Premium pricing for SMIC

Massive order book secures pricing right at the top of the range.

As expected, Semiconductor Manufacturing International Company (SMIC) achieved its aggressive pricing expectations on Friday, completing an upsized $1.8 billion IPO at the top end of the deal's indicative range. Joint bookrunners Credit Suisse First Boston and Deutsche Bank completed the world's third largest IPO of the year at HK$2.72 per share and $17.50 per ADS.

About 60% of the 5.15 billion share deal comprised primary shares (3.03 billion) and 40% secondary shares (2.12 billion). The 772 million share greenshoe will also be all secondary shares and on full completion, the company will have a freefloat of 31.8%.

At this level, SMIC has priced at 2.3 times 2004 book. At the time of pricing, TSMC was trading at 3.3 times 2004 book, UMC at 1.9 times and Chartered Semiconductor at 1.5 times (all UBS estimates). This means SMIC has come at a 30% discount to TSMC, but a 20% premium to UMC and an even more staggering 53% premium to Chartered.

One measure of SMIC's success can be seen in the context of the valuation it achieved relative to Chartered Semiconductor, which listed in Singapore and New York in October 1999. In absolute terms, the two valuations are very similar since tech bankers say Chartered was priced at about 2.2 times book.

However, on a relative basis, Chartered priced at a huge discount to TSMC and UMC, which were both trading at peak historical valuations in late 1999. According to Goldman Sachs research figures (see table 1), TSMC was trading at over eight times book and UMC over five times.

In 1999, Chartered had also ramped up production to about one third the output of its two Taiwanese rivals. At the end of that year, for example, it was producing 688,000 chips compared to TSMC's 1.9 million and UMC's 1.8 million.

By contrast, SMIC stands at an earlier stage of development, producing 153,000 chips at the end of 2003 relative to Chartered's 1.19 million, UMC's 2.6 million and TSMC's 4.01 million.

But the key driver for SMIC's valuation is where investors believe the tech cycle currently stands, since the three other foundries are all still trading below their historical averages. In late 1999, the foundry cycle was peaking, whereas in early 2004, most analysts believe it is still in recovery phase. And sector specialists say that of the four big Asian foundries, SMIC offers investors the most leveraged exposure to the sector.

"If the equity markets re-bound from last week's correction and the tech cycle moves onto a sustainable upswing, then investors stand to make a lot of money," says one. "SMIC is growing faster than the other foundries and from a much lower base. As a result all its earnings multiples should expand a lot faster."

However, if the markets are facing a more prolonged correction and the recent China equity bubble has burst, then many believe SMIC may have some difficulty holding its share price, at least over the short-term. Over the medium term, investors will also face overhang from the large number of seed investors that still want to cash out (roughly 40% of enlarged share capital), not to mention the constant threat of dilution given the sector's intense capital expenditure requirements.

But others believe the huge order book should provide ample buffer even if most of the institutional demand was vastly inflated and a lot of retail interest gets nervous and evaporates over the next few days. The institutional order book closed 20 times oversubscribed with demand of $30 billion and the Hong Kong retail IPO 277 times oversubscribed.

Because the Hong Kong IPO secured an oversubscription rate of more than 100 times, full clawbacks were triggered and retail investors have been given a final allocation of 20%. Of the remaining 80% of the deal, institutions have been allocated about 89% of the total, the Japanese POWL (Public Offer Without Listing) about 7% and US retail about 5%.

Demand for the POWL, led by Nomura, is said to have closed around the $1.6 billion level, with virtually no price sensitivity. The same was also true of the institutional book, which had over 700 accounts and 15 orders for more than $150 million. One investor is said to have placed an order for $500 million.

By demand, the institutional book was pretty evenly split between the three regions, but allocations were skewed away from Asia, where a lot of private banking demand (15% to 20% of the total) and momentum driven orders were coming from. The US received a final allocation of about 45%, with the remainder split equally between Asia and Europe.

Having spent 14 days enduring an exhausting schedule of one-on-one meetings, SMIC is said to have achieved a conversion rate of over 90%. Specialists say many of these accounts were global tech funds that hold the other foundries rather than the more traditional Asian IPO funds.

One major quibble was why SMIC priced right at the very top of the range when it could have left a few cents on the table and signaled some good faith to investors after four days of difficult equity markets. However, as one observer points out, it is also more difficult to make such a symbolic gesture when dealing with such a large proportion of selling shareholders that want to maximise proceeds above all else.

US equity markets also recovered slightly on Friday after four consecutive days of losses. But company supporters hope investors' pre-occupation with market momentum will not detract from SMIC's huge achievement.

The Chinese government has placed great strategic emphasis on the foundry sector and Richard Chang's company has moved up the technology chain and into profitability over a very short period of time. A number of analysts conclude that the main loser is likely to be Chartered Semiconductor rather than sector leader TSMC, which has the economies of scale, a wide-ranging IP library and client base, plus a Chinese fab of its own with which to support its margins.

In a recent research report, UBS estimated that SMIC is about 9 to 12 months behind TSMC in terms of process technology. As of the fourth quarter 2003, 72% of SMIC's wafer output was logic chips and 28% DRAM. At the end of the 2003, it reported that 10.4% of its sales were wafers of 0.13 micron and plans to move to 90nm by 2005.

As a comparative analysis, UBS said that, "SMIC was approximately 18% the size of TSMC at end 2003. If its capacity plan moves forward according to its schedule, this ratio would move up to 33% by end 2005."

And it concluded that, "If SMIC expands as planned it would be 22% larger in terms of capacity than Chartered."

Trading will begin in New York on March 17 and Hong Kong the following day.