China’s Corporate R&D: From Central Research Labs to a New Paradigm

Around 2024, China’s internet giants began to “withdraw” one after another from fundamental research in hard technology. Alibaba donated its entire quantum computing institute under DAMO Academy to Zhejiang University, and Baidu donated its quantum computing unit to the Beijing Academy of Quantum Information Sciences. These symbolic events raised a fundamental question for the industry: is the traditional “central research lab model” still viable for Chinese companies? An analytical paper by Sun Li and others published in Tsinghua Business Review examines this turning point in detail, organizes the evolution of global corporate R&D models into three stages, and presents new R&D paradigms suitable for Chinese companies.

The Golden Age Symbolized by Bell Labs

In the history of science and technology over the past half century, AT&T’s Bell Labs has been regarded as the benchmark for corporate R&D. With extremely high autonomy and abundant funding, relatively insulated from short-term commercial interests, it produced core achievements that laid the foundation of modern civilization, such as the transistor, laser, and information theory. The success of this model was rooted in AT&T’s monopolistic market position. Under government regulation, it could sustainably invest stable excess profits over the long term, forming a closed loop where basic research, systems engineering, and product development were closely linked.

For tech giants aiming to leap from “Made in China” to “Created in China,” establishing central research institutes similar to Bell Labs was the ultimate vision. In the 2010s, this vision began to materialize. Baidu’s Institute of Deep Learning (IDL), Tencent’s AI Lab, and Alibaba’s DAMO Academy were established successively, embodying Chinese companies’ technological ambitions to compete for influence in basic science.

The Prototype of Open Innovation Shown by

Xerox PARC

A representative example of the second stage is Xerox PARC. It gathered top computer scientists and produced disruptive outcomes such as the personal computer and Ethernet, but due to a disconnect between corporate strategy and technological vision, most of the technology leaked out. The combined market value of 11 spin-off companies reached more than twice that of Xerox. Xerox later attempted to participate in external projects through venture capital to gain returns, becoming a classic case of open innovation.

The lesson from this model is that if R&D achievements cannot be commercialized internally, much of the value will flow outward. Chinese companies’ central research lab models face similar difficulties in translating results.

The Industrialized Smart Science Embodied

by DeepMind

The third stage worth noting is the model of Alphabet’s DeepMind and Isomorphic Labs. After merging DeepMind with Google Brain, Alphabet pivoted toward commercialization and established Isomorphic Labs in the biomedical field. The core features of this model are threefold: the prediction of science, capitalization of R&D, and intelligent tooling. By vertically integrating computing resources, data, and algorithms, it builds a robust value-capture system, and future scientific outcomes may be transformed into pay-per-use cloud services.

This model can be considered an attempt to bridge the “valley of death” between basic research and commercialization using the power of AI and the cloud. In cases such as drug molecule design, processes that traditionally took months can be shortened to tens of minutes, making the industrialization of R&D a reality.

Structural Dilemmas of Chinese Companies’

Central Research Labs

Chinese companies’ central research lab models face three structural dilemmas.

First, a mismatch of financial cycles. China’s high-tech giants operate in a highly competitive internet market. After the traffic dividends peaked, profit margins have returned to normal levels. It is difficult for listed companies to sustain long-term projects like quantum computing, which require annual investments of hundreds of millions to billions of yuan and expect commercialization only after 2030. R&D investment must meet investors’ short-term return expectations, making it impossible to support research that does not generate positive cash flow for more than a decade.

Second, difficulty bridging the “valley of death” in technology transfer. Basic research results generally lack engineering and manufacturing process support, and business units under revenue pressure tend to exclude cutting-edge, high-risk technologies. Pilot testing and industrialization stages require substantial additional investment, but most companies lack the funds and willingness to continue investing.

Third, geopolitical and talent barriers have become apparent. After the intensification of US-China technology competition, advanced research equipment has become difficult to obtain due to export controls, and top overseas scientists face many barriers and legal risks when joining Chinese corporate labs. The global knowledge flows and free movement of talent on which central research labs relied have qualitatively changed.

These structural dilemmas explain why Alibaba and Baidu donated their quantum computing units to universities and research institutes. Balancing short-term commercial profits with long-term basic research is extremely challenging in the current competitive environment.

New Adaptive Paths for Chinese Companies

The paper presents three R&D paradigms for Chinese companies to adapt to the current environment.

Huawei’s Embedded Business-Oriented Model

Huawei’s 2012 Labs, with survival and security as its core driving force, focuses on applied basic research centered on core business. It achieves end-to-end transformation through early integration of R&D and products, along with internal profit-sharing. HiSilicon’s backup plan is a typical example. Although the barriers to this model are high, it offers a reference path for hard-tech companies.

AI for Science Platform-as-a-Service Model

This model is a light-asset, ecosystem-oriented path that upgrades traditional empirical science into AI-driven computable science. By building open-source, open platforms, companies aggregate global intelligence, forming a collaborative model where companies propose problems, the platform solves them, and outputs results on the cloud. This lowers the barrier to entry for innovation across society while creating new growth points. In China, the use of AI in healthcare is advancing, such as the organizational governance turning point indicated by the JCI Medical AI Guidelines, showing high compatibility with this model.

New Innovation Consortium Model

This model breaks down the “silo effect” among industry, academia, and government. The government provides top-down design, bears high-risk investment, and leading companies act as “problem proposers.” They collaborate with universities, research institutions, and upstream/downstream in the industrial chain to jointly tackle bottleneck technologies. Through mechanisms of shared demand, shared resources, shared results, and shared risks, they accelerate the transition from technological breakthroughs to industrial competitiveness.

Of particular note is the linkage with national-level projects such as China officially launching the space AI data center initiative. This consortium model combines government initiative with corporate execution, making long-term research feasible that would be too burdensome for any single entity.

Editorial Opinion

In the short term, China’s tech giants will accelerate the reallocation of resources away from extremely long-term basic research like quantum computing toward Huawei-style business-direct research or AI for Science platform models. From 2024 to 2026, more basic research units may be transferred to universities or national research institutes. Especially in geopolitically risky areas such as semiconductors, advanced materials, and quantum technology, government-led consortium-type research will become mainstream rather than corporate-only efforts. This restructuring will improve corporate R&D efficiency in the short term but carries the risk that basic scientific knowledge will be less accumulated within companies. From a long-term perspective, Chinese companies’ R&D model is likely to undergo a paradigm shift from “central research lab” to “distributed ecosystem.” AI for Science platforms rely on collaboration with open-source communities and universities, and their knowledge is provided as cloud services. If this model succeeds, a new business model will be established that recovers R&D investment returns through pay-per-use.

References

• 虎嗅網 “Chinese Corporate R&D Paradigm Shift” — Published 2026-06-22
• [Tsinghua Business Review (WeChat official account)—Author: Sun Li et al.]