SoftBank and Sprint combined, which have a subscriber total of ~88M, are dwarfed by the combined strength of Vodafone (450M), Airtel (266M), SingTel (265M), América Móvil (252M), Telefónica (250M), Orange (230M), VimpelCom (215M), TeliaSonera (160M), Telenor (150M), and Deutsche Telekom (130M). All of those operators participate in the GSM/UMTS/LTE ecosystem, and all of them are doing GSM/LTE or GSM/UMTS/LTE with LTE FDD and LTE TDD with Bands 7+38 instead of Band 41 LTE TDD. That is an ecosystem of 2368 million (~2.4 billion) subscribers.
Unfortunately enough, SoftBank can participate in this ecosystem and get good pricing on handsets with a swap to Band 41, but Sprint cannot. This is because SoftBank's handsets involve a simple filter swap on GSM/UMTS/LTE devices that support Band 38 to widen to Band 41 (and not include the Band 7 PA, which nearly all Band 38 devices currently do not have anyway). Also, since SoftBank uses bands for UMTS that are the same as the rest of Asia and Europe, there is a higher degree of reuse. This dramatically cuts the cost.
Sprint has several counts against it in the ecosystem. While it uses PCS A-F spectrum (which is widely used for UMTS service), it provides CDMA2000 service on that band instead. It also provides CDMA service on ESMR, with plans to provide LTE service on the band soon, too. Additionally, its PCS G block has not yet been auctioned elsewhere because the viability of the ecosystem is considered suspect, so the PCS G LTE network is considered "unusual". While it is true that most power amplifier parts are multi-mode, the procurement of CDMA devices and infrastructure is much more expensive because of the vastly reduced market for it. It doesn't help that Verizon's planned exit of the user device procurement market for CDMA/LTE devices will cause an ecosystem crash (it cuts the size of the CDMA/LTE market by more than half). Sprint will have to spend substantially more per device, which means Sprint has less money to spend on infrastructure.
3GPP infrastructure will be much cheaper for Sprint to acquire now, since it can use the combined strength of Sprint and SoftBank, but 3GPP2+3GPP gear will continue to get more expensive. That is why SoftBank wants to convert Sprint to 3GPP-only by 2017. It doesn't want to fund what it considers to be a waste (which it does consider the 3GPP2 gear to be that).