Every few years we hear that, at last, it's now the era of the "system on a chip" (SOC). I find this almost humorous, since there is no standard, widely accepted definition of the SOC entity. One person's system is another person's component or building block (for example, in the audio/video world, a 19" rack-mounted chassis is casually called a "component").
An SOC, to me, is more of a marketing label than a technical one. The reality is that one person's "system on a chip" is another's "more integrated solution." An SOC can combine a core function such as a processor, plus its associated peripheral functions such as I/O, onto a single IC.
For some applications, however, it's those so-called peripheral functions which are the real product differentiators, rather than the core processor itself. Alternatively, an SOC can encompass more of the signal chain, such as a front-end amplifier, filters, down-converter, ADC (and its buffer amplifier), and other functions of the serial signal path.
There's no question that integrating more functions onto a single IC can yield cost savings though the implications of Moore's conjecture (commonly known as Moore's Law); our industry has proven this repeatedly over the decades. But there is no free lunch, either. For the complex tradeoffs in the mixed-signal paths which are commonplace in the RF world, the cost and footprint savings of an SOC may not be worth what you have to give up, or the design inflexibility of most SOCs. The IC processes needed for a really good low-noise amplifier (LNA) front end may offer the right characteristics for a fast, wide dynamic range A/D converter or higher-power I/O drivers.
In the end, a multichip approach may be the best. That's why so many analog and RF ICs are single-function devices, and there are so many choices even for a single function (just look at how many LNAs are offered by a given vendor).
So while we all hail the SOC in principle, I'll retain the right to assess the "gain versus the pain" of each on a case-by-case basis! ♦