Oracle reportedly scraps $7 billion Tencent AI compute deal

By Tom Whitfield

Oracle reportedly canceled a five-year cloud computing agreement with Tencent valued at approximately $7 billion after US officials raised concerns about giving a Chinese customer remote access to advanced AI chips. The Financial Times, citing two people familiar with the matter, reported the cancellation.

Neither Oracle nor Tencent had publicly confirmed the agreement or its reported collapse at publication. The contract’s scale was significant, but its terms remained unconfirmed.

What would the $7 billion agreement have provided?

Under the reported plan, Oracle would have housed up to 100,000 advanced AI chips in data centers operated by Southeast Asian cloud providers. Tencent would rent the computing capacity remotely, gaining processing power that US export controls effectively restrict Chinese companies from buying directly.

The proposal converted direct hardware access into a remotely delivered cloud service. That made the server location and the customer’s identity relevant to export-control risk, rather than leaving geography as the only consideration.

The account did not identify chip models, process nodes or wafer sizes. It also did not present the approximately $7 billion contract value as confirmed revenue. Commercially, the plan would have shifted advanced compute from a restricted direct sale to remotely delivered capacity.

Did Washington take formal action?

Available reports identify informal pressure, not a formal enforcement order or legal ruling. US officials reportedly raised concerns with Oracle, while Washington signaled that overseas cloud infrastructure serving Chinese customers might fall within export-control scope.

No detailed public statement from Oracle, Tencent or the White House confirmed that signal at publication. If the policy hardens, cloud providers may need to reassess contracts that rely on offshore capacity for Chinese customers. For now, the government position remains a warning rather than a settled rule.

Why did Tencent face pressure from Beijing and Washington?

Business Times, citing the Financial Times account, reported that Guangdong officials pressed Tencent to shift toward domestically produced semiconductor alternatives. Beijing wants to reduce dependence on foreign chips and strengthen China’s domestic semiconductor sector.

Washington took the opposite approach, seeking to restrict Chinese access to advanced AI computing. Tencent therefore faced US pressure to limit access and Chinese pressure to promote local supply at the same time. The two policy goals worked against the same cross-border cloud arrangement.

What does Nvidia’s reported 51% threshold mean?

Separate reporting cited by Forbes and 247WallSt says Nvidia told cloud customers that at least 51% of their services must be sold to non-Chinese customers under new agreements. Nvidia had not publicly confirmed the reported requirement.

The company also had not explained how compliance would be measured. Revenue, capacity and customer count were identified as possible benchmarks, but the cited coverage did not confirm any of them as the required measure. The account placed this move amid a broader surge in AI-chip demand, although it supplied no shipment totals or pricing data.

The restriction, as reported, applied to future agreements; existing contracts were not necessarily affected. The reported percentage was specific, but the enforcement mechanism was not.

What remains at risk for ByteDance and Oracle?

ByteDance reportedly maintains active chip-leasing arrangements with US cloud providers. Oracle was also reportedly a significant supplier to ByteDance across Asian data centers, including facilities in Malaysia.

Those arrangements now face the same scrutiny applied to the Tencent case. The commercial mechanism is straightforward: a chip located outside China can still reach a Chinese customer through a cloud contract, making deployment location only one part of the review. For Oracle, its Asian business therefore retains regulatory exposure even without an announced enforcement action.

What can readers conclude from an unconfirmed cancellation?

No primary source confirmed the full arrangement, its cancellation or the governing policy framework. The available reporting showed concern and commercial uncertainty, but no formal enforcement action.

Still, the report marks a change in risk for providers hosting advanced processors outside China while serving Chinese customers. Cloud access can no longer be treated as an isolated service arrangement outside the reach of export-control scrutiny.

The cancellation may be an early signal, not a final policy, that overseas compute access will receive closer review. Until then, providers face uncertainty over compliance measurement and contract structure. If the reported Nvidia rule becomes written policy, capacity allocation becomes the key commercial variable: providers must reserve at least 51% of services for non-Chinese customers under new agreements.