Thursday, May 6, 2010

Thrive or survive…going for gold in post-recession recovery: Malcolm Penn @ IEF2010, Dresden

According to Malcom Penn, chairman and CEO, Future Horizons, 2010 — a barnstroming year — will likely see the global semiconductor industry grow by 31+ percent. He was delivering the company’s forecast at the ongoing 19th International Electronics Forum (IEF) 2010 in Dresden, Germany, which ends here tomorrow. He said it would take a disaster of the scale of Lehmann Brothers to derail this now!

Some of the other forecasts made by Malcolm Penn include:
* 2011: +28 percent; based on: peak of the structural cyclical boom (could stretch into 2012).
* 2012: +18 percent; based on: normal cyclical trash cycle starting 2H-2012 (1H-2013?).
* 2013: +3 percent based on: market correction in full flow (could be negative, cap ex overspend and inventory build depending).
* 2014: +12 percent; based on: start of the next cyclical recovery (single digit, if 2013 is negative).

Given the now unavoidable 2010-11 fab shortage, the growth upside for 2010-12 is huge!

The forecast track record of Future Horizons is quite interesting. As per forecasts made during the IFS2010 in Jan.2010, the chip fundamentals was said to be in very good shape. The industry was starting its recovery with shortages. Also, the ASPs had already stopped faling. The inventory levels were at an all-tme low. Finally, the capacity was tight, and spending, weak!

All of this added up to two years of very strong growth in prospect. Penn had said: “It doesn’t get much better than this. But, despite what the numbers say, still no-one believes beyond the next quarter! “Ah but” is still driving the industry consensus!

Industry fundamentals don’t lie — believe in them or die! The capacity famine was instigated two+ years ago — well before the crasj, today’s shortage was inevitable. The recovery dynamics will continue to strengthen. Future Horizons’ forecast is now +31 percent ~$300 billion. The next trash dynamic has still not yet triggered. It is unlikely to happen before 2011, meaning, 2012 impact. However, the economic uncertainty remains the biggest risk. Also, the global financial system is fundamentally flawed.

Current industry status and outlook
There is said to be underlying fear, uncertainty and doubt. Some of the reasons are:
* Is unit demand overheating or sustainable?
* Is inventory starting to get out of control?
* Will the economy slip back into recession?
* Is the capacity crunch a blip or more fundamental?
* What will happen to the economy when the stimulus funding dtops?
* Will the end market demand hold up or slip back?

All of these have been going around, largely due to the loss of collective confidence, after 'five bad years' of industry growth.

According to Penn, risk aversion is not risk management. As recessions draw to a close, few are ready to believe that it is over, exaggerating the catch-up reactions, amplifying the cycle peaks and troughs. He cautioned that fears of a 2H-2010 dip will exacerbate the next trash cycle. By the time the industry waits for 2010 clarity (September?), it could be too late to rescue 2011. Well, if you really can't stand the semiconductor heat, it is best for you to maybe, consider a career change!

Now let's analyze the industry fundamentals -- economy, unit demand, fab capacity and ASPs, respectively.

Regarding the economy, the industry had entered the recession in structurally good shape. The 2009 chip market eas the victim, not the cause of the recession. While the chip market depends on the economy, it marches to its own drum. It must be pointed out that the chip market actually recovered much faster than the economy. Asia, and not the USA, is now driving the global GDP growth. Also, business is set to replace consumer as the global growth driver.

Forecast health warning #1: Economic disruption will derail the chip market. The only questions are: "by how much and for how long?"

With regard to the unit demand, riding the IC unit shipments trends wave takes judgement. The monthly run rate varies dramatically from the trend line. It is impossible to balance supply with demand (demand changes in days, supply changes takes months). A mismatch makes it ‘feel’ that capacity expansion is out of control.

Forecast health warning #2: Need to watch inventory plus double ordering in tight demand cycles.

Turning attention to fab capacity, the long-term wafer supply security Is said to be fundamental. It is the fundamental fabless (Fablite) achilles heel, and the reason that FSA (now GSA) was formed in 1994.

Kit ordered today equals units out one year later. Then, there’s the ramp up time The next four quarter’s capacity is cast in stone. Even if we splurged cap ex today, there will be no new sales impact until May 2011. Therefore, 'sell the kit today' means capacity two quarters later.

On the semiconductor equipment sales trends, there have been two years of under investment. No amount of productivity gains, second-hand equipment sales, cannibalisation, or other ‘tricks’ will compensate for this decline.

Looking at the front end book-to-bill, equipment orders placed today equals new IC sales four quarters later. The Q1 cap ex growth won't impact until Q1-11. The front end book-to-bill has never been so slow for so long.

As for the cap ex spend and book-to-bill impact, the 2009 cap ex was down 48 percent on 2008 (down 31 percent on 2007). The 2010 cap ex is up 80 percent on 2009 (which is back to 2008's level). The 2010-11 capacity is condemned to leading edge famine.

Looking at the medium-term new capacity outlook, there is still no serious capacity build out yet in prospect!

If you look at the global MOS wafer fab capacity, Q1-09 was -8 percent, the biggest quarterly capacity fall in semiconductor history. Q2-09 was not much better at a further -2.7 percent. Q3-09 ‘bottomed’ at -0.7 percent. The Q3-09 capacity was down 12.5 percent on Q3-08’s peak and not increasing. Also, the Q3-09 demand was down only 10.6 percent, but recovering.

Comparing the capital spend and semiconductor market, the cap ex was well below 20 percent trend line (2008 = 12 percent/2009 = 7 percent). Even the 80 percent cap ex growth in 2010 is too low (still only 11-12 percent).

Looking at the MOS capacity build out by wafer size, there has been no change in volume ramp profile. The MOS capacity mix by feature size is 35 percent of the capacity Is <55nm vs. 18 percent this time last year (54 percent 65nm and under).

On the supply/demand balance, he said that during 2010, new capacity addition will be minimal. The era of cheap and readily available wafers is over. Excess capacity had zeroed out in Q4-2009, exactly as forecast!

Forecast health warning #3: Need to watch front end cap ex spending in 2011. Excess capacity will kill the market stone dead!

Finally, turning to ASPs, Penn said these are the least understood industry wild card. The ASPs can’t really keep falling forever. He listed five factors, all of whom having already run their course. These were:
* 130nm yield bust … destroying the node’s ASP price enhancement.
* 300mm wafer transition.
* Two to three year memory price war -- a result of 300mm conversion.
* Brutal 32-bit MPU price war -- ASPs fell from $100 to $70.
* The overall price pressure due to excess capacity.

According to Penn, there is lots of structural ASP recovery potential.

Forecast health warning #4: ASPs are diagnostic and Complex -- systemic (Moore's Law); structural (Capacity/Wafer Size); and sabotage (price wars).

Industry outlook
Coming to the outlook for 2010, according to Penn the four industry fundamentals are all positively aligned. We have a recovering global economy, there is a strong unit demand with no excess inventory, ASPs are in the early stage of recovery, while there is still tight fab capacity and past underinvestment to deal with. He added that 2010 could well be the (inverse) perfect storm!

Hence, Future Horizons forecasts 2010 to be 31+ percent -- up from Jan. 09's 22 percent number! Only a monumental disaster can now derail the recovery!

Key market drivers
On market drivers, Penn first touched upon the applications market. The good news is that demand is holding up well. Also, PCs and smartphones are booming. The really good news is that there are loads of emerging applications on the horizon. The not so good news is that automotive and (some) consumer are still slow.

Strong memory
The 2010 memory market outlook also looks good. DRAM GBs are up 45-50 percent, NAND GBs are up 80-90 percent, and ASPs are trending up as well. There was a strong Q1, followed by a record Q2. There was no sign of 2H-2010 weakness. Penn added that memory sales could top $75 billion during 2010. Similarly, after-tax profits are likely to be in the range of $20-25 billion. A tight market is expected well into 2011.

The memory cap ex has resumed. It was $23.7 billion in 2006,$31.5 billion in 2007, $20.7 billion in 2008 and $6.3 billion in 2009. (~2/3 DRAM). There is a huge DRAM cap ex cost to convert to 50nm and better process. During 2010, the DRAM market will grow from $22.4 billion to $45.7 billion. The NAND vendors have so far shown uncharacteristically modest investment plans. The NAND market will grow from $13.9 billion to $22 billion during 2010.

Industry issues
Finally, turning to the latest industry issues, Penn said that there were downside risks to trhe world economic outlook. The financial community lessons have clearly not yet been curbed.

There are also certain perils of dancing to the tune of the Wall Street. The Street's 'one size fits all' semiconductor five-point plan is:
* Become specialised -- increases your commercial risk.
* Go Fab Lite -- lose your competitive differentiator.
* Merge with each other -- yet, few are successful.
* Narrow scope of R&D -- limits tomorrow's opportunities.
* Cull the product line -- restricts market opportunities.

Plus, outsource everything for best operational efficiency!

Then, there is also the fabless/fablite extrapolation. What if the wafer prices increased, or wafers were unavailable or you are not front of the line? It is one thing having to pay a higher price for your wafers, not getting the wafers, getting them late or getting your new part to market six months after your competitor is another matter entirely.

Fabless/fablite alternative
So, what's the fabless/fablite alternative? Well, there is a non-fablite IDM alternative. It is recommended to build a joint fab with a consortium of competitors. Run and operated properly, it combines the advantages of a foundry, along with the benefits of an IDM.

There are also design and foundry challenges to deal with at 16nm. That is: how to design and deliver reliable ICs and system solutions out of increasingly variable and unreliable components at a reasonable cost and design effort? There will be issues like:
* Increasing transistor variability.
* Decreasing signal-to-noise ratio.
* Susceptibility to manufacturing faults.
* Advent of carbon chip technology?

Given the now unavoidable 2010-11 fab shortage, the growth upside for 2010-12 is huge! The cycles are dead! Long live the cycles!

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.