FOOL CONFERENCE CALL SYNOPSIS*
By Debora Tidwell (MF Debit)

Advanced Micro Devices (NYSE: AMD)
One AMD Place
Sunnyvale, CA 94088-3453
http://www.amd.com/

UNION CITY, Ca., January 16, 1997/FOOLWIRE/ ---Advanced Micro Devices released their Q4 and fiscal 1996 results on Monday after the market close. AMD reported a net loss of $21,243,000 on sales of $496,868,000 for its fourth quarter. The loss amounted to $0.15 per share. Sales for the same period of 1995 were $595,178,000, which resulted in net income of $9,344,000, or $0.07 per share. In the immediate-prior quarter, AMD reported sales of $456,862,000, which resulted in a loss of $38,362,000, or $0.28 per share.

Revenues for 1996 amounted to $1,953,019,000, which resulted in a net loss of $68,950,000, or $0.51 per common share. In 1995 AMD reported revenues of $2,468,379,000, and net income of $216,326,000, or $1.57 per common share on a fully diluted basis.

BREAKDOWN OF SALES RESULTS. The computation products group (their Microsoft Windows compatible microprocessor group) was 22% of sales in the quarter, up from 18% in the prior quarter. The communications and components group was 66%, down from 69% in Q3 even though the sales in terms of dollars were up. Programmable logic was 12%, down from 13% in the prior quarter on relatively flat sales. Depreciation in the quarter was $84 million, almost exactly flat with the third quarter. Capital expenditures were $144 million in the quarter and $493 million for the year. This is slightly less than they expected but the difference will probably flow into the first half of 1997, so there are really no changes in their capital expenditure plans. Headcount at quarter end was 12,181, down about 200 from the end of the third quarter. International sales were 55% in the quarter, up from 53% in the previous quarter.

INCOME STATEMENT. Quarter-to-quarter, AMD had a sales growth of almost 9% or $40 million. Looking at the non-GAAP income statement, 80% of that or $32 million flowed through to the operating income line. AMD is very happy with that kind of flow-through.

BALANCE SHEET. Cash ended the year at $386 million, up $25 million from Q3. Inventory was down about $10 million in the quarter and for the year inventories were flat with the prior year. Accounts receivable were down both for the quarter and for the year. EBITDA for the quarter was approximately $35 million.

ESTIMATES FOR 1997. Depreciation for 1997 is estimated to be about $350 million and they don't have a better number than that right now. The tax rate for 1997 is anticipated to be 30%. Capital expenditures not including their Dresden plant (because it is funded separately) should be in the $400-450 million range because they have some carryover from 1996 into 1997. They have the capability to take down the $250 million term loan until the end of January and are the process of taking that down now. By the end of January they will have drawn down that $250 million. The interest expense will, therefore, be slightly higher on a run rate than the fourth quarter. While the majority of the costs for Fab 25 have been transferred into cost of sales, not all of them have, so there is a slight adjustment still to be made there. Going forward they don't believe we will see material changes in the R&D number.

PLD PRODUCT LINE. Sales were essentially flat with Q2. The story in PLD is the emergence of the MACH family as the primary driver. MACH bookings have been greater than the small density CMOS PALs for the last two quarters and were a record in Q4. MACH shipments for Q4 were over 50% of PLDs for the first time and AMD believes that their PLD division has commenced gaining market share and that they look more like a C-PLD (complex-PLD) company going forward.

COMMUNICATIONS PRODUCT LINE. Local area networking products, both Ethernet terminals (in the box, in the PCs) and hubs (the infrastructure of Ethernet) have both turned strong. Customers inventories, they believe, are in-line, the glut is over, and business has returned to it's growth path. In the public infrastructure (line cards), the SLIC and SLAC demand remains at record levels as it has for most of 1996. This particular product line was not affected by the inventory bubble that they saw in 1996. The cordless telephone is becoming significant as a design win for AMD's 900 mHz chip controller for the US market started to translate into shipments. The VTEC brand is an OEM brand for about 50% of the US marketplace.

NON-VOLATILE MEMORY (FLASH & EPROMs). EPROM sales were flat. Flash shipments year-to-year for the entire year were essentially flat at slightly over $500 million. Flash shipments in Q4 continued to grow at about the same rate as in Q3 (to about the $135 million level). AMD had records both in shipments and bookings. Prices reflect normal learning curve. AMD's single voltage power supply devices will remain at about 85% of their volume. The 2.7 volt only family continues to ramp for portables and mostly in the cellular phone marketplace. Finally, in non-volatile memory, they are commencing the conversion from 0.5 micron to 0.35 micron in their FASL wafer fab operation. They shipped during Q4, their first units of 16-meg flash on 0.35 micron. This gives them more capacity and lower cost. Business is improving, but the limitation is still the rate the market is growing, not AMD's ability to produce. They are not capacity constrained and, in fact, with their 2.7 volts they are seizing the market. All new 16-megabit wafer starts will use the 0.35 micron technology and over the next 12-14 months the whole line will convert over.

COMPUTER PRODUCTS GROUP. Sales were higher on a richer mix as well as increased volume, up 36% quarter over quarter. The unit volume of K5s more than doubled to 1.1 million units from 0.5 million units in Q3 with a richer mix as they shipped PR133, PR120, and PR150s for the first time during the quarter and this combination of performance ranges was the highest fraction of AMD's sales in Q4. Average selling prices in the K5s were up smartly in the quarter even though AMD took the prices down since the mix is getting richer. The largest dollar segment was the higher speed PR120, 133, and 150s. AMD is doing everything they can to get the optimum pricing, but since they have been bottom-feeding and competing with IBM and Cyrix for some of the PR133s and maybe in the future PR166s, they have had to offer product prices which, while higher than theirs, were substantially below Intel's. It's not really clear when that will abate. Fourth generation microprocessor shipments were in the 1.7 million unit range for a total Microsoft Windows compatible unit shipment of over 2.9 million units for the quarter and approximately 10 million units for the year. They have maintained this rate of 10 million over the last 5 years. With the PR166 announcement Monday on the K5s, AMD addresses solidly the mainstream desktop PC market. They have added one speed grade per quarter since introduction earlier in 1996 with PR75 and PR90 in Q2; PR100 in Q3; PR133, PR120, and PR150 in Q4; and the PR166 in Q1 1997.

STATUS ON K6. They continue to sample the 6th generation microprocessor code named K6 to customers. All of the customers who currently use the K5s are interested in using the K6. In some cases, yet to be announced, customers who want to use the K6 are going to be using the K5 as a segue to the K6. They are on their schedule to commence production this quarter with significant revenues in Q2. The K6 is a plug-in replacement for the Pentium Socket 7 infrastructure that gives AMD's customers time-to-market advantages because they can use their existing infrastructure to get to market with the K6. It lowers customers' development costs and will provide 6th generation performance within the existing infrastructure. The performance of the K6 is designed to be competitive with the forthcoming Pentium Pro Klamath and it is higher performance than the P55C. AMD has never seen a Klamath, so they can only speculate on performance. They are somewhat better than a Pentium Pro. They believe that a Pentium Pro Klamath will be lower performance because it doesn't have the advantage of the close coupling to the L2 cache. Relative to the best P55C at 200 mHz, AMD's K6 should be substantially and meaningfully better. Their real competition will be Klamath when it comes and at this point they don't know if Klamath will even be available before the Christmas season of 1997. One of the reasons for the superior performance of the K6 is the large cache. It also has two other benefits -- it allows them a small die size because memory is more dense than logic and it burns less power. This combination of parameters means that AMD appears to have a product which can be a 6th generation mobile product. They have a good candidate for mobile applications that will perform better than anything they have seen to-date. AMD demonstrated 200 mHz K6s running versus a Pentium Pro at Comdex and they have announced that the chip size is 162mm square. Once they move to the CS44 process, the chip will be less than half that size. AMD has a number of companies they are working with that will build chipsets that will run with the K6. The K6 was demonstrated with an Intel chipset at Comdex, so the K6 will run with any chipset that runs an Intel Pentium or P55C. For competitive reasons, AMD does not want to reveal at this point all of the features of the chipsets that their chipset partners are going to offer. Virtually all of the wafer starts at Fab 25 will be committed to K6 production and they are going to ramp production on the K6 as fast as they can. Fab 25 will probably not get up to its ultimate 6000 wafer production level before the end of 1998.

MICROPROCESSOR CUSTOMERS. Most of their customers do not want AMD to announce that they are using AMD's processors until they have actually announced the product. Monday, Acer announced another family with AMD products. There will be a number of the top 10-15 players announcing K5 usage this quarter. Most of them began ordering last quarter and some of them are ordering now. With the advent of the PR133 and now the PR166, AMD's customer base in the US is getting much richer.

MICROPROCESSOR SOCKET STRATEGIES. AMD wants to make it as easy as possible for the customers to use their products. All indications are that customers prefer the existing infrastructure. These new infrastructures are ways to get around product deficiencies rather than bringing a benefit to the user. So, it is AMD's intention at least through the 6th generation to say with the so-called Socket 7. At the present time, Intel's Pentium Pro and future generation bus structures are intended to exclude industry standards. AMD doesn't benefit by trying to play into those sockets. The industry standard is the Socket 7 which has a number of chipset makers, is well understood by the customer base, is lower cost, and is a cheaper implementation than the daughter cards and modules that Intel is promoting to make up for the deficiency in performance in their Pentium Pro and Pentium Pro Klamath that is forthcoming. AMD doesn't think they are disadvantaged by sticking with the Socket 7 design and look on the K6 as follows: it's smaller, it's faster, it's more energy efficient, it's easier to use, and it's going to sell at at least a 25% discount to Intel. AMD doesn't think that using a more expensive bus structure is going to make a difference. Moving forward to the K7, AMD could stay with the same socket, but then it turns out the bus structure getting in and out of the memory would be the limitation. So, AMD is working on an alternative bus structure to that. But there certainly still would be an opportunity for many years to come for the Socket 7. They don't think that the current generation is bus limited. They think they will have AGP capability coincident with Intel.

THE MMX TECHNOLOGY/TRADEMARK DISPUTE. There is no dispute with Intel over AMD's rights to MMX. There is no IP issue or dispute. The question is whether MMX is a trademarkable term or is it merely a generic term. AMD believes it is a generic term since MMX stands for Multimedia Extensions just as 386 was the numeric definition for the 386. AMD went through a battle with Intel over the 386 and was victorious -- the 386 is the generic term. They also believe that MMX is a generic term. There is no question that AMD will be providing MMX solutions -- the K6 is an MMX processor. They hope they can, on an amicable basis, resolve the issue with Intel because they want to promote MMX -- it's to Intel's benefit and to AMD's benefit. AMD does not see MMX as hype, they see it as a meaningful addition to the capabilities of a Windows-compatible microprocessor. The rate at which it will be adopted, in AMD's view, is the rate at which Intel converts their production. If the price is the same, why wouldn't people use it. It will run all the software. They think AMD and Intel will co-exist in 1997 and as AMD moves their production to K6, they will be moving all of their business to MMX. However, the rate at which the industry changes will be dependent on Intel's ability to convert their production from standard products to all MMX products and that is a question for Intel.

VIEWS ON THE INDUSTRY OVERALL. Relative to the industry overall, business is clearly dramatically better. AMD's bookings in the quarter just completed were better than the bookings they had a year ago. It is the first time AMD has had a book-to-bill ratio in excess of 1-to-1 since the second quarter of 1995. That booking strength continues. Inventories seem to have been worked down in the network areas and business was strong there. They think the inventory bubble has been absorbed and AMD expects that the outlook is quite bright for the industry. For the next year or so, AMD's expectations are 80-85 million PCs as far as the total market. They have been saying that 100 million units is the number for 1998, but they have to say that estimating 1998 at this point is pretty foolhardy. They think the 80-85 million units for 1997 is something they have some conviction on. AMD shipped 10 million processors in 1996.

SUMMARY. While AMD is pleased with their operational progress in the quarter, they believe that the most meaningful way to evaluate their financial performance is the non-GAAP consolidated statement of operations. In the fourth quarter AMD cut their operating loss in half compared to the September. To eliminate that operating loss, AMD must grow their quarterly revenues in the $540-550 million range. They can't predict if they will hit that number at this point, but bookings are strong and they are off to a good start. They are hopeful and it is certainly their desire to return to operational profitability as soon as possible. The company was asked, given that assumption of revenue, what they expect the gross margin to be under normal circumstances, non-GAAP. The company answered that the way they looked at it last quarter, they think the gross margin was in the 37% range. What happened last quarter they are very happy with -- 80% of the incremental sales flowed through to the operating income and certainly to the gross margin line because R&D and SG&A were flat. They advised calculating the estimate using something in the 60-80% range of the flow-through. The key to hitting the breakeven revenue level is increasing their Windows-compatible microprocessor sales. AMD's PC microprocessor fortunes and outlook are improving. The largest contributor to K5 revenues came from the higher speed parts -- K5-PR120s, 133s, and 150s. This acceptance of AMD's higher performance parts and the PR rating concept is gratifying. With Monday's announcement of the PR166, they continued to demonstrate their ability to execute their roadmap for the desktop. They plan to sample PR200 K5s in the current quarter and ship K5-PR200s for revenues next quarter. Most gratifying, however, is the progress they are making on the K6 MMX processor. They are continuing the sampling process. The development process is proceeding well and they believe all the significant issues have been identified and resolved. They are making the final revisions necessary to implement a production mask set. They anticipate no snags in the process and currently expect to commence production within the next several weeks. This is consistent with their long-held expectations that production would commence in the first quarter with significant revenues in Q2. They expect K6 unit volume to exceed K5s by the Q3-Q4 1997 timeframe and their goal is still to do $1 billion in revenues on the K6 this year. They do need to win some of the tier one players to get the ASPs and the high-end mix to get there. They are committed to the $1 billion goal and to revenues of $5 billion in 1998 which will be importantly dependent on K6 sales.

* A Fool conference call synopsis represents an effort to highlight the salient points of a conference call and should not be taken as an authoritative accounting or transcription of the entire event. Note: Statements made by a company other than historical information may constitute forward-looking statements for which the company can claim protection under the Safe Harbor Act. Please consult the company's filings with the SEC for information on risk factors which might cause actual results to differ materially from the information contained in these forward-looking statements.