Wednesday, March 23, 2011

The capex derivative for ICs (Part III)

Last week, I reviewed how solar charts can point up and up, yet the sales of lasers into solar fab tools can languish because they go as the derivative, not directly with sales. In this post, I will show how the same thing happens in semiconductor chips. And for that matter, in any capital equipment business.

The shipments of the end product made from semiconductor fab tools tend to go up and up (the Great Recession notwithstanding) because the world keeps getting bigger.  The installed base of tools tracks that trend.  (The installed base and chip revenues don't completely march in step, since installed equipment can sit idle, or chip prices can fluctuate.)  But the shipments of new tools tracks the 1st derivative of the installed base--you only ship new tools to add capacity or upgrade dated equipment.  The laser sales track this trend--the 1st derivative.  (It's actually the 2nd derivative of the revenues generated by the electronics, but that's not important here.)

That's shown in the figure below, using actual data for the semiconductor industry over the last several years. The installed base (in units of 10 million 200-mm equivalent wafer starts per month--got that?) ramps up and up. The current recession was an exception, when so many companies closed fabs that the installed base actually declined. But that's rare.

Source: SEMI

While the end product shipments grew and grew, the capex spending itself oscillated dramatically during that time.  While the capex business is a sizable business of its own, it isn't really growing so much as it's cyclic. Let that be a lesson.

Of course, we could make similar charts for displays, data storage, and any capital equipment business you like.

The important thing to remember is that the equipment shipments don't scale with the production, they go as the derivative. That's how component sales can languish even as forecasts for a downstream product go up and up.

Tuesday, March 15, 2011

The 2nd Derivative Capex Paradox Update

Jay Liebowitz recently commented on my Second Derivative Paradox where I explain how capital equipment markets can gyrate wildly even though the end-use product grows steadily.  Among other things, he points out that equipment sales can surge even when there is excess capacity because companies need to retool for new products.  This can happen because of the steady progress of technology: smaller via holes or more thin-film solar modules.  It can also happen if certain lines are specially qualified and others aren't. 

I can add that the inefficiencies of capitalism play in the equipment makers' favor: the churn in end-product manufacturers moves the manufacturing from company to company, creating shortage in new places and surplus in others.  So, even though a manufacturer has excess overall capacity, it may have to tool up a new line because that line has different requirements than its other ones.   

I am re-running the earlier posts below if you haven't seen them or can't link to them.  They are here and here.  I didn't update the solar numbers since my point is more conceptual.


How could equipment sales in an exponentially-growing market be anything but upward? It happens all the time. Welcome to the 2nd-Derivative Paradox. That's my name for the trap that one can fall into when it comes to capital equipment markets. Solar is a great example. It's hard to explain the paradox, though, so bear with me.

Start with installed capacity. If you are a power generator, you think in terms of the cumulative installed generating capacity in the world. This is what the users actually use. The figure shows three scenarios how that might play out, and they all look pretty much the same in this chart. Nice, steep slopes. Note how they all start at the same point and end up at the same point.

Then look at panel shipments. But the solar panel industry isn't interested in what's already out there. It needs to ship new panels every year. The shipments amount to a 1st derivative: the new capacity that's added to the infrastructure every year. Now the differences in the scenarios show through, as shown in the second figure. But the scenarios all show steep upward growth. What's to worry about?

Now look at panel manufacturing equipment. The solar manufacturing equipment industry, and that includes lasers--isn't even interested in solar shipments, but the need for more manufacturing capacity to make the panels. You only need more equipment when you are shipping more panels than before. That amounts to a 2nd derivative of the cumulative generating capacity, and can give wildly different results. New equipment is shipped in all three scenarios, but in the "sustaining" scenario the equipment shipments are flat year after year, while in the "slowing" scenario they start out strong, but then decline. Ouch.

Other traps. Of course we would all like to live in the "growing" scenario. The trouble is, strong positive exponential growth doesn't last indefinitely, no matter what they say. And that's not even considering some ups and downs along the way, like this year. A slight shift in the solar panel shipments wreaks total havoc for equipment shipments.

Other things that juice equipment sales. The same trap exists in other industries, too. But there are other details to consider. First, there is usually some churn in suppliers. Machines also get obsolete. And there is also the early obsolescence forced by things like Moore's Law. These all have to be considered.

Watch that 2nd derivative. Don't get me wrong. I love solar. I had a summer job at TI testing solar cells back in the Jimmy Carter era. We all believe it's going to be a great thing in coming decades. But it's not enough that the cumulative generating capacity will be on a steep upward slope for years to come, because when it comes to manufacturing equipment, it's the 2nd derivative that counts.

Some real numbers.  What happens when we plug in some numbers that may be more or less what we expect the solar market to be?

I’ve done that in this figure. The first thing to notice is that the cumulative generating capacity—the top curve and what the power companies think about—goes up all through the forecast.

The next thing you notice is that the new module shipments—that’s the middle curve—takes a dip in 2009. This isn’t too surprising, given the recession, tight credit, and low oil prices. The dip isn’t too big and it’s in record territory again by 2011.

But what is really interesting is the bottom curve. That’s the new factory capacity that’s needed to make the modules each year. This correlates directly to lasers sold for making cells. That curve actually goes to zero, even negative, for a couple of years. And even in the recovery it only hangs around the 2008 level through 2013. In other words, the laser sales will not rocket upwards like the module sales through 2013.

Of course, there are some problems with this simple chart. The new factory capacity (laser sales) probably don’t go negative. That would mean companies were taking equipment out of commission. While I have heard of this happening in 2009, it’s not widespread. Companies want to be ready for the recovery. And, there are always new suppliers, and old suppliers expanding and upgrading equipment. That raises sales above zero.

On the other hand, there is also inventory in the supply chain and used equipment for sale. That pushes the recovery further into the future.

To a first approximation, the chart is a good model, and a good example of what I call the "second derivative paradox." At least it’s better than looking at the other two curves and assuming something similar.