Although it gained only a couple points of market share, AMD's gains are greater when measured in dollars because the company charges more for Athlon than for its older K6 processors. Taking advantage of a weak point, AMD caught its archrival at the tail end of a product cycle. Intel tried to counter with its aging Pentium III design, but that chip barely made it to 1GHz even as Athlon soared as high as 1.2GHz.
Intel subsequently brought out the faster Pentium 4 at the end of last year, but that chip remains expensive and available only in limited quantities. The initial Pentium 4, code-named Willamette, has a relatively large die size, which increases its cost and limits production. Furthermore, Pentium 4 is available only with Rambus memory, which is also expensive and in short supply. In brief, all this has made it difficult for Intel to position its fast processor against Athlon.
Ramping up P4 and beyond
The problem should be alleviated over the course of this year, but it will still take another year before the Pentium 4 really hits its stride. First, Intel is increasing capacity in the 0.18-micron fabrication plants that can build Pentium 4. In fact, the shortfall in PC sales is actually accelerating the transition to Pentium 4. This is because the fewer chips Intel needs to build, the more capacity it has left over for the large Willamette die.
This summer, Intel is expected to introduce a new chipset, code-named Brookdale, which combines Pentium 4 with low-cost SDRAM memory. While SDRAM will reduce the cost of Pentium 4 systems (compared with Rambus memory), it also reduces system performance significantly. Some consumers may not notice or care, but others will avoid the SDRAM boxes.
In the fall, Intel plans to roll out a 0.13-micron version of the Pentium 4 known as Northwood. In this next-generation process, Northwood will be a fraction of the size of Willamette. This change will reduce Intel's cost to build the Pentium 4 and will allow the company to greatly increase production levels. In addition, Northwood will push the Pentium 4's clock speed to 2GHz and beyond.
Finally, in early 2002, Intel plans to upgrade Brookdale to support DDR SDRAM. This memory combines high performance with low cost. Northwood systems with DDR SDRAM will offer price and performance well beyond what AMD will have. This combination should allow Intel to take back some of the market share it has lost to AMD.
Intel is laying the groundwork to gain market share with a huge investment in manufacturing facilities. The company plans $7.5 billion in capital expenditures this year and has not trimmed these plans in the face of falling revenue. This investment is crucial for the transition to 0.13-micron manufacturing and to 300-millimeter wafers. As noted above, 0.13 micron offers both lower cost and higher speed, an irresistible combination.
The larger 300-millimeter wafers cut manufacturing costs by an additional 30 percent or so. Intel underestimated PC demand for 1999 and early 2000. In 2002, the company does not want to be in a similar position and leave market share on the table. The huge cap-ex plan is critical to ensure that capacity is not a limiting factor in 2002 and beyond. With a six- to 12-month lead over AMD and others in both 0.13-micron and 300-millimeter wafers, Intel will be the lowest-cost producer, enhancing its gross margins and affording the company an edge in any price war.
Looking ahead to 2002
Intel's 2002 shipments will also be boosted by Microsoft's Xbox, which is expected to debut late this year. If Xbox is successful, shipments could easily exceed 10 million in 2002. These units will use a low-cost version of Pentium III, reducing the revenue impact, but Xbox could still add a nice bump to Intel's revenue.
Even without the economic downturn, 2001 was shaping up to be a lackluster year for Intel. We knew that the large Willamette die would reduce Intel's margins and hinder its ability to increase shipments. The technology transitions from Willamette/Rambus to Willamette/SDRAM to Northwood/SDRAM to Northwood/DDR will tax the resources of many PC makers, further inhibiting Pentium 4 sales.
But by the start of next year, Intel will be pumping out Northwoods at 2-plus GHz and will have a cost and performance advantage over AMD and plenty of manufacturing capacity. If the general economy turns around by that time, Intel will be ready for a rapid turnaround.
Besides the supply-side fix there are other potential drivers of demand. Among them are the W2K upgrade cycle and the secular trend toward notebooks in the commercial market. In addition, the launch later this year of Windows XP Home Edition, which blends the stability and reliability of W2K with the consumer-oriented features of the Windows 9x line, should resuscitate consumer demand. However, we do not expect the macro environment and the technology cycle to align favorably until 2002.
With Intel shares down more than 60 percent from their 52-week highs, is it too late to sell? We have seen with Intel and other technology companies that the E in P/E has fallen at a faster rate. If demand does not pick up in the second half of 2001, it is likely that Intel will retest the valuation lows of 1998, taking the stock down to the high teens. When we do come out on the other side of this bear market, investors should be in a position to "buy" and not "sell on rallies."