come.

After the recent price cuts I assume gross margins will not continue to improve that much from extra 65nm production / cutting back expenditure so it will be interesting to see what happens in Q3. K10 will have to get up to speed quickly but I think they will make less loss next time by a big percentage, won't get into the black again though.

Yes, I agree with Geo, how can the ATi buy still be incurring costs of such large amounts ?
 
Frankly i don't understand what does Gapp-nonGaap means...
can someone exp-lain in plain english why they are loosing money from gpus?
Are they selling GPUs below the production cost?
To me it looks like selling gpus for 195mln$, while spending 245mln to produce them ... does that number includes salaries and research expenses for gpu?
 
http://seekingalpha.com/article/41686 There's the transcript.

GAAP is "Generally Accepted Accounting Principles" and is a legal standard that everyone has to follow in the US. Additionally, many companies will also report non-GAAP which is usually intended to exclude various stuff and give a truer picture of ongoing operations.

The problem is non-GAAP can be diddled to look a little rosier than reality in various manners. . . like by stuffing losses into one time acquisition charges and then excluding them on that basis. . .
 
Longtime AMD folk like myself are further alienated by knowing that after 'arriving,' AMD did nothing with their ~2 years of advantage but sit on it and settle for marginal architectural improvements while they knew Intel was throwing money at a NetBurst replacement.

They settled for incremental design improvements because several new designs failed to come to fruition.

K8 as we know it wasn't the original K8, and K9, K10, and perhaps another version of both were cancelled.

The more radical K10 was cancelled after six months of development time. The time gap does line up nicely for several false starts and then settling back on incremental improvement.

So it's not that AMD didn't do anything, just that what it did failed.

Failure is pretty common, but here again Intel's size allows it to fail more often and often concurrently.

AMD's limited resources means it can't fail as often as it does, and its failures have a higher time cost.
 
can someone exp-lain in plain english why they are loosing money from gpus? Are they selling GPUs below the production cost?
To me it looks like selling gpus for 195mln$, while spending 245mln to produce them ... does that number includes salaries and research expenses for gpu?[/QUOTE]

The problem is two fold: low volumes and low margins. They are not selling GPU's below production cost, but as you suspected, they are not making enough money per die and not selling enough dies to compensate voor salaries and R&D.

The chip industry has very large upfront expenses. If you can pull it off and sell tons of chips with a large markup, then you can make very large profits. If you don't, it's very easy to lose out.

This is why 2900 is somewhat of a problem: for the consumer it's a great deal wrt price/performance, but for ATI it must be a really expensive affair: very large die, more expensive process than the competition, more expensive PCB and no lower grade derivative (like a GTS) to off-load faulty dies.
 
How much longer can they last? 1.2billion in the 1st half of this year?

That merger was the worst idea ever! Well besides 3dfx buying STB. Anyways I felt the merger between AMD and ATI was bad and this is only backing up my assumptions.
 
How much longer can they last? 1.2billion in the 1st half of this year?

That merger was the worst idea ever! Well besides 3dfx buying STB. Anyways I felt the merger between AMD and ATI was bad and this is only backing up my assumptions.

I'm not expert, but i think the real problem is AMD have no "world saving products" in they pocket, Barcelona/Phenom won't change anythig because its late, Intel can answer any time for this CPU's, in GPU side is the same, when AMD coming out with rv670 performance part, NV answer will be follow shortly, selling products cheaper to stay competative not good for long term, and AMD do this alredy with Athlon x2 CPU's and Rx6xx based VGA's.

There is real chance AMD colapse this year and when this happening ATi sink with AMD too, this makes the situation from user aspect very bad.
 
I'm not expert, but i think the real problem is AMD have no "world saving products"
Let's be honest with ourselves, now. 3DFX didn't either. Rampage was one half-node behind, less integrated (4x1 vs 4x2) and had no early-Z or Z/Color-compression capabilities. Its memory controller was also isochronous, IIRC. Its only chance would have been in the OEM market as the lowest-end DX8 product of its time. While hardly an insignificant market segment, I doubt that it would have saved the company. Would it have put them in a position to be acquired on better terms? Maybe, who knows.

Anyway, sorry for that slight OT. Call it revisionism if you want (but I will disagree heavily), however I think it's certainly worth pointing out, and for one very good reason... If a company is not in a favorable position, then clearly, it and its fanboys will try to claim that its upcoming products will be greatly superior.

While this is sometimes true, I would however like to play the Devil's Advocate and claim that this is actually not true in the majority of cases: companies that did badly in the previous cycle are likely not to do much better in the next one, unless their innovation cycles differ from that of their competitors.

This was the case with the NV30: it was an evolutionary refresh of the NV20, arch-wise (don't tempt me to go into the details...) while R300 was a revolutionary Tabula Rasa. This was not as obviously the case with 3DFX regarding VSA-100 vs NV10. In the case of AMD, the cycles do diverge: Brisbane was a small change, while Conroe was a large change. And in the previous 'large changes' cycle, AMD did extremely well for a variety of reasons.

However, Barcelona is also a small change compared to Brisbane, relative to P4->Conroe. As such, the cycles of innovation do diverge, but they are *not* realigning in AMD's favor anytime soon (even though their historical engineering record is very good). This is the key problem in AMD's case, IMO. The time when the cycles will realign is with AMD's true next-gen architecture in 2009, which is an AMD California project.

I may sound a little bit crazy trying to apply speculative and highly theoretical models to this case. However, while far from perfect, I find this kind of reasoning to be much more probable than the alternatives. Your mileage may vary.

The question, based on this, is whether AMD will be able to survive (in its current form) until 2H09 or so, since I would be extremely surprised if the company became profitable again before then, no matter what the management team is claiming. Can they significantly reduce their quarterly loss? Sure. But actually turning it around and having a positive cash flow seems like a much harder thing to me.

In terms of today's negative news, there's this http://www.eetimes.com/news/latest/...d=R0RFUWBRUECUQQSNDLSCKHA?articleID=201200778 - but I'm not sure where they're getting the -$600B free cash flow. My calculations actually put it at -$1B, although I was unable to figure out why they would have burned through so much cash. Because AMD took extra debt in Q1, it does complicate matters; the Q2->Q3 evolution will be a much more straightforward one to analyze, so I'm looking towards that one.

If AMD keeps burning through cash and their debt rating keeps going down, that doesn't bode well for 2008, obviously... PE Angels, Oh, Where Art Thou?
 
While this is sometimes true, I would however like to play the Devil's Advocate and claim that this is actually not true in the majority of cases: companies that did badly in the previous cycle are likely not to do much better in the next one, unless their innovation cycles differ from that of their competitors.

This was the case with the NV30: it was an evolutionary refresh of the NV20, arch-wise (don't tempt me to go into the details...) while R300 was a revolutionary Tabula Rasa. This was not as obviously the case with 3DFX regarding VSA-100 vs NV10. In the case of AMD, the cycles do diverge: Brisbane was a small change, while Conroe was a large change. And in the previous 'large changes' cycle, AMD did extremely well for a variety of reasons.
That trend might hold for competitors that are close to being evenly matched.

AMD's good times basically coincided with whenever Intel screwed up, and Intel's bad times were usually much greater than AMD's best times.

Even in the depths of the Prescott fiasco, Intel's mobile fortunes were splendid.
When Intel was screwing up with one architecture, it was basically raking in cash with another architecture.

The question, based on this, is whether AMD will be able to survive (in its current form) until 2H09 or so, since I would be extremely surprised if the company became profitable again before then, no matter what the management team is claiming. Can they significantly reduce their quarterly loss? Sure. But actually turning it around and having a positive cash flow seems like a much harder thing to me.
Makes me wonder what AMD's going to sell off to save some pocket change.
It is getting to that point.
Shares in Spansion, real estate, various divisions that didn't quite have the right "synergies"...
 
Makes me wonder what AMD's going to sell off to save some pocket change.
It is getting to that point.
Shares in Spansion, real estate, various divisions that didn't quite have the right "synergies"...
Well, according to AMD's CFO...
http://seekingalpha.com/article/41686 said:
In addition, we did not receive any cash for the sale of our 200-millimeter tools, our Spansion investment, or from the sale of administrative land and buildings in the quarter. These items represent in excess of $800 million of cash generating opportunities in the future.

At the rate AMD is burning cash though, that's only 1-1.5 quarters of relief, AFAICT. The rumours regarding Fusion at TSMC are certainly getting interesting though. This might actually be a pretty good fab-light strategy: keep doing business as usual for the 45nm generation, except that Fusion is a single-chip solution manufactured at TSMC. But that doesn't really collide with the rumours of the initial Fusion platform being a MCM.
 
At the rate AMD is burning cash though, that's only 1-1.5 quarters of relief, AFAICT. The rumours regarding Fusion at TSMC are certainly getting interesting though. This might actually be a pretty good fab-light strategy: keep doing business as usual for the 45nm generation, except that Fusion is a single-chip solution manufactured at TSMC. But that doesn't really collide with the rumours of the initial Fusion platform being a MCM.

Aside from the uncertain percentage of AMD's x86 production that can be fabbed at TSMC, the implication of a single-chip Fusion on TSMC's 45nm node is that it would be a non-SOI x86.

It seems unlikely that there is a parallel K8 design for bulk SI that has never been manufactured.
Either they're going back to refactor K8, or *gasp* we're looking at a reprise of K7 (Geode?).
Using Geode like that would cement Fusion's targeting of the low-end, but transitioning the design to 45nm would need significant reworking as well, given how far back the Geode NX is process-wise.

Maybe they'd retool the K7 design to feature some of K8's design changes, a sort of K8 done all over again (or counting Griffin, again-again).
 
Well, what are the alternatives? Either the analysts are wrong, or they do not realize Fusion might be a MCM and it's just the GPU being manufactured at TSMC, or the process at TSMC is actually going to be SOI-based.

The latter is far from impossible, but then the timeframes don't make a lot of sense. I won't try to comment on how likely it is anyway; I am obviously biased there, since I'd like SOI to become more common in order to make Z-RAM a more viable alternative, with all that entails. heh.
 
I'd bet on it being a simpler MCM solution, with the K10-family CPU die packaged together with a 790G IGP.
Perhaps that's where the TSMC rumor comes from, since this last portion of the package is made at their factories.

I always had a hard time swallowing the marketing speak of a revolutionary approach in such a short amount of time, when both the K10 and 690/790 projects were already well under way, with independent management and R&D teams.
If the time frame was longer, i could have bought it as the real thing, but not like this.
 
Back
Top