NextFin News — Baidu reported second-quarter revenue of about 31.3 billion yuan, down roughly 4 percent from a year earlier. Inside that total, the picture diverged.

AI-related activity—cloud infrastructure, applications and AI-native marketing—reached approximately 12.5 billion yuan and represented half of Baidu’s core operations. Intelligent-cloud infrastructure alone rose 50 percent to 7.3 billion yuan; GPU-cloud revenue grew more than threefold. Online marketing, once the dominant profit engine, continued to decline.

The same quarter showed the cost of that shift. Capital expenditure roughly tripled to 11.4 billion yuan. Net profit dropped sharply, in part because of a high year-earlier base in other income, but also because heavier depreciation and operating costs from the AI build-out are now flowing through the income statement. Operating cash flow remained positive; free cash flow was negative as investment accelerated.

The pattern is familiar across the industry: AI infrastructure and model development require large, sustained outlays long before commensurate profits appear. What distinguishes Baidu is the weight of the legacy business that still sits beside the new one. Search and online marketing generate cash and still contribute the bulk of near-term profit, yet they face structural pressure from changing user behavior and competition for attention. AI cloud and applications are growing quickly and now form half of core revenue, yet they demand continuous capital and still operate at lower margins during the scale-up phase.

Keeping both inside a single listed entity creates a practical tension. Investors who want pure exposure to AI infrastructure and models must also own a mature advertising franchise whose growth has slowed. Investors who value the cash generation of the older business must accept the earnings volatility and heavy spending of the newer one. The result is a blended multiple that often satisfies neither camp fully. Several pure-play or more narrowly focused AI companies have commanded higher growth valuations; Baidu’s shares have lagged the enthusiasm that greeted its early large-model releases in 2023.

Management has chosen to fund the AI expansion from the existing balance sheet and operating cash flow rather than through a structural separation. The company retains a substantial cash and investment position. In 2025 it raised debt in both renminbi and dollars, partly to refinance and partly to support compute capacity. In 2026 the emphasis has shifted toward deploying that capacity: the jump in second-quarter capital spending is the clearest signal yet that Baidu is trying to close the gap with peers that have spent more aggressively in earlier periods.

Whether the current structure is optimal remains an open debate among analysts. A separation of the AI cloud and model businesses could, in theory, allow that segment to be valued on its own growth and margins and potentially ease access to dedicated capital. It would also remove a large cash generator from the parent and create transition costs. Baidu has so far preferred integration—using search traffic and existing enterprise relationships as distribution for AI products—over a clean split.

The competitive backdrop leaves little room for half-measures on spending. Demand for GPU cloud and model serving is rising across financial services, gaming, industrial software and other verticals. Providers that cannot offer sufficient capacity or competitive pricing risk losing workloads that may prove sticky once deployed. Baidu’s own figures show GPU-cloud growth accelerating from an already high base, suggesting the constraint has been supply as much as demand. The tripling of capital expenditure is an attempt to relieve that constraint.

At the same time, the advertising decline is not cyclical in any simple sense. Users increasingly obtain information through conversational interfaces and short-form video; advertisers follow attention. Baidu has slowed monetization of AI-powered search features in order to protect user experience, accepting near-term revenue pressure. That choice reinforces the identity shift the company describes—from an internet franchise adding AI features to an AI company that still operates a large search and marketing business.

The second-quarter results therefore mark a milestone rather than a conclusion. AI has become the majority of Baidu’s core revenue. Capital spending has finally stepped up in a visible way. Profitability is absorbing the consequences. The market is still deciding how to price a company that is no longer primarily a search advertiser and not yet a pure AI-infrastructure pure play.

Until one of those identities becomes dominant in both the numbers and the narrative, the shares are likely to reflect the average of two different stories rather than the upside of either one alone.