Pingtouge to supply most of Alibaba's new AI chips from 2027
11/19/2026 — 11/20, 12:21·5 sources (all)·5 reports (all)
Key takeaways
- Alibaba CEO Wu Yongming said at the November 19 earnings call that Pingtouge chips will supply the majority of Alibaba's new AI chip demand…
Background & analysis
On November 19, 2026, Alibaba outlined a domestic computing strategy centered on Pingtouge during its earnings call for the second quarter of fiscal year 2027. CEO Wu Yongming said Pingtouge's third-generation GPU roadmap is moving faster: Zhenwu M890 went fully live this year, with V900 and J900 heading into mass production in 2027 and 2028 respectively. He added that by the first quarter of 2027, a new generation of Pingtouge chips will fully support training for ultra-large models, and Alibaba has already started training its next-generation models on this year's Pingtouge chips. Qwen is already training its next-generation models on Pingtouge chips, and Bailian already offers model inference services built on domestic chips.
In the same call, Wu said 2027 and 2028 will be the two years when computing capacity in China is tightest. From 2027 onward, chip supply for Alibaba Cloud will be anchored on Pingtouge's in-house chips, supplemented by purchases of other domestic chips for internal inference workloads, while flagship chips will be reserved first for large-model training. He also forecast that most models from Alibaba, its partners and its customers will depend entirely on domestic chips for both pre-training and post-training. Alibaba management further noted that shifts in computing policies across global regions could lead to cuts to its long-term 20GW data center target and overall capex plan. The peak of this computing investment cycle is expected in 2027 and the first half of 2028, with cloud revenue growth expected to return above 50% once domestic computing output scales in 2028, and cash flow turning positive in the second half of 2028. For the quarter, revenue from AI, cloud and computing services rose 53% year over year to RMB 61 billion.
On November 20, 2026, reports confirmed that from 2027 onward the bulk of Alibaba's new AI chip supply will come from Pingtouge, positioning it as the company's primary in-house source of AI computing capacity. The arrangement remains at the stage of implementation.
AI-generated from 5 reports · updated 2 days ago
Latest turnAlibaba Cloud's growth accelerated in its latest earnings, according to TMTPost, which also reports that after 2027 the vast majority of Alibaba's new AI chip supply will come from its in-house T-Head chips. The shift brings T-Head, previously a behind-the-scenes unit, into the foreground as Alibaba's main self-developed AI compute supplier.
Related AI tools
- Heat index
- 26
- All sources
- 5
- All reports
- 5
- First seen
- 3 days ago
Reports on this story headlines open the original
Alibaba Cloud's growth accelerated in its latest earnings, according to TMTPost, which also reports that after 2027 the vast majority of Alibaba's new AI chip supply will come from its in-house T-Head chips. The shift brings T-Head, previously a behind-the-scenes unit, into the foreground as Alibaba's main self-developed AI compute supplier.
钛媒体AI score 60
Alibaba CEO Eddie Wu said at the company's earnings call that 2027 and 2028 will be the tightest years for compute supply in China, according to Jiemian News. From 2027 onward, Alibaba Cloud will mainly rely on its in-house T-Head chips plus other domestic chips for inference, while flagship chips go to model training. Wu added that Qwen is already training its next model on T-Head chips and that the Bailian platform already serves inference on domestic silicon.
IT之家AI score 60Alibaba CEO Wu Yongming said at an earnings call that its next-generation Pingtouge chips will support ultra-large-scale model training in Q1 2027, and the company has already begun training its next model on this year's Pingtouge chips. He expects most models from Alibaba, partners and customers to rely entirely on domestic chips for both pre-training and post-training.
东方财富AI score 67Alibaba CEO Wu Yongming said on the November 19 earnings call that Pingtouge's third-generation GPU roadmap is accelerating: Zhenwu M890 is fully launched this year, with V900 and J900 moving toward mass production in 2027 and 2028. Qwen is already training its next model on Pingtouge silicon, and Bailian serves inference on domestic chips.
36氪AI score 58On its fiscal second-quarter earnings call, Alibaba said its long-term 20GW data center goal and overall capex plan may be scaled down as compute policies shift across regions. It will accelerate a domestic chip supply chain built around T-Head, expecting cloud revenue growth to return above 50% once that compute scales in 2028 and free cash flow to turn positive in the second half of 2028. AI, cloud and compute services revenue rose 53% year over year to RMB 61 billion in the quarter.
华尔街见闻AI score 70
Other stories people are talking about
- Heat index 49A roundup of the 10 hottest open-source video projects on GitHub in 20261 sources
- Heat index 63Nasdaq 100 enters bear market as AI doubts hit tech stocks1 sources
- Heat index 46POCKET-27B-Pi coding model released with 16.5GB quantized build1 sources
- Heat index 25Developer recounts 10 runtime pitfalls from six months of a homegrown AI coding agent1 sources
- Heat index 38Together AI launches Qwen3 TTS with 9 voices across 10 languages1 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion

