News & Analysis
Comment
peter.clarke
Warren
I am caused to wonder if the analysts' prediction error for May averages (that I ...
Analysts lag 'actual' chip market growth
Peter Clarke
7/5/2010 7:56 AM EDT
The record figures were ahead of the predictions made by analysts despite the fact that they display below-average month-on-month growth and look set to push predictions of annual growth for 2010 yet higher into the mid-30s of percentage points. All of this is must be tempered with the possibility that semiconductor sales may be supply constrained in the second half of the year and that if that should happen a usual consequence would be rising average selling prices.
Bruce Diesen, senior analyst with Carnegie Group (Oslo, Norway) predicted on June 24 that the three-month average of global chip sales for May would be called at $24.5 billion.
Meanwhile independent analyst Mike Cowan was predicting actual global sales of $22.743 billion in May equivalent to a three-month average of $24.227 billion.
Both were behind the actual and three-month figures. Indeed whereas the figures given by Cowan equate — under his market model — to an annual growth for the global semiconductor market in 2010 of 33.2 percent, the figures published by ESIA would equate to annual growth in 2010 of 34.6 percent.
And yet the figures predicted and returned were all behind one historical indicator. At $24.02 billion May's actual sales were just 2.7 percent ahead of April's actual sales of $23.39 billion. Over the last ten years the sequential growth for May's actual semiconductor sales has averaged 4.9 percent. With the best and worst years taken out of that averaging process May actual sales are, on average, 5.2 percent ahead of those in April.
June actual sales, coming at the end of the quarter, are normally very strong and have shown average sequential growth over the last ten years of 22.8 percent.
Related links and articles:
Lead times for some chips at 20 weeks, says iSuppli



Warren
7/5/2010 8:24 PM EDT
I am caused to wonder if the analysts' prediction error for May averages (that I calculate at 0.7% and 1.8%) is particularly unusual.
Sign in to Reply
peter.clarke
7/19/2010 12:41 PM EDT
The short answer is "no."
The market statistics on a monthly basis tend to swing around quite a lot. On a quarterly basis the wilder swings tend to get averaged out.
Sign in to Reply