You are spamming;  advertising your product is off-topic for this thread, doubly so since its already been spamvertised once here;  but since you've been so bold--  I inquired and found out that your product is based off the same weak, barely tested/reviewed, and slow as heck naive cryptographic code used in the product being discussed here.  The information leak here is so severe that I am very doubtful that your (quite laudable) improved hardware isolation can prevent-- e.g. the code in question leaks several bits of information about the key from just the time it takes.

Furthermore, Your "directly in silicon" is an FPGA with a loading procedure 'under the seal', this is potentially yet another back door vector, it sinks a lot of power, and really seems to be of dubious value. I would have preferably seen all the external interfaces over simple low-ish-speed serial interfaces with good electrical isolation, rather than a huge power sucking FPGA under the secure-area can.  Use of a BGA probably also means you need a 4 layer board for signals routing and thus probably can't use an extra layer as a separate ground to complete the shield can. The FPGA just seems like a costly gimmick to me, and that you're misrepresenting this as a solution to bad cryptographic code (which you have made a similar failure by selecting to use it) doesn't bode well for the security of your product.