--- title: "Baidu claims more than 30% cheaper AI inference, but the M300 chip behind the prices lands in 2027" description: "The silicon meant to lock those prices in, a 1,024-card supernode and the M300 flagship chip, is dated 2027. Third-quarter results on November 16 will show who pays for the wait until then." publisher: "The Inference" section: "Chips" published: 2026-11-12T15:38:29.372Z modified: 2026-11-12T15:38:29.372Z canonical: https://theinference.org/article/baidu-claims-more-than-30-cheaper-ai-inference-but-the-m300-chip-behind-the-prices-lands-in-2027 language: en keywords: "Baidu, AI Chips, Inference, China AI, Earnings" --- # Baidu claims more than 30% cheaper AI inference, but the M300 chip behind the prices lands in 2027 > The silicon meant to lock those prices in, a 1,024-card supernode and the M300 flagship chip, is dated 2027. Third-quarter results on November 16 will show who pays for the wait until then. ## A price list on stage in Beijing At [Baidu](https://theinference.org/markets/companies/baidu) World 2026 on November 12, founder and chief executive Robin Li said inference, not training, will be "the primary driver of sustained long-term growth in the AI chip market" [1][2]. His prices followed. Two new application models, ERNIE Light Deep Thinking and ERNIE Wizshot Micro-Expression, cut the cost of delivering inference by more than 30% and video production cost by 50%, Baidu said [1][2]. Customers benefit first: finished video from the Baidu Yijing agent now costs RMB 0.2, about three US cents, a second [2][3]. [The Token Exchange marketplace for model calls](https://www.yicaiglobal.com/news/baidus-new-llm-trading-platform-exceeds-50-million-average-daily-calls-in-first-month), one month old, already averages more than 700 billion tokens a day [2][3]. ## The cheap silicon is dated 2027 The roadmap under those prices is a 2027 product [1][2]. Baidu says the Tianchi supernode, 1,024 of its own Kunlunxin accelerator cards chained over an all-optical interconnect, launches in 2027 alongside the flagship M300 chip, with an "N architecture" generation to follow in 2028; the M100 inference chip has already shipped as planned [1][2]. The claimed gains, three times the chip-to-chip bandwidth, 2.5 times the training performance and three times the inference performance, arrive without a named comparison machine, a chip price or a capex figure [1][2]. Until then, Baidu's data centers keep running a mix of Kunlunxin and [Nvidia](https://theinference.org/markets/companies/nvidia) hardware [8]. Kunlunxin held [4% of China's data-center AI chip market in the first half of 2026](https://finance.yahoo.com/technology/ai/articles/u-export-restrictions-drive-nvidia-124325286.html) [7]. Nvidia's shipments into that market fell 44% in the first half, to 678,000 units; [Huawei](https://theinference.org/markets/companies/huawei), the nearest rival at 28% share, shipped 572,000, up from more than 300,000 [7]. *Chart: **Kunlunxin holds 4% of the AI chip market it needs to win** Bar chart of China's data-center AI chip market by unit share in the first half of 2026: Nvidia 34%, Huawei 28%, Alibaba T-Head 7%, Cambricon 5%, Baidu's Kunlunxin 4%, Hygon 3%.* *Share of China's data-center AI chip market by units shipped, January to June 2026. IDC data reported October 13, 2026. [7]* ## Monday prices the wait [Third-quarter results land on November 16](https://www.marketwatch.com/story/baidu-s-revenue-likely-dropped-in-third-quarter-earnings-preview-49119e78) [5]. [FactSet](https://theinference.org/markets/companies/factset) consensus puts revenue at RMB 30.8 billion, slightly below the 31.2 billion yuan a year earlier, and net profit at RMB 1.65 billion [5]. [Citi](https://theinference.org/markets/companies/citigroup) expects AI cloud infrastructure revenue up 99%; Deutsche Bank has cut its 2026 and 2027 profit forecasts by 15% and 14% on the shift toward lower-margin cloud [5]. [The second quarter showed both edges of the trade](https://finance.yahoo.com/markets/stocks/articles/baidu-bidu-q2-2026-earnings-002835072.html): total revenue down 4%, GPU cloud revenue up 283%, AI cloud profit and margin rising [6]. [Bank of America](https://theinference.org/markets/companies/bank-of-america) reads China's price war as a move from token prices to cost per completed task, with the durable economics in chips and cloud platforms rather than model vendors [9]. That is the bull case for Baidu's stack, and it stands only if Kunlunxin is genuinely cheaper than the Nvidia silicon it replaces; no third party has tested that yet. *Chart: **Baidu's revenue has sat between $4.4 billion and $4.8 billion for two years** Line chart of Baidu's quarterly revenue in US dollars from Q3 2024 to Q2 2026, between $4.379 billion and $4.782 billion each quarter, flat to slightly down.* *Baidu's total revenue by quarter, US dollars, Q3 2024 to Q2 2026. Figures from Baidu's SEC filings, retrieved November 12, 2026; third-quarter 2026 results are due November 16. [10]* ## The cut may buy share, not margin The bear risk runs through rivals: Chinese model vendors have been cutting token prices aggressively since [DeepSeek](https://theinference.org/markets/companies/deepseek) kicked off the race [9], so a cheaper floor helps them as much as Baidu, and the 30% cut may buy share rather than margin. Watch the AI cloud margin line on Monday, [the November 25 release of the design files for ERNIE 5.5 Flash](https://cryptobriefing.com/baidu-ernie-5-5-flash-preview-launch/), Baidu's new model [2][4], and the M300 date. Until the M300 lands, Baidu is selling 2027 economics on 2026 silicon [2]. ## Takeaway Baidu cut the price of inference by more than 30% on hardware that does not arrive until 2027: the M300 chip and the Tianchi supernode are dated then, and the November 16 results will show who pays for the wait. ## Sources 1. [ChosunBiz, "Baidu unveils AI chip roadmap, to launch 1,024-accelerator system in 2027", 12 November 2026](https://biz.chosun.com/en/en-international/2026/11/12/U3ZYVLUTYZEATCCAH35K7CSX3Q/) 2. [Baidu press release via PR Newswire, "Baidu Unveils Latest ERNIE Models Amid Broader Full-Stack AI Lineup, and Debuts Token Exchange", 12 November 2026](https://www.prnewswire.com/news-releases/baidu-unveils-latest-ernie-models-amid-broader-full-stack-ai-lineup-and-debuts-token-exchange-to-power-intelligent-organizations-302937959.html) 3. [Yicai Global, "Baidu's New LLM Trading Platform Exceeds 50 Million Average Daily Calls in First Month", 12 November 2026](https://www.yicaiglobal.com/news/baidus-new-llm-trading-platform-exceeds-50-million-average-daily-calls-in-first-month) 4. [Crypto Briefing, "Baidu launches ERNIE 5.5 Flash Preview with 280 billion parameters", 11 November 2026](https://cryptobriefing.com/baidu-ernie-5-5-flash-preview-launch/) 5. [MarketWatch (Dow Jones Newswires), "Baidu's Revenue Likely Dropped in Third Quarter - Earnings Preview", 11 November 2026](https://www.marketwatch.com/story/baidu-s-revenue-likely-dropped-in-third-quarter-earnings-preview-49119e78) 6. [Baidu Q2 2026 earnings call transcript, Yahoo Finance / The Motley Fool, 18 August 2026](https://finance.yahoo.com/markets/stocks/articles/baidu-bidu-q2-2026-earnings-002835072.html) 7. [Yahoo Finance, "U.S. Export Restrictions Drive Nvidia China AI Chip Market Share Down to 34%" (IDC data), 13 October 2026](https://finance.yahoo.com/technology/ai/articles/u-export-restrictions-drive-nvidia-124325286.html) 8. [CNBC, "Baidu is emerging as a major AI chip player in China to fill the Nvidia gap", 27 November 2025](https://www.msn.com/en-us/money/markets/baidu-is-emerging-as-a-major-ai-chip-player-in-china-to-fill-the-nvidia-gap/ar-AA1RjkxO) 9. [Business Insider, "China's AI price war is entering a new phase" (Bank of America report), 17 September 2026](https://www.businessinsider.com/china-ai-price-war-new-phase-bofa-2026-9) 10. [Baidu, quarterly figures from its SEC filings, retrieved 12 November 2026](https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001329099) --- Baidu claims more than 30% cheaper AI inference, but the M300 chip behind the prices lands in 2027 — The Inference. Canonical: https://theinference.org/article/baidu-claims-more-than-30-cheaper-ai-inference-but-the-m300-chip-behind-the-prices-lands-in-2027