TikTok USDS Joint Venture is pursuing four gigawatts of computing capacity by the early 2030s, a substantial infrastructure play that signals the company's commitment to American compute independence and new revenue streams.
The demand stems from two sources: a flagship B2B AI product focused on large-model inference and creative automation, and the retraining of TikTok's core recommendation algorithm on U.S. user data within Oracle's American cloud environment. Both efforts require sustained, massive compute footprints.
TikTok USDS will rely on Oracle Cloud Infrastructure (OCI) as its primary provider, according to a company spokesperson. Oracle typically operates OCI facilities through leased data centers, meaning TikTok's capacity ambitions will flow through Oracle's infrastructure agreements.
The timing is deliberate. TikTok USDS was established in 2026 to comply with President Donald Trump's executive order mandating divestiture of TikTok's U.S. operations from Chinese parent ByteDance. In January, the company announced plans to retrain and update its recommendation algorithm using U.S. user data—a regulatory requirement that doubles as a technical overhaul.
Recent job postings reveal the scale of the engineering push: TikTok USDS has listed numerous machine learning positions supporting the development and deployment of advanced AI capabilities. These roles underscore a strategic pivot into B2B products, a diversification play away from pure ad-supported social media.
The ownership structure reflects the complexity of the deal. Silver Lake, Oracle and MGX each hold 15 percent stakes as managing investors. Michael Dell's family office also holds a position. ByteDance retained 19.9 percent, the maximum allowed under the divestiture terms.
The compute demand places TikTok USDS in direct competition with Meta, Google and Amazon for long-term capacity access during a period of GPU scarcity. A 4-gigawatt footprint—roughly equivalent to the power consumption of a small country—is not trivial. Whether TikTok can secure that capacity reliably, and at what cost, will determine whether the B2B AI bet succeeds or becomes a capital-intensive drag.
