An article from EE Times today described a speech from the president of Nvidia, who made the following statement, which EE Times noted, raised a lot of eyebrows:
"A small fabless company is fundamentally long-term intractable," he said. Huang argued that given the increasing resources necessary to field a chip and the diminishing markets for application-specific products, the only survivors in the fabless world would be companies that "identified a broad swath of applications that could be met with a single piece of configurable or programmable hardware. Only then can they achieve the volumes necessary to survive."
This is one of those high-level general statements which might apply to a few fields, but does not apply to every single segment. He is correct that it is taking increasing amounts of capital to get an IC company off the ground, mainly due to the skyrocketing costs of mask sets, which are now over $1 million at the latest process nodes. But plenty of niches still exist where a nimble company with unique technology can create a "one-off" product and generate a large number of sales. They then have to maintain this lead as new companies and eventually the majors enter this field.