Governing Epistemic Sustainability: Epistemic Enclosure, Double Concealment, and Knowledge Sovereignty in India's Viksit Bharat 2047

India's Viksit Bharat 2047 framework (2026) and the IndiaAI Mission (2024) conflate compute sovereignty with knowledge sovereignty, embedding India within AI infrastructure whose epistemic consequences existing sustainability governance cannot register. Reading India's AI governance corpus as performative political narratives, the paper identifies the rhetorical operations through which compute capacity is narrated as knowledge sovereignty, and the silences through which epistemic costs are rendered external to the governance frame. To name what this discourse cannot see, the paper develops three interlocking concepts, each addressing a different question: what must be sustained, what happens to knowledge under enclosure, and why that enclosure persists ungoverned. Epistemic sustainability is the capacity of a knowledge system to remain durable, plural, and transmissible across time without structural dependence on extractive, proprietary, or epistemically homogenising infrastructure, extending Scheman's (2012) norm from the ethics of inquiry to the political economy of knowledge infrastructure. Epistemic enclosure names the mechanism: the print-to-digital-to-AI trajectory, narrated as democratising, renders postcolonial knowledge traditions legible only on terms they did not set, through architectures optimised for English-language statistical patterns. Indian languages account for approximately 0.5 per cent of Common Crawl data, a major source corpus for language-model training; traditions including Nyaya, Mimamsa, and Ayurveda are rendered legible through patent classification, formal logic, and metadata standards external to their conditions of intelligibility. Double concealment names the ideological operation sustaining this enclosure: material infrastructural costs are externalised from green AI accounting, while the foreclosure of non-Western epistemic plurality is rendered invisible behind narratives of democratised access. The paper proposes three governance instruments: epistemic impact assessments for public AI infrastructure investment; a policy framework distinguishing compute from knowledge sovereignty; and multilateral mechanisms protecting the epistemic commons of postcolonial traditions in global AI regimes. As India approaches its 2047 centenary, the governance question its AI discourse has not yet asked is whether the knowledge it produces will remain answerable to India's plural epistemic traditions, or merely processed through infrastructures designed elsewhere.