Tencent President: US chip ban to slow China AI training
Editorial responsibility
- No named human review is recorded for this page.
- Source reporting is collected, normalized, translated or condensed automatically when needed.
- Automatically published source-backed update
Tencent President Martin Lau stated that Washington's restrictions on offshore Nvidia chip access will significantly slow model training across China. The US Commerce Department now mandates licenses for advanced chips destined for foreign data centers serving Chinese firms, blocking the use of Blackwell chips previously rented abroad for flagship model development. Lau warned this could delay model releases and eventually force all Chinese developers onto domestic chips. In contrast, Tencent Chief Strategy Officer James Mitchell noted that inference work, which comprises most of Tencent's cloud operations, can already run on domestic chips unaffected by the new rules. This warning coincides with Tencent's third-quarter capital spending surging over six times to $12.2 billion, representing 39% of revenue. While revenue rose 8%, net profit growth was limited to 0.8%, missing forecasts. Mitchell indicated that spending remains at peak levels expected to persist through 2027.
Source report
Tencent President Martin Lau has cautioned that Washington's restrictions on offshore access to Nvidia chips will significantly slow down artificial intelligence model training across China. The US Commerce Department now mandates licenses for advanced chips destined for any foreign data center capable of serving Chinese firms, effectively blocking the export of high-performance hardware like the Blackwell chips.
Impact on Model Development
Many Chinese laboratories have historically relied on Blackwell chips—now ineligible for direct shipment to China—to train their flagship models. While these chips cannot be physically delivered to domestic facilities, they were previously accessible via rental services abroad.
Lau highlighted the specific consequences of these tighter curbs:
- Training Delays: The new restrictions will directly impact training capabilities, potentially delaying the release of new AI models.
- Forced Domestic Transition: Over time, stricter regulations will compel every Chinese developer to transition exclusively to domestic chip solutions.
Inference Capabilities Remain Stable
In contrast to training operations, Chief Strategy Officer James Mitchell noted that inference workloads—which constitute the majority of Tencent's cloud business—are already compatible with domestic chips. Consequently, these operations remain unaffected by the current export rules.
Financial Implications and Outlook
The strategic shift comes as Tencent reported a dramatic surge in capital expenditure for the third quarter:
- Capital Spending: Jumped more than 6x year-over-year to $12.2 billion, representing 39% of revenue.
- Profitability Impact: This heavy investment constrained net profit growth to just 0.8%, despite revenue rising by 8% and missing analyst forecasts.
Looking ahead, Mitchell stated that spending levels are currently at their peak and are projected to remain elevated through 2027.
Source
rohanpaul_aiNeutral / independent
Part of this Story
Tencent Warns US Chip Bans Will Delay China AI Training Amid Spending Surge