Kuaishou, Baidu Split AI Units to Unlock Value

Kuaishou and Baidu are executing major structural changes by splitting off their primary AI divisions. This move signals a strategic pivot to unlock value for high-growth AI units within larger conglomerates.

The decision reflects the immense capital requirements of modern AI development. Traditional parent companies can no longer sustain the burn rates of cutting-edge generative models alone.

Key Facts: The AI Spin-Off Trend

• Kuaishou (market cap $29 billion) is spinning off Kling AI, valued at approximately $20 billion.
• Baidu (market cap $48 billion) is separating its core AI chip and computing unit, Kunlun Xin.
• The primary driver is accessing external financial resources independent of the parent company's slower growth.
• Chinese internet firms now operate under two conflicting valuation logics simultaneously.
• AI video generation requires massive infrastructure investment, creating a 'capital intensity' crisis.
• Global competitors like OpenAI face similar pressure, with Sora remaining restricted due to costs.

Valuation Logic Clashes Within Conglomerates

Traditional internet companies and AI startups command vastly different market multiples. Legacy platforms often trade at lower revenue multiples due to mature, saturated markets. In contrast, AI ventures attract premium valuations based on future potential and technological breakthroughs.

This discrepancy creates a 'conglomerate discount.' Investors struggle to price a single entity that houses both slow-growth legacy businesses and hyper-growth AI arms. By keeping these units together, companies inadvertently suppress the stock price of their most valuable innovations.

Kuaishou’s announcement highlights this tension directly. The $29 billion market cap of the parent company cannot fully reflect the $20 billion standalone value of Kling. Separating them allows each entity to be priced according to its specific risk and growth profile.

This structural adjustment is not merely financial engineering. It represents a fundamental shift in how technology firms organize themselves. The era of the all-encompassing super-app is giving way to specialized, agile AI entities.

The Capital Intensity of Generative Video

Generative AI, particularly video synthesis, is an expensive arms race. Training large multimodal models requires unprecedented computational power. The cost of inference and training continues to outpace revenue generation in the early stages.

Even well-funded giants like OpenAI have struggled with this dynamic. The initial release of Sora was delayed repeatedly. When it finally saw limited deployment, it operated as a constrained service rather than a mass-market product.

The competitive landscape is intensifying rapidly. Recent launches from major players include:

• Google Veo-3.1: Offering high-fidelity video generation with improved temporal consistency.
• ByteDance Seedance-2.0: Leveraging TikTok’s ecosystem for rapid user adoption and feedback loops.
• Kling AI: Competing on resolution, duration, and physical realism in generated clips.

These developments require continuous hardware upgrades. Without independent funding streams, parent companies may hesitate to invest aggressively. A spin-off allows the AI unit to raise dedicated venture capital or pursue IPOs specifically for R&D.

Strategic Independence for Kunlun Xin

Baidu’s decision to split Kunlun Xin follows a similar logic but focuses on infrastructure. Kunlun Xin develops AI chips and cloud computing services tailored for deep learning workloads.

Hardware and software development have different capital cycles. Chip design requires long-term, heavy upfront investment. Software applications benefit from faster iteration and network effects. Combining them dilutes focus and obscures financial performance.

By becoming an independent entity, Kunlun Xin can partner more freely with other tech firms. Currently, many potential clients hesitate to use Baidu’s proprietary chips due to competitive concerns with Baidu’s search and AI apps.

Independence removes this conflict of interest. It positions Kunlun Xin as a neutral infrastructure provider. This mirrors the strategy of NVIDIA, which supplies GPUs to virtually every AI competitor without direct application-level competition.

For Baidu, this separation clarifies its core business. The parent company can focus on search, autonomous driving, and consumer AI products. Meanwhile, Kunlun Xin can scale its chip business across the broader Chinese tech sector.

Implications for Global AI Markets

This trend in China has significant implications for global investors and developers. It suggests that vertical integration is reaching its limits in the AI era. Specialization is becoming the preferred model for scaling complex technologies.

Western companies should monitor these developments closely. If Chinese AI units succeed as independent entities, they may accelerate innovation through focused capital allocation. This could intensify competition in areas like video generation and AI hardware.

Developers and enterprises benefit from this specialization too. Independent AI units often offer better APIs, more transparent pricing, and dedicated support. They are incentivized to build robust ecosystems to attract external users.

However, risks remain. Spin-offs can lead to fragmentation. Developers may need to manage multiple relationships instead of a single vendor contract. Compatibility issues between separate hardware and software stacks could emerge.

Future Outlook: Fragmentation vs. Integration

The next 12 to 24 months will test this new organizational model. Success depends on the ability of spun-off units to raise capital independently. If funding dries up, these units may struggle to keep pace with integrated rivals.

We expect to see further consolidation in the AI infrastructure layer. Smaller players may merge to achieve the scale necessary for competitive chip or model development. Large parent companies will likely retain minority stakes to benefit from upside potential.

Key trends to watch include:

• IPO Activity: Potential public listings for Kling AI or Kunlun Xin to validate valuations.
• Partnership Shifts: Increased collaboration between formerly internal units and external tech giants.
• Regulatory Scrutiny: Antitrust bodies may examine whether spin-offs create unfair competitive advantages.
• Talent Migration: Key engineers may move between parent companies and new independent entities.

Ultimately, this restructuring acknowledges a simple truth. AI is too expensive and too fast-moving to be buried inside traditional tech conglomerates. The industry is evolving toward a modular architecture where specialized units drive innovation. This shift will define the competitive landscape for the remainder of the decade.