OSATs eyeing chiplet opportunities

Julian Ho, Taipei; Rodney Chan, DIGITIMES Asia 0


ASE Technology and other OSATs capable of providing advanced packaging technology are all gearing up for a boom in demand for chiplets.

Deglobalization will shoot up production costs for the semiconductor industry, and as the focus is turning toward enterprise technology, particularly AI and machine learning, the advanced packaging needed to support some of the chiplet products will be handled by OSATs, according to ASE group chairman Jason Chang.

Major cloud services providers, including Meta, Microsoft, Google and Amazon, have been investing large sums in constructing datacenter infrastructures, beefing up demand for HPC chips. The growth in demand for HPC chips may be slower in 2023, but stronger growth can be expected beginning next year. AI accelerator chips from Nvidia, AMD and Intel will play an increasingly important role in the industry.

Development of advanced semiconductors continues trending towards higher density and highly heterogeneous integration. The ASE group's R&D continues to focus on SiP, 2.5/3D IC and fan-out packaging, and it also places emphases on high I/O pin counts, fine pitch, and optimal heat dissipation.

Advanced packaging technology is now essential to production of high-end AI chips, and the concept of chiplets will take center stage in the medium to long term. According to ASE, chiplets will no longer be a sector dominated by top foundry houses or IDMs because of the following reasons.

First, the impact of deglobalization is becoming more obvious in the semiconductor supply chain. From the standpoint of OSATs, the traditional semiconductor industry embraces a vertical division of work between designers, manufacturers and packaging/testing service providers. In the chiplet era, the top foundry houses will initially be able to dominate both the manufacturing and packaging/testing sectors because of their better capability of maintaining yield rates.

Then driven by packaging costs, major IDMs will expand into the contract chipmaking sector, and traditional OSATs will also be able to improve their chiplet production yield rates. By then, some chiplet products will switch back to the traditional process where the designing, manufacturing and packaging/testing are separately handled.

Second, advanced wafer fab capacity is expanding in the US and Europe. While some chiplet production will switch back to the traditional division of labor, the advanced packaging capacity for chiplets will be lacking in the US. If the US does not increase the local packaging capacity in tandem with the rise of its chipmaking capacity, its chips will still have to be sent to Asia for packaging and testing, lengthening the time to market.

Third, packaging/testing service providers will change the way of assessing fab expansion projects: emphasis will be placed more on catering for supply chain changes and customer needs, rather than simply on cost concerns. The deglobalization will prompt the world's top-3 foundry houses with advanced manufacturing capabilities to build fabs in the US and Europe.

The world's major OSATs will also set up advanced packaging/testing plants in the US and Europe to serve the local wafer foundries. They will work with US IC design customers concerning their deployments, seeking to maintain Taiwan's leadership in the global packaging and testing market.