SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Technology Stocks : Intel Corporation (INTC) -- Ignore unavailable to you. Want to Upgrade?


To: Joseph Pareti who wrote (43530)1/2/1998 10:02:00 AM
From: gnuman  Read Replies (2) | Respond to of 186894
 
Joseph, re; AMD questions
I can't give an intelligent response on AMD's costs/yields. You might want to ask these questions of Paul Engel.
Lehman Bros believes that Intel and AMD will be 100% 0.25 by the end of 1998.
Everyones estimates for Intel 1997 earnings are $3.80, +/- a few cents.
Earnings estimates for 1998 range from $3.80 (Lehman Bros) to $4.11 (consensus). It should be noted that in early October these estimates were $4.35 and $4.69 respectively. These are pretty significant cuts.
My guess is that AMD can produce >8 Mil units of K6 this year, slightly less than 10% of total CPU shipments into PC's. Additionally, Cyrix/NSM will capture some additional share of market with their 6X86/mediaGX products. IDT will capture some portion of the very low end market, but I have no idea what share this represents. Throw in IBM and one can extrapolate that the competition has the capability to take >25% market share.
Lehman Bros projects Intel will come out of 1998 with 77% of Q4 units, AMD 14% and Cyrix/Others at 9%. Note that they forecast Intel will have 90% of dollars, representing the higher prices on the Intel chips. It's also apparent that Lehman see's AMD shipping more than 8 Mil units this year.
AMD's main threat is the significant price erosion in CPU's. Intel will remain the dominant supplier of PC chips, if for no other reason than their fantastic manufacturing capacity. So the real question for the investor is what impact will AMD have on forward looking margins? If recent history is any indicator, this may be significant.
While sub-zeros have captured the attention of the analysts, I think we need to look at PC prices across the board. Who would have imagined six months ago that HP would sell a fully featured high end PII 266 for $1799? Or that HP would be selling P55C 166 MMX for $899, with monitor and printer! (Todays CompUSA ad). I've seen PII 233's for $1299. And with the recent price cuts on PII, it wouldn't surprise me to see a sub-0 PII 233 in the near future.
While PC's have long been commodity products, it's only recently that CPU's have become commodities. This is a new experience for Intel. While AMD has long been a competitor, in the past they were always competing with the prior generation product.
I'd prefer not to publish my earnings estimates, better to look at the pro's on that.
The downgrades after the last earnings miss may be a clue. Hopefully we won't see any more misses. I think PUTS represent some cheap insurance.
As usual, these are my opinions and interpertations, and don't necessarily represent facts.
Regards, Gene



To: Joseph Pareti who wrote (43530)1/2/1998 2:59:00 PM
From: Time Traveler  Read Replies (3) | Respond to of 186894
 
Joseph,

AMD and Intel:

The cost breakdown of AMD and Intel is very difficult to obtain. A lot of information is deemed proprietary by both companies. We as outsiders can only guess/estimate through scattered data made available to us.

The following parameters are the most important:

- Cost per wafer in manufacturing from the beginning to the end
- Cost of packaging/handling/testing for each die
- Number of wafers per week of processing capacity
- Total number of dice on a given wafer that can be packaged
- Total number of good CPUs produced and perhaps sold
- ASP, the average selling price of each CPU
- Cost of goods (COG)

From Lehman Brothers' Q3 reports for AMD and Intel,

- Cost of K6 = $180.7M, of P-MMX = $1,325M, of P-II = $492M, and of P-Pro = $326M. This cost includes COG, M&A (marketing & administration), and R&D (research & development).
- For AMD (all together), COG = $324.2M, M&A = $100.9M, and R&D = $125.9M.
- For Intel (all together), COG = $2,045M, M&A = $676M, and R&D = $586M.

We then can go ahead calculate the percentage of COG from (COG + G&A + R&D). COG for K6 = $106.3M, for P-MMX = $819M, for P-II = $304M, and for P-Pro = $202M. What is this COG? The following model may be too simple but at least a start.

COG = 13 * wafers/week * (dice/wafer * cost/die + cost/wafer)

Now, we have to start guessing at the numbers, plug into our model, cross-check with existing data, and see if they all make sense. If not, we have to guess another set of numbers until everything seems fall into place. Just as Patient Engineer has found out, it is much easier to do this exercise on a Excel spreadsheet.

The high Intel COG indicates a high cost per wafer, in which implies a low wafer capacity for K6. Since K6 has more "advanced" manufacturing processes, Patient Engineer assessed K6 wafer at $400 higher than normal 0.35um (Intel). Also Intel has about 20K wafers per week of capacity. Say the x86 production is at 19K wafers per week. The socket 7 packaging should be between $5 to $20. Finally, the average ASP for P-MMX, P-II, and P-Pro mix should be about $210.

How many dice are on a wafer? The majority of folks on AMD thread insisted 140 dice per wafer for K6, but Paul counted 112 personally. Paul was probably looking at a pre-production wafer, so I have to go with the majority. With die size given, the number of dice per wafer of P-MMX is derived to be 178, P-II = 112, and P-Pro = 115.

I guess K6 wafer capacity to be at 1,800 per week and P-MMX at 14,000. Then solve a two-equation-two-unknown matrix. K6cost per wafer is $3,284, P-MMX is at $2,884, and Socket 7 packaging/handling/testing cost is $9. The number of wafers per week for P-II is guessed at 3,800, P-Pro at 1,200. Using the same cost per wafer for P-II and P-Pro, their packing cost come out to be $29 and $87 respectively.

In Q3, AMD shipped 1M K6, and Intel shipped over 22M x86 CPUs. I guess P-MMX at 18M, P-II at 3.8M, and P-Pro at 1.2M. Let's cross-check some numbers.

- The total yield of K6 = 31%, P-MMX = 71%, P-II = 55%, and P-Pro = 55%.
- ASP = revenue / total number. K6 = $150 (given), P-MMX = $171, P-II = $269, and P-Pro = $613. With the mix = $210.
- Cost of manufacture for K6 = $106, for P-MMX = $46, P-II = $80, and P-Pro = $168.

Again, these numbers did have a lot of guess work in them. You just have to play with them until they fit the existing data.

For Q4, Intel's number is a nightmare because of more mix of processes, namely 0.25um stuff. However, AMD is relatively easier to assess.

At 1,800 wafers per week, AMD is putting a lot of effort to ramp up this production to an eventual 5,000 of total capacity. Since AMD management is still very disappoint of its own production line this quarter, that indicates an about-the-same total yield for Q4 (31% I guessed/calculated) at 1.6M units. A 2,800 wafers per week for Q4 is thus reasonable. Given all other non-K6 business to stay the same and ASP for K6 to be $140 (generous here since it was $150 in Q3), a $52M loss before tax for Q4 is not far off to believe. In Q3, the dreadful IRS gave AMD more than $30M cash back due to $67M before tax loss. This quarter, the IRS may not be that generous.

The success of AMD will affect the bottom line of Intel. The more successful AMD is, the deeper the cuts Intel is going to on the Pentium line. Keep an eye on AMD. Although AMD is a threat to Intel, you have to sort through a lot of hypes. If Jerry Sanders (CEO of AMD) were an Intel CEO, Intel stock price would be $300 and up by now.

John.