Why I think the market is mispricing Oracle

Why I think the market is mispricing Oracle Oracle trades at roughly 13.8x forward earnings while sitting on $638B of contracted revenue (RPO) . The obvious concern is that fulfilling those contracts requires an extraordinary amount of capital. That concern is legitimate, but Oracle is not financing the buildout alone. Consider Michigan. The new $16B data-center campus is being developed by Related Digital with outside equity and debt; Oracle is the tenant. Across its pipeline, Oracle has committed to roughly $260B of future data-center lease payments , generally over 15–19 years. Meanwhile, Oracle says customers are prepaying for or supplying $75B of computing hardware tied to large AI contracts. Oracle is effectively combining landlord capital, customer capital and its own capex to build capacity faster than its balance sheet alone would allow. The risk is straightforward: Oracle still owes the rent if demand disappears. So the thesis ultimately depends on utilization. Today, Oracle reports 97.5% GPU utilization . When 35,000 GPUs across 59 customers recently came up for renewal, customers representing 92% of that capacity renewed , and Oracle says most of the remainder was resold within the quarter. What about cheaper/open-source AI? That may actually increase infrastructure demand. If AI becomes 10x cheaper but usage increases 50x , compute consumption still increases 5x. We have seen this economics before with bandwidth and storage: falling unit costs created vastly more consumption. So my bet isn’t that OpenAI wins. It’s that global compute consumption grows faster than compute gets cheaper , while Oracle uses other people’s capital alongside its own to supply it. At 13.8x forward earnings , I like that bet.