Investment Options

The semiconductor manufacturing market is beginning to resemble a derivatives trading scheme.

popularity

It’s clear that something fundamental has changed in the semiconductor manufacturing industry. What’s less clear is how this will play out over the long term.

Intel’s agreement to invest more than $4 billion in ASML to ensure the continued development of EUV and 450mm wafer technology is more than just a one-off deal. It’s a very public recognition that the astronomical cost of design and manufacturing for just a few customers has far outpaced any possibility of an adequate return on investment. That applies to equipment, to EDA tools, and to foundry processes.

We’ve been witnessing this trend unfolding for the past several process nodes. The number of chipmakers that can afford to stay on the Moore’s Law road map has dwindled significantly. That distinction may become fuzzier as the industry moves to stacked die, subsystems and platforms, where a 130nm analog chip may be bolted on a 14nm processor platform with 20nm memory. Still, the number of companies actually developing chips at 14nm will continue to shrink. It’s simply too expensive, and with markets for devices becoming ever-more fragmented, the sales volume required to generate reasonable profits is almost impossible to attain with planar devices.

Intel is looking to cut some of its costs with bigger wafers, which is why it’s subsidizing that effort. And almost all the major companies have opened their doors to third-party customers for their fabs, which is basically the equivalent of a homeowner taking in boarders to offset renovation expenses. IBM adopted a similar strategy a decade ago in research, and AMD spun off GlobalFoundries because it needed to generate cash in its fight against Intel.

But this is just the start of open investment in companies. ASML and UMC are both looking for additional investors, and most of those investments are expected to come from customers that need certain technologies that are suddenly non-mainstream. In effect, they are turning themselves into derivatives options, and if you’ve been reading the general financial press on this subject over the last four years you’ll realize quickly this adds a new kind of uncertainty into the market.

EDA tool vendors, at least so far, are playing it safe. At a time when many of their customers internally developed tools are running out of steam and there is an opportunity for taking over large chunks of that market with commercial tools, EDA vendors have chosen to limit their exposure. There are simply too few big customers, too many uncertainties, and the rates of return are too low to place big bets on technology approaches that may not work. Will it be EUV, DSA, gridded design, double patterning, multi-patterning or stacked die, for example?

At the moment, the best return for EDA companies appears to be from better integration of their own tools, as well as investments in complementary adjacent markets. Really pushing the limit for a few customers that may do things one way versus another is well beyond what constitutes good business sense, even if there is a possibility for a big payoff someday.

So will other companies begin pitching in on this development the way they have with the capital equipment vendors? That remains to be seen. Risk is being evaluated and re-evaluated everywhere, and for the first time it’s really beginning to impact the advancement of new technology that is commercially viable rather than just laboratory-proven.

Moore’s Law, which has dominated semiconductor design and manufacturing for the past five decades, is approaching its practical end of life. What comes next is less obvious, but what is clear is that it will involve radically different economics, business models and potentially huge changes in technology. And what we’ve viewed as discrete companies might well be viewed as parts of portfolios in the future, where risk and reward are balanced, distributed and much harder to gauge.