News, views and contacts from the global Semiconductor industry

Capital Expenditure Excesses Curbed


1 September 2004


The massive boom/bust that was characteristic of the previous chip cycle looks unlikely to repeat itself this time around. Len Jelinek of iSuppli reports on the market dynamics at the heart of this change.


In January 2004, when revenue and profits for the chip industry exceeded all expectations, Wall Street semiconductor analysts remembered the boom years of 1999/2000. As they pondered market conditions, they wondered if new capacity would be available in time to meet soaring demand. They also speculated about whether the industry was headed for allocation and – if so – how long it might last.

Meanwhile, financial controllers at semi-conductor manufacturers harboured lingering fears from the aftermath of the previous chip boom. They looked to 2005 with trepidation and remembered 2001 when book-to-bill ratios plummeted to below 0.5:1, when product returns exceeded factory shipments and the term 'downsizing' dominated corporate vocabulary.

By the third quarter, both Wall Street and the chip manufacturers began to change their view. It became clear that 2004 was not a replay of 2000 and would not bring a return to the capital-spending excesses of that year and the subsequent severe downturn. In just three short quarters in 2004, the industry had shifted from insufficient supply to excess capacity.

So what factors have changed between 2000 and 2004 that have resulted in such dramatically different capital-spending scenarios? Significant improvements in yields at 300mm facilities, rising amounts of available capacity and changes in fab construction and equipping strategies have generated dramatically different market dynamics. It is important to examine these changes and how they affect conditions in the semiconductor industry in 2004 and beyond.

DEMAND

" Just as in the boom of 1999/2000, the expansion of 2003/04 brought record unit shipments for semiconductors.

Just as in the boom of 1999/2000, the expansion of 2003/04 brought record unit shipments for semiconductors. Figure 1 shows the quarterly semiconductor unit shipment figures from World Semiconductor Trade Statistics.

As presented, the pattern of semiconductor unit shipment growth from the first quarter through to the third quarter of 1999 was nearly identical to that during the period from the first quarter of 2003 to the second quarter of 2004. So why was the industry in a state of hard allocation in the fourth quarter of 2000 and not now?

CONTROLLED CAPITAL SPENDING

One of the causes of the change in market conditions between growth cycles has been a big change in capital-spending patterns. Figure 1 also shows worldwide capital expenditures as a percentage of global semiconductor revenue.

So what factors have changed between 2000 and 2004 that have resulted in such dramatically different capital-spending scenarios? Significant improvements in yields at 300mm facilities, rising amounts of available capacity and changes in fab construction and equipping strategies have generated dramatically different market dynamics. It is important to examine these changes and how they affect conditions in the semiconductor industry in 2004 and beyond.

DEMAND

Just as in the boom of 1999/2000, the expansion of 2003/04 brought record unit shipments for semiconductors. Figure 1 shows the quarterly semiconductor unit shipment figures from World Semiconductor Trade Statistics.

As presented, the pattern of semiconductor unit shipment growth from the first quarter through to the third quarter of 1999 was nearly identical to that during the period from the first quarter of 2003 to the second quarter of 2004. So why was the industry in a state of hard allocation in the fourth quarter of 2000 and not now?

CONTROLLED CAPITAL SPENDING

One of the causes of the change in market conditions between growth cycles has been a big change in capital-spending patterns. Figure 1 also shows worldwide capital expenditures as a percentage of global semiconductor revenue.

In periods of expansion, the percentage of capital expenditures to total revenue has exceeded 22%. Many analysts expected the industry to achieve a similar rate in 2004. But iSuppli predicts that capital spending as a percentage of revenue will not exceed 20% in 2004 or in the next few years. The area circled in red in Figure 1 shows how global semiconductor capital spending as a percentage of semiconductor revenue is declining.

CHEAP, USED EQUIPMENT

A major reason for this shift in spending is the extensive amount of used equipment that became available in the wake of the industry's dramatic downsizing in 2001. Chip makers that did not need cutting-edge process technology snapped up as much used equipment as they could, even if it meant putting it in storage for the near term. For start-up companies, such as the China-based foundries, the wide availability of this low-cost equipment reduced the traditional barriers to entry into the industry.

The flood of used equipment has been a major factor in reducing capital expenditures in 2004 – one that has often been overlooked. Such equipment is inexpensive, a factor that has reduced overall capital spending levels. Furthermore, when the industry returned to growth, this additional capacity minimised double booking, kept the industry out of hard allocation and suppressed typical panic pricing increases when lead times extended. The effect was a minimal increase in capital expenditures.

300MM: SLOW PROGRESS

Meanwhile, cutting-edge chip makers have been installing 300mm manufacturing capacity that uses advanced manufacturing technology. But the transition to leading-edge technology has not taken place as quickly as many chipmakers had expected. This resulted in a backward migration of technology into next-generation fabs.

" One of the causes of the change in market conditions between growth cycles has been a big change in capital-spending patterns.

Initially, most companies believed that 300mm fabs would use technology that required copper processing and 0.13µm or smaller feature sizes. The reality is that they have installed some aluminium technology at the 0.15µm process node. This also depressed capital spending to levels lower than historical trends would indicate. The transition to 300mm manufacturing has spurred a shift in how fab tools are installed.

Because of the high cost of equipment, which is estimated to be 75% to 80% of the total price of a fab, companies are completing modular construction of facilities and then equipping them in increments of about 6,000 wafers per month. As demand materialises, additional equipment is being installed. By extending the time to fully equip a fab, capital expenditures are made at a slower pace and equipment companies are better able to forecast their production. This has also suppressed panic pricing.

CAPACITY IMPROVEMENTS

Probably the most significant factor determining the level of capital expenditure is the availability of semiconductor die. A 300mm wafer generates a minimum of 2.5 times the number of good die as a 200mm wafer – every fully loaded 300mm fab equates to three 200mm fabs. The rise in die production lowers the need to make further equipment purchases.

There is also the issue of 300mm fab efficiency. Most 300mm fabs are still using a large portion of their capacity to make process monitor wafers. As they complete multiple cycles of learning, however, the ability of engineers to maintain processes within tight tolerances increases. This results in lower numbers of monitor wafers – the fab gains free capacity. The average 300mm logic fab is running close to 40% of its 300mm wafer capacity as monitor wafers. Another factor is the huge increase in die production using existing facilities and equipment, thereby reducing chip makers' need to buy more equipment.

SUM OF PARTS

Potential yield improvements, capacity, fab efficiency, tool installation rates and unit demand lead to the conclusion that the industry has changed how it spends capital to meet demand, and less spending is required to meet demand in 2004 than in 2000. With the wide availability of used equipment able to support trailing-edge capacity, the industry averted allocation and transitioned quickly from insufficient supply to overcapacity in 2004.

iSuppli anticipates that 300mm fabs will exceed present die yields on existing 200mm

lines during the next two years. The run rate of monitor wafers will decline to the 20% range, which is more in line with existing 200mm fabs. These events will result in each 300mm fab having the potential to output the equivalent die of three 200mm fabs. This will further reduce the capital required to support industry growth.