DeepSeek and Huawei unveiled open-source programming tools built for Huawei's Ascend AI chips, led by a new language called TileLang that targets Nvidia's CUDA software. DeepSeek is also preparing to deploy at least 160,000 Ascend 950DT chips in Inner Mongolia as part of a wider move away from Nvidia hardware.
On September 30, 2026, Chinese AI startup DeepSeek and Huawei Technologies unveiled a suite of open-source programming tools built specifically for Huawei's Ascend AI chips. Bloomberg reports that the pair are chipping away at Nvidia's competitive edge.
The centerpiece is TileLang, a high-level programming language pitched as a simpler alternative to Nvidia's CUDA framework.
What the new toolkit includes
The software package includes six modules covering compute, communication and programming capabilities for Ascend hardware. TileLang is designed to make development easier than it is with CUDA, while still aiming to extract full performance from the underlying chips.
The hardware commitment behind the software is substantial. DeepSeek is preparing to deploy at least 160,000 Huawei Ascend 950DT chips in a new data center in Inner Mongolia. The Ascend 950DT is set to launch in Q4 2026, with the chips expected to come partially online by late 2027 or early 2028.
Building on an earlier shift
This announcement builds on work already in motion. DeepSeek previously optimized its V4 AI model for Huawei's Ascend architecture, and Huawei's Ascend 950 chips were used in parts of V4's training. That made V4 something of a test run for the hardware transition DeepSeek is now formalizing with tooling and infrastructure.
Nvidia has noticed. CEO Jensen Huang has voiced concerns about what the DeepSeek-Huawei partnership means for US interests, and Nvidia views the collaboration as a competitive threat.
The real contest is over developer habits
Hardware can be swapped out on a purchase order, but years of code, tutorials and engineering muscle memory cannot. TileLang is a direct attempt to lower that switching cost. If developers can write code more easily while still getting full performance from Ascend chips, one of the strongest reasons to stay on Nvidia hardware weakens, at least inside China.
Releasing the tools as open source also invites outside developers to build on, test and improve the stack. For DeepSeek, the bet ties its future model training to domestic supply, which could reduce exposure to export controls, though it also concentrates dependence on a single hardware partner.
Whether TileLang can match CUDA on real-world performance benchmarks, not just ease of use, remains an open question. Years of CUDA tooling got Nvidia here; TileLang is Huawei's bet that open-source code can erode that head start inside China.
Source: Crypto Briefing
Trading involves risk.
Risk Disclaimer
