Unless the temperature differential is huge, hundreds of degrees C, the efficiency of these is tiny - around 5-7%. Incorporating things like stirling generators

Unless the temperature differential is huge, hundreds of degrees C, the efficiency of these is tiny - around 5-7%. Incorporating things like stirling generators could theoretically get up to around 35-45% efficiency, but that adds weight and complexity, with the maintenance and production costs offsetting the potential economic gains in reduced fuel usage. Better thermoelectric generator technology is possible - there are lots of exciting possibilities with graphene, and heat pumps, and so forth, but saving $10,000 in fuel costs over 5 years for $15,000 worth of technology in a car doesn't make sense. Auto manufacturers optimize for price and performance and space and maintenance and all sorts of competing factors, so it's not always as easy as engineering a technical solution and slapping it on top of everything else that goes into a product.