The gemara at first tries to learn brachos from pesukim, implying that they are d'oraysa. There seems to be two opposite approaches being used by the gemara regarding the bracha rishona vs. the bracha achrona. On one hand the gemara implies that had we only had one source for bracha, we would assume it referred to bracha achrona, indicating that bracha achrona is more compelling. Yet, the gemara also introduces a kal v'chomer that if we make a bracha achrona, we should certainly be making a bracha rishona, implying that bracha rishona is more compelling. Rashi seems to be bothered by this question and explains that although the gemara assumes the rationale to require a bracha rishona is stronger than for a bracha achrona (therefore is able to make a kal v'chomer from bracha achrona to bracha rishona), nevertheless, had we only had one source for bracha we would assume that the bracha achrona was d'oraysa because we find a precedent for bracha achrona being d'oraysa on bread - ואכלת ושבעת וברכת.
At the conclusion of the gemara it seems to use pesukim as a source for bracha achrona on all 7 species, but is unable to find a source for bracha achrona for other foods and is forced to conclude that it is just logic - סברא הוא, אסור לו לאדם שיהנה מן העוה"ז בלא ברכה. According to this it should come out that bracha achrona on all 7 species, not just bread is d'oraysa. This is actually the opinion of Rabbeinu Yona (although the chidushei anshei sheim points out that Rabbeinu Yona at the end of the perek 32a in dafei ha'rif writes that the ברכה אחת מעין שלש is only d'rabonon), however, the Rosh implies that at the conclusion of the gemara our only source for bracha (other than birchas hamazon on bread) is just logic, implying that even on the 7 species it is only d'rabonon. I also found that the Rabbeinu Chananel (on the page in some printings) writes explicitly that according to the opinon who holds kerem r'vai the gemara concludes that ואכלת ושבעת וברכת is going on all 7 species, so the bracha achrona will be d'oraysa. And the opinion who holds of neta revai will hold that bracha achrona on all trees that are subject to revai is going to be d'oraysa. The sevara is only needed to fill in for other items and for bracha rishona.
The gemara also seems to conclude that on whatever the bracha achrona would be d'oraysa, the bracha rishona on that item would also be d'oraysa based on the kal v'chomer. Meaning, that according to most rishonim who consider bracha achrona on everything except bread to be d'rabonon, the bracha rishona on bread my still be d'oraysa based on the kal v'chomer. And according to Rabbeinu Yona that the bracha achrona on all 7 species is d'oraysa, we could use the kal v'chomer to say that the bracha rishona on all 7 species is also d'oraysa. However, Tosafos and Rabbeinu Yona both point out that the gemara 21a seems to reject this logic because it considers bracha rishona d'rabonon even on bread where bracha achrona is certainly d'roraysa. We see that ultimately the gemara doesn't accept this kal v'chomer (for an explanation as to why we don't accept the kal v'chomer, see Meshech Chochma on the pasuk in parshas eikev ואכלת ושבעת וברכת). The Tzlach proves that even bracha rishona on bread is only d'rabonon from the mishna 42a where we allow one to make a bracha rishona for others. If bracha rishona were learned from bracha achrona using a kal v'chomer it would only be d'oraysa when eating a shiur of שביעה, so how can we simply say that one can make a bracha rishona for others - perhaps the me'vareich is not eating a shiur so his bracha is only d'rabonon whereas the others are eating a shiur and are required m'doraysa (there is no arvus for bracha rishona). Clearly, we always assume bracha rishona is only d'rabonon.
Another point raised by the Pnei Yehoshua is that elsewhere we find that a sevara can be in place of a verse in the Torah. The gemara in kesubos says that the source of פה שאסור הוא הפה שהתיר, of believing someone to be telling the truth because they created the doubt to begin with, and the concept of המוציא מחבירו עליו הראיה, which are both Torah concepts are learned from sevara. Therefore, although the gemara concludes that the source is sevara, bracha rishona may still be d'oraysa. The Tzlach disagrees and says that this is not possible. Although sevara can replace a pasuk to teach a din, it cannot be a source of a mitzvah or an obligation. If sevara were able to be a source of a mitzvah, we wouldn't need the Torah to write any מצות שכליות (logical mitzvos such as stealing and killing) since we can learn them out from sevara. Furthermore, if the sevara can introduce a mitzvah, it should apply even to goyim, yet the requirement to make brachos is only for Jews. Since the gemara concludes that bracha rishona is derived from sevara, we assume that all bracha rishona are only d'rabonon.