TikTok’s US business is seeking to secure as much as four gigawatts of computing capacity by the early 2030s, positioning the social media platform among companies competing for long-term access to the infrastructure needed to support artificial intelligence and increasingly demanding digital services.
The company, officially known as TikTok USDS Joint Venture, has explored opportunities to obtain additional computing capacity over the coming years, according to two people familiar with the discussions who spoke to CNBC on condition of anonymity. One source said the company was targeting four gigawatts of capacity by the early 2030s.
The scale of the reported requirement points to how the race for computing infrastructure is expanding beyond specialist AI developers and cloud providers. Social media platforms are also seeking access to the processing power needed to operate machine-learning systems, develop AI products and support complex digital workloads.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
TikTok’s plans could add another substantial source of demand to an infrastructure market already facing pressure from companies seeking computing resources over multiyear periods. Securing capacity well in advance can help businesses prepare for future workloads.
A TikTok spokesperson told CNBC that the company “runs primarily on Oracle Cloud Infrastructure in the U.S. and Oracle will continue to be our primary cloud infrastructure provider.”
Oracle Cloud Infrastructure, or OCI, typically operates in facilities that Oracle leases.
The statement confirms that Oracle remains central to TikTok US’s US infrastructure strategy, but it does not establish whether the company’s reported search for additional capacity would involve new Oracle arrangements, other infrastructure providers or a combination of options.
TikTok US’s reported capacity target comes as the company signals ambitions to expand its use of artificial intelligence beyond the recommendation systems that underpin its social media platform. The company has advertised multiple machine-learning positions on its website. One job listing says TikTok is “shipping a flagship B2B AI product at the frontier of large-model inference and creative automation.”
The description suggests that TikTok US is developing AI capabilities aimed at business customers, although the company has not disclosed enough detail in the cited information to establish the product’s precise features, launch timetable, or expected commercial contribution.
Large-model inference involves running trained AI models to generate outputs in response to user requests. Depending on the model, workload, and number of users, inference can require substantial computing resources. Creative automation can also involve demanding tasks such as generating or modifying digital content, potentially creating additional infrastructure needs as products scale.
For TikTok, the opportunity would be to apply AI capabilities to a platform already built around video creation, discovery, and recommendation. AI tools could potentially support content production and business services, while machine-learning systems remain important to ranking and delivering content to users.
However, the presence of AI-related job advertisements does not by itself establish how much computing power a new product will consume or whether the reported four-gigawatt target is primarily intended for those initiatives. The company’s infrastructure requirements could also reflect its existing platform, future user demand, and other computing workloads.
Long-term capacity planning involves uncertainty. Companies must anticipate future demand before it materializes, but committing to too much infrastructure can leave them paying for resources that are underused. Securing too little capacity, meanwhile, can constrain product launches and increase competition for resources when demand rises.
TikTok US’s reported target therefore illustrates both the scale of its ambitions and the difficulty of forecasting the infrastructure needs of AI-driven products several years ahead.
Oracle’s Role Extends Beyond Cloud Hosting
Oracle occupies an important position in TikTok US’s infrastructure arrangements because it is both the company’s primary cloud provider and one of the principal investors in the US joint venture.
TikTok USDS Joint Venture was established in 2026 to comply with US President Donald Trump’s executive order mandating the divestiture of TikTok’s US operations from its Chinese parent, ByteDance.
In January, TikTok US said it would “retrain, test, and update” its recommendation algorithm using US user data. It also said the algorithm would be “secured in Oracle’s U.S. cloud environment.”
These commitments place computing infrastructure at the center of the new entity’s operational structure. The company must support its platform while implementing arrangements for algorithm training and data security within the United States.
The reported search for additional capacity comes against that background. If TikTok US expands its AI activities and computing requirements, Oracle’s role as its primary cloud infrastructure provider could become more significant. The available information, however, does not establish that Oracle will supply all the additional capacity being explored.
Oracle’s involvement also extends to ownership. The company is one of three managing investors in TikTok USDS Joint Venture, alongside private-equity firm Silver Lake and Abu Dhabi-backed MGX. Each holds a 15% stake. The Dell Family Office, the investment firm of Michael Dell, is also an investor, while ByteDance retained 19.9% of the new entity.
The ownership structure connects TikTok US’s infrastructure strategy with the interests of investors involved in technology, finance and AI-related infrastructure. Yet the joint venture’s reported computing ambitions should not be treated as a confirmed infrastructure contract or a commitment by any particular supplier.
A four-gigawatt computing target amounts to a substantial infrastructure requirement, but its significance needs to be assessed carefully. A gigawatt is a unit of power, not a direct measure of computing performance. The actual processing capacity delivered by a given amount of power is determined by the equipment, facility design, cooling systems, and workload involved.
The target also concerns a future period extending into the early 2030s. It does not mean TikTok US currently operates that amount of capacity or has already secured the full requirement.
Nevertheless, the reported ambition highlights a broader shift in the technology industry. Access to computing resources is becoming a long-term planning issue for companies that want to develop and deploy AI products at scale. Demand is no longer limited to building models; it also includes running them continuously, integrating them into commercial services, and supporting the data infrastructure behind those services.
TikTok US’s challenge is compounded by the need to maintain its existing platform while adapting its technology operations to the requirements of the US joint venture. Its decision to localize algorithm training and secure the recommendation system within Oracle’s US cloud environment makes infrastructure availability, reliability, and governance central to the business.