In the last one year, we have witnessed a spate of traditional IDM companies changing their business strategies towards fablite, asset-light, foundry partnership or fabless business models. Notable ones include STMicro, TI, Sony, Freescale, AMD, NXP, Cypress, Spansion and WJ. Consequently, a huge business opportunity is luring at the foundries. This has prompted the world largest chip fabrication foundry TSMC to expand aggressively into the processor (GPU and MPU) and memory segments in recent months.

On the GPU segment, TSMC announced yesterday that AMD has selected its 65nm process to manufacture for a broad segment of AMD’s GPU products. The announcement coincides with an important milestone in the TSMC/AMD relationship — TSMC’s shipment of two million wafers (200mm equivalent) to AMD’s in less than a decade (Ref, Ref). AMD’s graphics chip competitor, Nvidia, is also tapping TSMC’s 65nm embedded DRAM process for their latest mobile GPU (Ref)

On the memory segment, Spansion announced a week ago that it had signed an agreement with TSMC to jointly develop its MirrorBit flash memory chips at 40-nanometers and below (Ref). This news came shortly after TSMC announced two months ago it has officially rolled out its 45-nm process technology for customers, with plans to enter production as early as September 2007 (Ref).

During the ISSCC 2007, TSMC’s founder Morris Chang talked about the future role of foundry in the following four part interview with EE Times:

Finally, the long anticipated rumor is true. Yesterday, both Intel and STMicro jointly announced that they would sell all their flash memory assets to a newly created independent NOR flash memory company (Intel PR, STMicro PR). While the news is not surprising, there are still questions hovering on the future move of the new entity.

First, both Intel and STMico will get a total cash payment of $900 million and hold a combined equity of 93.7% of the new company. The remaining equity is owned by a private equity firm, Francisco Partners which will pump in $150 million. My guess is that Intel and STMicro would probably prefer to secure more cash payments rather than holding substantial equity of the new company. The cash payout is largely funded by a loan of $1.3 billion from a consortium of banks. This means the new company would only have $550 million left to run its operation and this is clearly not sufficient for filling up the 300mm Fab in Catania. I guess the new company might sell off some of its assets to secure more cash or raise funds through other means. Second, the integration of the two companies, in terms of technology, people and culture will be quite a challenge. Most likely, the new company would adopt the Intel NOR Flash technology. Thirdly, the new company also acquire advanced phase change memory (PCM) technology from both Intel and STMicro. This is certainly important for the company to remain competitive in the long term. Lastly, the NOR Flash memory industry would become pretty much like the microprocessor industry, with two big players dominating the market. This certainly help to tilt the price elasticity of the NOR flash memory in favor of the suppliers than the buyers. Hopfully