Institutional logic theory emphasizes the bidirectional interaction between organizations and institutions, demonstrating enduring theoretical vitality in explaining organizational agency and strategic variation. However, this theory implicitly presumes the reconcilability of multiple institutional logics, yet lacks a systematic examination of interaction processes-particularly organizational agency-under conditions of extreme institutional logic conflict. This paper introduces the novel concept of “buffered institutional work,” revealing a fundamental shift in organizational agency: when institutions exceed an organization’s capacity for transformation, the focus of agency shifts from “changing institutions” to “altering how institutions apply to the organization,” thereby creating temporary buffer spaces within rigidly conflicting institutional logics. A tool-based case study of NVIDIA’s lobbying practices under the context of the “digital Cold War” reveals the action strategies and objectives of buffered institutional work: cognitive buffering, material buffering, and relational buffering serve as core tactics, with the goal being “temporary mitigation” rather than “complete resolution.” Further deductive analysis shows that the effectiveness of buffered institutional work is profoundly influenced by the organization’s available “buffering” resources and the degree of coordination among buffering strategies; its boundaries are shaped not only by explicit constraints of institutional logics but also by implicit limits arising from “strategic lock-in.” The theoretical contribution of this paper lies in uncovering the critical conditions under which the classical assumption of “reconcilability” in institutional logic theory breaks down, and in introducing “buffering” as a new form of organizational agency beyond the traditional typology of creation, maintenance, and destruction.