From the Industrial Age to the Digital Age: Rethinking R&D in a Platform Economy

John H Howard, Victor Pantano, 18 December 2025

Research and development is undergoing a fundamental transition from the industrial-era laboratory to digital platforms driven by data, software, and artificial intelligence. This shift transforms the nature of discovery from codified knowledge published in journals to executable knowledge embedded in code. As global technology firms consolidate control over the computational infrastructure of science, nations face new challenges regarding capability, governance, and sovereignty.

Australia sits in a precarious position within this landscape. While the public research sector is strong, it is constrained by a narrow industrial base and policy settings that reflect an earlier era of innovation. The rapid decline in sovereign high-performance computing capabilities exposes a strategic blind spot, risking a future where the nation becomes dependent on "licensed science" and proprietary black boxes to validate its own research.

To secure epistemic agency, Australia must recognise scientific data as a first-order strategic asset and invest in a scalable digital research commons. Adapting to this platform economy requires a shift in policy architecture to value software-intensive R&D and ensure national capabilities remain resilient in a global system.

---

Research and development (R&D) is being reshaped by a steady transition from industrial era laboratories to digital platforms supported by data, software, and AI. This is, in turn, driving a shift in the nature of knowledge production itself, from a system of codified knowledge, published in journals, towards executable knowledge embedded in software.

As cloud infrastructure, modular digital tools, and machine learning influence the way scientific discovery is conducted, there is an increasing presence of firms operating global digital R&D environments. This, in turn, raises important national issues about capability, governance, and sovereignty.

Australia occupies a precarious position within this landscape. It has a strong public research sector but is constrained by a narrow industrial base, exacerbated by policy settings that reflect an earlier era of innovation, built around physical laboratories and discrete R&D projects. As platform-based models reshape the research landscape, Australia has to consider how to adapt.

1.    Trends and Trajectories

The Digital Transformation of Research

The industrial model that shaped science and innovation policy for most of the twentieth century was anchored in physical laboratories and specialised equipment. Firms relied on internal scientific and engineering capabilities. Governments funded public research institutions to acquire these assets and advance public knowledge through programs like the National Collaborative Infrastructure Scheme and the Education Investment Fund.

State Government initiatives, through the Victorian STI initiative, the Queensland Smart State Strategy, and some generous philanthropy, also supported these investments. Taken together, this architecture supported significant advances in clinical procedures, vaccines, pharmaceuticals, materials, electronics, and manufacturing.

While still significant, this model no longer fully describes how a large share of modern research is organised or conducted. A shift towards platform-centred, data-driven, and AI-enabled research is reshaping the global landscape. These environments constitute R&D ecosystems for thousands of businesses, start-ups, researchers, and public institutions.

The scale of these platforms gives them significant influence over research methods, standards, and tools. This can shape the direction of innovation in fields ranging from biotechnology and materials to energy systems and advanced manufacturing. For Australia, this presents a significant challenge. Business R&D spending remains low, and many firms rely on imported technologies rather than building internal capabilities.

An Epistemic Shift: From Open Science to Licensed Science

This transition presents a challenge to the Scholarship of Discovery. Historically, public science funding favoured hypothesis-driven, human-led inquiry. The emerging model suggests an "automation of invention" where discovery is data-driven and validated by the performance of code rather than the status of the author.

Biology and materials science are increasingly treated as information science problems. While efficient, this creates a genuine risk of epistemic agency loss. If the tools of discovery are proprietary "black boxes" that researchers cannot inspect, the scientific method itself becomes compromised.

There is a risk of moving from a tradition of open science to a future of "licensed science"^. In this scenario, valid research would require a corporate subscription. If a private firm holds the keys to the microscope of the twenty-first century, the public sector risks losing the ability to verify and reproduce knowledge. This creates a sovereignty risk just as acute as a loss of manufacturing capability in an earlier era.

Platforms as the New Innovation Architecture

Digital platforms have become central to the performance of innovation ecosystems. A platform is a digital environment that coordinates interactions among users, data, tools, and applications. They provide the computational ground layer on which a growing share of research tasks is undertaken. In R&D, they integrate cloud computing, storage, collaboration tools, workflow automation, and AI services.

A few global technology firms now have the computational resources, datasets, and software environments that underpin advanced R&D. Companies such as Alphabet, Microsoft, and Amazon operate as de facto global laboratories. They own large cloud networks, AI supercomputers, proprietary models, and tightly integrated software platforms that support scientific computation and digital product development.

This concentration of capability reflects three defining features.

• Scale. Platforms pool infrastructure in ways that no individual laboratory or agency can achieve, sharing the cost of high-performance computing among global user bases.  

• Modularity. Platforms allow R&D tasks to be decomposed and recombined efficiently. Researchers can draw on libraries of models, datasets, and simulation environments to design and execute experiments rapidly.  

• Orchestrated interaction. Platform owners set the standards, interfaces, and protocols that govern how participants operate. This digital coordination influences what is possible, how data is handled, and how knowledge is shared or restricted.

Consequently, the coordination of R&D is shifting from traditional institutional structures to these digital architectures, where software protocols rather than physical facilities increasingly determine the pace and direction of scientific discovery.

Building on this platform architecture, digital technologies are now evolving beyond their traditional role as analytical support to become active participants in the acceleration of scientific discovery.

Digital Acceleration and the Automation of Discovery

Digital technologies have evolved beyond their original role as analytical tools to participate directly in the discovery process. Large language models and autonomous laboratories influence all stages of research, from hypothesis generation to validation. In chemistry, AI models identify promising molecular structures before physical synthesis is attempted. In genomics, machine learning predicts gene interactions with accuracy that reshapes drug discovery timelines.

Digital twins replicate physical systems, enabling real-time testing without the need to build prototypes. These tools reduce uncertainty and shorten development cycles. Autonomous laboratories combine robotics, sensors, and AI to conduct experiments continuously, adjusting parameters based on incoming data.

This shifts scientific labour away from manual tasks toward conceptual reasoning and model interpretation. Organisations with access to advanced computational resources iterate more quickly, and their productivity advantage compounds with each generation of tools.

Physical facilities are increasingly serving as validation nodes for digital models rather than as the primary sites of discovery. A physical experiment may only be conducted when a digital model requires calibration or final verification. Australia’s investment in physical infrastructure should be designed to integrate directly with these digital feedback loops, rather than standing apart as isolated assets.

These observations cannot be taken as rendering physical infrastructure obsolete. The platform economy still requires scale-up facilities and clinical trial sites. One cannot digitise the final validation of a new vaccine or the stress-testing of a new alloy. However, the relationship between digital and physical environments has changed. But there is also a strategic blind spot, as recent commentary from Australia’s Tier-1 high-performance computing facilities attests. This is addressed below.

2.    Issues and Implications

Sovereign Compute as a National Capability

While global firms are investing hundreds of billions of dollars in AI infrastructure, Australia’s national supercomputing capability has slipped rapidly down international rankings. Gadi, once the 24th fastest computer in the world, is now outside the top 150; Setonix has fallen from 15th to 49th in three years. Both are oversubscribed and constrained by supply-chain delays for critical GPU components.

This matters because certain categories of scientific and industrial research simply cannot be executed on commercial cloud platforms alone. The most demanding workloads in climate modelling, materials design, genomics, and physics-informed AI require tightly coupled CPU–GPU systems with low-latency interconnects, bespoke cooling, and sovereign control over data pathways. These are not merely computational luxuries; they are prerequisites for participation in the next wave of scientific discovery.

If domestic compute capability erodes, Australia risks becoming dependent on foreign infrastructure for foundational research tasks, a form of digital-era capability loss analogous to the offshoring of manufacturing in previous decades. Without locally controlled HPC–AI hybrid facilities, Australia cannot guarantee timely scientific outputs, nor can it ensure that sensitive or sovereign datasets remain within national custody.

The broader implication is that cloud access is necessary but not sufficient; countries that aspire to shape their own innovation trajectories require on-the-ground computational assets that are predictable, expandable, and aligned with national priorities.

The Missing Layer: Scientific Data as a Strategic Asset

An emerging insight from global research agencies is that the bottleneck in AI-enabled science is not only compute capacity but the structure and quality of scientific datasets. As a former Director of Los Alamos National Laboratory recently observed, governments hold vast archives of scientific data that are not well organised for training foundation models. Equally, without deliberate design choices, future datasets risk being generated in formats that are incompatible with machine learning.

This point goes to the heart of the digital research transition. AI systems for chemistry, materials, climate, energy, and geoscience depend on data that is standardised, richly annotated, and generated through validated physical instrumentation. Australia’s physical research infrastructure, from advanced microscopy and synchrotron beamlines to national CT imaging platforms and supercomputers, is an under-recognised strategic asset in this context. These systems produce the experimental observations required to build science-informed models, a category of AI that blends empirical data with the governing laws of nature.

If Australia wishes to build or adapt scientific foundation models rather than simply consume models developed elsewhere, it must treat scientific data generation and curation as a first-order policy issue. This requires metadata standards, interoperable formats, custodial architectures, and investments that ensure datasets are “AI-ready” from the moment of capture.

 In many fields, the dataset, not the algorithm, becomes the locus of competitive advantage. Some cutting-edge research groups in Australia, have now sort to own (through patenting for instance) the means of data creation, fusion and curation for specific applications. A national strategy for digital R&D must therefore recognise structured, sovereign scientific data as a core component of research infrastructure.

3.    Concluding Comment

The transition from the industrial laboratory to the digital platform marks a permanent change in the structure of discovery. Innovation is no longer solely the product of human intuition and physical experimentation; it is increasingly the output of executable knowledge, driven by massive computation and governed by software protocols.

For Australia, this transition presents a serious challenge. Continuing with current settings risks a slow drift into dependency, where national research capabilities are hollowed out and the validation of scientific truth becomes a service leased from global technology firms.

To avoid this loss of epistemic agency, policymakers must recognise that digital sovereignty is now as critical as physical security. This requires cultivating essential domestic capabilities that allow Australia to engage with global platforms on its own terms. The nation’s immediate priority must be to arrest the decline in sovereign high-performance computing and to leverage its substantial physical research infrastructure to generate the high-quality, standardised datasets that will define competitive advantage in the coming decade.

Ultimately, successful adaptation requires reimagining the national innovation architecture. It calls for a move beyond the grant cycles and capital works programs of the past towards a strategy that integrates skills, data, and digital infrastructure into a cohesive whole. By building a scalable digital research commons and treating scientific data as a core national asset, Australia can ensure it remains a maker of knowledge rather than a mere user, securing its economic future in a platform-driven world.

---

[1] This Insight draws on research undertaken on Australia’s R&D performance over many years, and concentrated in the early part of 2024, summarised in Australian innovation and the crossroads: The slump in national R&D since 2008—causes, consequences, and subsequent research, investigation, and analysis. Several relevant documents are located on the Acton Institute for Innovation Website. It also draws on recent Innovation Insight Atoms to Algorithms: Building Australia's New Innovation Infrastructure