What Went Wrong?
by Phil Dembowski of Dow Corning
When IBM introduced copper interconnects in 1997, it opened the gates to a new era in semiconductor manufacturing - one that is defined and paced by materials innovation. One of the industrys first forays into this new era came with the push for low-k dielectrics, the next "big" thing.
Amid the fanfare around the low-k opportunity, many materials developers dove into the challenge early, aiming to deliver ultra low-k materials (k<2.0) for volume manufacturing by 2003. Those plans were overly optimistic, to say the least. Todays best estimates dont see ultra low-k materials entering production before 2013, and many in the industry believe that a sub 2.0 dielectric will never be implemented. The experience sparked a wake-up call about the financial feasibility of semiconductor materials development. Most suppliers who invested in low-k early have not seen any revenue, and even those who have seen some revenue have not recovered their investments in full.
So what went wrong? A close examination of the low-k experience reveals a few fundamental lessons that should be considered as the industry embarks on future material development efforts.
1. Remember: incumbency rules
One of the unspoken laws in semiconductor manufacturing is that the industry will avoid change as long as performance permits. When it came to low-k materials, manufacturers found ways to work around integration concerns by moving instead to fluorinated silicate glass (FSG), which behaves a lot like SiO2, and designing around the need for low-k altogether.
Its an important lesson, for as long as the industry can make an existing technology work, the introduction of its successor will be delayed. (Look no further than optical lithography and 300mm wafers for evidence.)
With typical new business opportunities, cash flow delays of one to two years are manageable from a return-on-investment (ROI) perspective, but delays of six and seven years make it nearly impossible for a company to recoup investments, let alone earn a profit.
2. Get closer to your customers
Successful suppliers are those who go deep with customers to understand the full breadth and depth of their needs. It is important to strive for an open and honest dialogue between the material supplier and the IC manufacturer regarding performance issues, alternative approaches and the realistic prospects that a material will eventually be adopted.
At Dow Corning, we originally tried to position our low-k spin-on dielectric material, FOx(r) Flowable Oxide, for the copper dual damascene interconnect opportunity. It became apparent to us that the early adopters of low-k would be the advanced logic and microprocessor IC manufacturers, none of whom were traditional spin-on dielectric users.
Because of this, we shifted our focus to develop CVD precursor materials and technology, like low-k films based on trimethylsilane, which could meet customers needs without forcing them to shift equipment platforms. Also, by listening to the needs of customers who were already using spin-on dielectric materials, we discovered a significant market need for carbon-free planarising films, for which FOx was an ideal fit.
3. Manage risk
Materials suppliers approached low-k development with a traditional sales-based business model, where the developing company creates, patents (when applicable) and scales a material for sale in volume manufacturing. The developing company owned all the risk, but, in theory, also could lay claim to a huge jackpot if successful.
In hindsight, its easy to see that the low-k materials risk was far too high and the reward not clearly enough defined to support a reasonable return on investment for most companies. At least 17 to 18 materials were in the running and the timeframes for market adoption were dubious, at best. Low-k programmes looked more like a game of high-stakes gambling than a smart investment opportunity.
This represents a threat for chipmakers - and to the health of the industry at large. When materials development becomes undesirable from an investment perspective, the pace of innovation surely cools down. Although the rewards may be large for the company that develops a winning technology, the majority of companies in the supplier base lose. And when the majority of suppliers cannot thrive, the health of the industry they support begins to wane.
The low-k experience provides good reason for why suppliers can no longer shoulder the burden of materials development alone. Looking ahead, chip makers will need to adopt new materials at a blistering rate of one to two per technology node, according to the International Technology Roadmap for Semiconductors. Technical hurdles remain a concern, but the financial issues raise a bigger question: who can afford to pay for it all?
The industry is hard pressed for an answer. Companies are, of course, focused on achieving a realistic ROI, but todays business models are ill-suited for the challenge; a fresh approach involving early stage collaborations throughout the industry is needed.
An even better answer is to formulate new business models around early collaborations. When chipmakers, equipment companies and materials developers can partner early on materials development, they gain an opportunity to balance risk and reward among everyone involved.
New models may involve partnerships and royalty arrangements, licensing agreements or novel cost-sharing deals.
No matter what the specifics look like, early collaboration will be essential for the industry to gain a greater understanding of the business challenges at hand - and will help increase our chances of success (both financially and technically) in the long run.