OpenAI, Oracle Won't Expand Flagship AI Data Center in Texas
Summary
- Oracle and OpenAI have terminated plans to expand the "Stargate" flagship data center in Abilene, Texas, halting the move from 1.2GW to 2GW.
- Oracle is reportedly facing a "cash crunch" and cutting thousands of jobs, leading to the cancellation of this specific capital-intensive expansion.
- Meta Platforms has entered negotiations to replace Oracle as the tenant for the capacity being developed by Crusoe.
- Nvidia proactively paid $150 million to the developer (Crusoe) to contractually ensure the facility uses Nvidia GPUs rather than AMD chips.
Trade Ideas
"Oracle and Open Air have scrapped plans to expand that flagship data center in Texas... Oracle was planning to cut thousands of jobs as it was dealing with a cash crunch." The cancellation of a marquee AI infrastructure project due to financing issues and a "cash crunch" directly contradicts the high-growth AI narrative priced into the stock. It signals that Oracle may lack the capital depth to compete with hyperscalers (Microsoft, Amazon, Google) in the race for gigawatt-scale clusters. Short-term bearish catalyst as the market reprices Oracle's AI execution risk. Oracle still retains a separate 4.5GW agreement with OpenAI; this may be an isolated site issue rather than a systemic failure.
"Meta has come into the equation and is negotiating in early talks with Crusoe... to be the tenant to take on to lease that capacity... now that Oracle is going to walk away from it." While Oracle retreats due to capital constraints, Meta is stepping in to scoop up premium, shovel-ready power capacity. This confirms Meta's massive capex flexibility and aggressive posture in securing infrastructure, allowing them to deploy compute faster than constrained peers. Long on Meta's infrastructure dominance and balance sheet strength. Negotiations are in "early talks" and could fall through; overspending on capex remains a concern for some investors.
"Nvidia made a $150 million deposit or payment to Crusoe to ensure that whatever happens in that next phase of expansion in Abilene, it is their computers, their GPUs that power it and not AMD." Nvidia is not just competing on specs; they are using their balance sheet to pay "slotting fees" that lock competitors out of physical data centers. This aggressive defensive maneuver creates a formidable moat, ensuring that even as tenancy changes (from Oracle to Meta), the silicon remains Nvidia's. Long on Nvidia's pricing power and competitive entrenchment. Regulatory scrutiny regarding anti-competitive practices if they are paying specifically to block rivals.
Nvidia paid a deposit specifically to ensure the site uses Nvidia hardware "and not AMD." If the market leader is successfully paying developers to physically exclude AMD from large-scale clusters, AMD faces a structural barrier to entry that superior technology alone cannot solve. This limits AMD's Total Addressable Market (TAM) in premium, gigawatt-scale facilities. Bearish on AMD's ability to capture high-end data center market share. AMD could secure wins in other data centers where developers refuse exclusive lock-in deals.
