Baidu

Baidu’s free cash flow dropped by CNY 5 billion sequentially to negative CNY 8 billion, driven by a 92% quarter-on-quarter surge in AI-related capital expenditure. While cash-generative advertising revenue fell 19% year over year, revenue from its capital-intensive AI businesses grew by only 25%.

Why it matters: Capital expenditure exceeded our expectations, prompting us to increase the combined capex over the next five years by 55%. We think medium-term pressure on AI infrastructure costs will reduce the AI revenue/capex ratio during 2026-29. Search advertising revenue outlook is worse-than-expected, worsening Baidu's free cash flow outlook along with higher AI investments. We cut non-AI Baidu General Business revenue, which includes search revenue, by an average of 14% during 2026-30. AI revenue missed our estimate. We trimmed AI-powered business revenue during 2026-30 by 2%, as higher AI cloud infrastructure revenue, supported by strong demand and increased capacity, will be offset by lower AI native marketing and application revenue.

The bottom line: We cut our fair value estimate for no-moat Baidu by 10% to USD 113 per ADS or HKD 110 per share as we cut our non-GAAP EBIT forecasts by an average of 12% during 2026-30. The shares are moderately undervalued in our view. We think the long-term earnings growth prospects of Baidu's AI cloud and Kunlunxin businesses remain underappreciated by the market. We now expect Baidu's non-GAAP EBIT to grow to CNY 20 billion by 2030 from an estimated CNY 13 billion this year, mainly driven by Baidu's long-term growth opportunities in AI cloud infrastructure, offset by a decline in search advertising.

Long view: We now expect free cash flow to turn positive starting 2028. We think the AI revenue/capex ratio will gradually recover starting 2027.

Our fair value estimate for Baidu is USD 113 per ADS, or HKD 110 per share.

We expect Baidu’s revenue growth to be driven primarily by its general business segment, which comprises both AI-powered and non-AI businesses. As of the second quarter of 2026, 40% of total revenue still comes from non-AI businesses, mainly search, which will weigh on Baidu’s near-term growth trajectory. We forecast a negative 15% CAGR for non-AI General Business revenue over 2026-30, reflecting a rapid erosion of search-related demand in China in the AI era. The decline will be largely offset by strong momentum in AI-powered businesses, which we project will deliver a 26% CAGR over the same period, driven by China’s substantial unmet compute demand. Overall, we expect total revenue to rise at a 6% CAGR from 2026-30. We also forecast net margin expansion to 8% in 2030 from 4% in 2025, supported by the scaling of AI-powered businesses and increasing operating leverage as AI demand accelerates. Net income is expected to grow at a 21% CAGR, benefiting from both margin expansion and a low earnings base in 2025 following significant asset impairments.

We assess Baidu as a no-moat company. Baidu’s online marketing services revenue, predominantly traditional search, accounted for 49% of revenue in the third quarter of 2025. This segment’s revenue fell 18% in the third quarter and was no longer disclosed in the fourth quarter. While Baidu still held an estimated 9% revenue share in China’s advertising industry in the first half of 2025, we think it will lose share to other apps and emerging generative AI chatbots over the next 10 years. Therefore, we do not believe the online marketing services/traditional search segment has a moat. Baidu’s AI cloud services for enterprises benefit from switching costs, but we estimate this business will account for a minority of total revenue at mid-cycle, insufficient to support an economic moat for the company as a whole.

Chinese users are increasingly searching within other app verticals, including social networking, short video, and interest-based communities such as Little Red Book. For instance, according to Zhoutian Financial, the daily search volume on Little Red Book doubled to 600 million between 2023 to 2024, compared with Baidu’s 1 billion.

As vertical-specific apps continue to flourish in China, more content and data are concentrated within individual ecosystems that Baidu cannot crawl. This makes Baidu’s search results less comprehensive and less relevant. The resulting decline in search usage reduces Baidu’s access to consumer data, which further wakens result quality, creating a self-reinforcing negative cycle.

As generative AI becomes more widely used, we expect Baidu’s share of search traffic to decline even more rapidly over the next five years.

Mobile phone–based AI assistants could further divert search traffic away from Baidu.

Both Douyin and Tencent, each with strong MAU for their AI apps and plugins, are larger internet giants with deeper data resources, broader ecosystems, and stronger cash flow generation. Their commitment to investing heavily in generative AI does not bode well for Baidu’s traditional search. In such a competitive environment, Baidu’s share of China’s advertising revenue could continue to decline.

For enterprises that build AI applications and models on Baidu Cloud and its self-developed Kunlunxin graphic processing units, switching to an alternative provider would entail significant costs and operational risks. Baidu charged egress fees for transferring data out of its cloud. In addition, engineers could require months to rewrite APIs, code, and programs to adapt to a new cloud environment, and the migrated AI models and applications may not achieve the same level of performance as they did on Baidu cloud. However, intense competition from large companies like Bytedance, Alibaba, Huawei and Tencent also poses a major headwind for Baidu AI Cloud. We expect that enterprise AI cloud services, where customer switching costs are substantial, will contribute only a minority of Baidu’s total revenue at mid‑cycle.

We view Baidu’s Kunlunxin as one of China’s leading GPU companies in terms of revenue scale (comparable to Cambricon) and customers quality (with clients such as Tencent and China Mobile). Kunlunxin stands to benefit from China’s push for self-sufficiency in AI hardware. However, disclosure remains limited, particularly around the business’s profitability ahead of its IPO. We will reassess Kunlunxin’s moat once more detailed information becomes available.

We assign a no-moat to Baidu’s remaining businesses, the most notable being its Robotaxi operations. The timing and scope of large-scale monetization remain uncertain. Government concerns about mass unemployment among drivers could delay the scaling of Robotaxis, especially in China.

Baidu becomes one of the top two AI chatbot providers in China, enabling large-scale monetization of AI-generated search results and driving a revival in online advertising revenue.

Baidu maintains its leadership in autonomous driving based on number of rides. Faster-than-expected issuance of autonomous driving licenses and accelerated monetization could serve as major growth catalysts.

Kunlunxin emerges as the undisputed GPU leader in China in both technological capability and revenue scale, providing Baidu with a durable long-term competitive advantage.

Baidu’s core search business declines faster than expected, putting significant pressure on the company’s cash position and resulting in insufficient investment capacity for AI initiatives.

Despite numerous growth initiatives, substantial uncertainty remains around whether Baidu’s new businesses can achieve mass-scale monetization. Failure to do so would create lingering margin pressure.

Kunlunxin loses its technology edge, leading to a rapid decline in chip orders and revenue for both Kunlunxin and Baidu’s AI cloud, which relies heavily on Kunlunxin’s processors.