Chinese AI Startup Accesses 2,300 Banned Nvidia Blackwell GPUs Through Indonesian Cloud Loophole

Reviewed byNidhi Govil

2 Sources

Share

A Shanghai-based AI startup successfully obtained access to restricted Nvidia Blackwell GPUs by renting compute power from an Indonesian telecommunications company, exploiting a legal loophole in U.S. export controls.

Complex Supply Chain Circumvents Export Restrictions

A Shanghai-based AI startup has successfully gained access to 2,300 banned Nvidia Blackwell GPUs through a sophisticated but legal arrangement involving multiple international partners. INF Tech, founded by Chinese-born American citizen Qi Yuan who heads the AI institute at Fudan University, secured access to the restricted hardware by renting compute power from Indonesian telecommunications company Indosat Ooredoo Hutchison

1

.

Source: pcgamer

Source: pcgamer

The arrangement began in California, where Nvidia sold its latest Blackwell chips to Aivres, a Silicon Valley-based AI server manufacturer. Aivres, which operates under a parent company reportedly one-third owned by blacklisted Chinese tech company Inspur, then sold 32 Nvidia GB200 server racks to the Indonesian telecommunications provider for approximately $100 million

2

.

Strategic Pre-Arrangement of Customer Base

According to Wall Street Journal sources, Indosat did not purchase the GPU servers without first securing a guaranteed customer. Aivres reportedly helped facilitate the connection between the Indonesian telecommunications company and INF Tech before the purchase was finalized. Representatives from Fudan University were said to be present during negotiations, though INF Tech ultimately signed the contract as the official client

1

.

The 32 server racks, each containing 72 Blackwell chips for a total of 2,304 GPUs, were installed at Indosat's Jakarta facility in October 2025. While this represents a relatively modest deployment compared to the massive data centers operated by AI giants like OpenAI and xAI, it provides significant computational resources for INF Tech's AI development in finance and health applications

2

.

Legal Compliance Despite Regulatory Concerns

Lawyers familiar with export controls confirmed that the arrangement complies with current U.S. export restrictions, as none of the involved parties—Indosat, INF Tech, or Fudan University—appear on the U.S. Entity List. However, the deal raises concerns among those opposed to Chinese access to advanced American hardware, particularly given Beijing's ability to compel domestic companies to cooperate with state interests

1

.

The Biden administration's AI Diffusion Rule, designed to tighten controls over advanced chip sales to China, would have provided additional scrutiny over such arrangements. However, the Trump administration chose not to implement this rule, leaving companies to conduct their own due diligence. Thea Kindler, former U.S. assistant secretary of commerce for export under the Biden administration, noted that the abandoned rule would have given the government greater oversight of customers purchasing chips, not just exporters

2

.

Industry Responses and Policy Implications

Nvidia has advocated for more lenient export controls, arguing that allowing nations access to its hardware helps maintain U.S. technological leadership. A company spokesperson stated support for "the Trump administration's vision to secure U.S. AI leadership and create American jobs," while criticizing Biden-era controls as costly and innovation-stifling

1

.

INF Tech emphasized its compliance with U.S. export controls and stated it conducts no research with military applications. Indosat CEO Vikram Sinha defended the company's approach, explaining that all international customers undergo the same regulatory review process regardless of nationality, with support provided only after clearing all applicable regulations

2

.

TheOutpost.ai

Your Daily Dose of Curated AI News

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.

© 2025 Triveous Technologies Private Limited
Instagram logo
LinkedIn logo