SEMI ISS: The Outlook

I spent Monday at the SEMI ISS meeting in Half Moon Bay. There is a lot of stuff to choose from, the entire day was interesting throughout. The day started with a keynote from Scott McGregor, the CEO of Broadcom. I’ll cover that in a separate posting.

Next up was Andrea Lati of VLSI Research on The Cycle: Is It Different This Time?He started with a review of 2014. Things that went right:

• Apple’s A8 in 20nm at TSMC
• Intel’s 14nm FinFET
• EUV moving to production insertion at N10

And what went wrong:

• 450mm put on back-burner, won’t happen until 2023 or later (or never in my opinion)
• 3D TSV still three years out
• Electronics market continued to grapple with falling ASPs (commoditization of smartphones, tablets fizzled, electronic market overall growing slower than semiconductor)

The answer to the question in the title of his talk was that magnitude of swings in the silicon cycle are much dampened when looking back as far as the 1980s. Equipment seems to be on a 2 year cycle for a couple of decades and based on history 2015 should be a good year. The big unknown: China goes nuclear to expand its semiconductor industry (for the 3rd time, maybe it will work this time).

Next up, Mario Morales of IDC on What’s Next for the Semiconductor Market. By the end of 2015 he expects:

• A base of 680M iOS and 1.7B Android devices
• 3.3M apps available, 45-50 apps downloaded per year per device
• But key market indicators are slowing
• Tablet market his wall (iPad shipments actually declined) partially due to cannibalization by Phablets

IoT will start in industrial where there are already massive amounts of data coming online. One machine at one plant might produce 1TB/day of data to be analyzed. IoT will then expand into the home. Challenges are low power, cost and security.

Mike Corbett of Linx-consulting talked about materials. One interesting datapoint is that wafer cost is growing 41% node to node but the material node to node growth rate is 43%. As a result, materials are growing as a portion of wafer cost, especially sub 20nm due largely to all the additional material required for double patterning. DRAM is similar, with 9% node to node wafer cost growth but materials cost growing at 14%.

Nariman Behravesh had his Top 10 Economics Predictions for 2015. So here they are:
[LIST=1]