“This is a serious problem,” said Wally Rhines, chairman and CEO of Mentor Graphics, “especially if you are looking to invest or develop new products.”
The figures for 2016 underline Rhines’ point. Experts have had to revise forecasts made only last year downwards; from predicted growth of 3 to 3.5% to a contraction of 0.8%.
So, why is it so difficult to predict what is happening with any accuracy?
According to Rhines: “Despite the semiconductor industry’s maturity, it remains very volatile and revenue forecasting combines two variables – the unit volume and average price – which can change rapidly.”
Speaking in Paris, Rhines said the industry was not only vulnerable to broader economic trends, but also to industry specific challenges. “It is vulnerable at both the macro and micro level,” he suggested.
IC unit demand is, however, predictable. “Unit, volume and revenue per unit area have remained pretty constant, making it possible to make accurate linear predictions.
“Where there is a dislocation,” Rhines continued, “is when it comes to calculating revenue growth. Why? Because companies like Apple and Samsung are now developing their own custom processors. Those sales disappear from the industry figures and are not included. It’s a significant and growing trend and it is a major source of error in the figures.”
Is there a more reliable way in which to forecast future growth in semiconductor markets? Rhines believes so and his solution is based on the Gompertz Curve.
“Devised by Benjamin Gompertz in 1825, it has proven to be remarkably accurate in predicting future growth trends – whether that’s mobile phone uptake or population growth. The equation comprises of a double exponential with three coefficients and, when plotted, looks like an S curve. Those coefficients can be modified, but the equation will always plot an S,” Rhines explains.
By using it, it is possible to identify a product’s life-cycle based on constant unit growth.
“According to the equation, growth accelerates up to a mid-point then, at 35 to 40% of life time volumes, growth slows. We’ve used it to plot 40 different markets and it has proved remarkably accurate.”
The equation has shown that products that may have shifted millions of units have still to enter a period of decline.
“The solid state image sensor market, for example, is mature, but, according to the Gompertz equation, remains in its growth phase.”
Looking to the future, it is possible to predict IoT life cycles and to identify emerging opportunities from smart meters to wearables.
“If you can identify the applications then you will be able to predict future growth more accurately,” explained Rhines. “The three major areas of IoT semiconductor growth going forward include data centres, gateways and IoT nodes.”
Crucially, the equation suggests that demand for silicon transistors is expected to remain strong.
“When we look at unit growth rates, peak acceleration is a long way off,” said Rhines, “and while we may have to start looking at a replacement for silicon transistors, the silicon transistor as we know it, will continue to grow rapidly.”