From imagination to productivity: AI's path to scale

ByBelunn Se

From June 23 to 25, Dalian is hosting the 17th Annual Meeting of the New Champions of the World Economic Forum – better known as "Summer Davos."

For years, this gathering has served as a bellwether for the world's next growth frontiers, and this year's theme, "innovation at scale," carries a pointed message: Global innovation is shifting from technological breakthrough to value realization, from proof of concept to mass deployment – from chasing imagination to building productivity.

Notably, capital has poured into frontier fields like artificial intelligence (AI) at a staggering pace, with valuations and funding rounds repeatedly breaking records. Yet the conversion rate from experimental technology into replicable, scalable applications has lagged behind the surge in investment.

That gap, however, is beginning to narrow. As technical bottlenecks are gradually overcome, as costs continue to fall, and the supporting industrial ecosystems mature, a more pragmatic path to commercialization is taking shape – one driven by scenario-specific deployment.

From tool to infrastructure

AI is no longer just a productivity aid; it is becoming a foundational infrastructure for the digital economy. In advanced manufacturing, AI is rebuilding entire production systems, coordinating everything from production management and quality inspection to logistics in real time, replacing processes once governed by human judgment and manual inspection.

In drug discovery, AI's impact is arguably more historic. Traditional pharmaceutical research and development (R&D) is notoriously inefficient – researchers often need to screen millions, even hundreds of millions, of candidate molecules, a process that can take a decade.

AI excels precisely at finding patterns in vast datasets, rapidly analyzing chemical structures, biological data and clinical information. Today, according to some industrial estimates, AI-assisted drug discovery can cut discovery timelines by roughly 80% and R&D costs by about 70%, with the number of AI-designed drugs already in the pipeline quickly expanding. More significant still, AI can identify molecular structures that have historically eluded human researchers, opening entirely new therapeutic paths.

A similar story is unfolding in materials science, where screening and validation work that once took years – sometimes over a decade – can now be dramatically compressed, with major implications for breaking through critical technology bottlenecks.

What matters most is how this technology is spreading: from "showcase" to ecosystem-wide diffusion. As leading firms complete their intelligent upgrades, the benefits cascade down supply chains to smaller enterprises. Industrial intelligence is moving from a handful of corporate showpieces to becoming a feature of the broader economic landscape.

The path to scale: From 'can it work' to 'does it pay'

How does innovation actually translate into tangible business growth? A clear pathway is emerging: from technical validation to scenario fit, from cost reduction to ecosystem-wide replication.

Most frontier technologies can clear the first hurdle, technical validation, under controlled conditions. The real test begins at the second stage, scenario fit, where a technology must be embedded into existing management systems, supply chains and talent structures, not just demonstrated.

The third stage is cost reduction: Scale only becomes commercially viable once deployment costs fall within a payback period businesses can accept.

The final stage is ecosystem replication. Once a flagship case proves itself, the question becomes whether the surrounding ecosystem – the technology, talent, supply chain and standards – can be rapidly copied across companies and regions. That determines how far an innovation can ultimately travel.

At its core, this is a return on investment (ROI) calculation. And it is the deeper logic behind this year's "scale" theme: Capital is no longer paying for imagination. It wants every dollar invested to be measurable, verifiable, and repeatable.

When AI gets a body

If large language models completed AI's cognitive revolution in the digital world, the rise of embodied intelligence marks its transition into the physical one. The logic is straightforward: Large models have endowed machines with genuine comprehension, reasoning, visual perception, autonomous decision-making and motion planning. These capabilities combined are laying the foundation for the age of humanoid robotics.

Over the past two years, leading players worldwide have rushed in – from Tesla, Figure, and Agility Robotics to China's Unitree, AgiBot and UBTECH. In 2025, Chinese robotics firms emerged as the largest producers of humanoid robots worldwide.

Behind that lead is a distinct industrial logic: mass production at scale, paired with scenario-specific adaptation.

Drawing on a manufacturing base that spans the entire industrial chain, Chinese firms have pursued a "hardware-first, software-iterates" model – deploying production capacity rapidly and expanding the surrounding ecosystem in step. That approach keeps driving down hardware costs and lifting mass-production efficiency, and it's what has put Chinese players at the front of rolling out humanoid robots in structured industrial settings first by refining their products in live environments – an iterative, "learn while deploying" approach that moves far faster than closed-door lab research.

Meanwhile, technology bottlenecks such as battery life (as most units still require a swap every two to four hours), dexterity in handling irregular objects, and autonomous decision-making in complex settings remain real technical bottlenecks. But they are being closed at a visible pace.

The commercialization pathway is following a clear, scenario-driven sequence tied to ROI. From now through 2027, factories are expected to serve as the proving ground – automotive assembly lines, electronics manufacturing and warehousing/logistics. From 2028 to 2030, adoption is likely to expand into broader business-facing services, commercial services, specialized industrial tasks and elderly care. After 2030 comes the final destination: the home, the true mass-market opportunity.

Morgan Stanley projects the global humanoid robot market could surpass $5 trillion by 2050, with deployments hitting one billion units, and with China expected to have the highest number of humanoid robots in use.

Accelerating translational efficiency: Building a social ecosystem for industrial maturity

History tends to rhyme. The internet, the smartphone, electric vehicles – every transformative technology endured a long stretch of skepticism before its breakthrough. AI's development has never followed a straight line; it moves in waves of hype and disillusionment. And every genuine wave of industrialization that follows a technical breakthrough must clear a harder threshold: organizational adaptation and social restructuring. This is precisely where today's conversion efficiency bottlenecks.

Deploying humanoid robots in a warehouse isn't simply swapping out one task; it requires systematically rebuilding workflows, management models, training systems and even labor regulations.

China's 'scalable innovation' in AI benefits the planet

Behind "innovation at scale" lies a profound shift in the logic of global economic growth. In this transformation, China is both a major participant and an undeniable driver – from the accelerating pace of AI adoption in manufacturing, to the rapid rise of its humanoid robotics supply chain, to the strength of its energy and power infrastructure in supporting AI computations, to its deep investments in drug discovery and new materials. China's industrial advantages and scale effects carry distinct strategic weight in the context of innovation at scale.

As capital grows more focused on real returns, as businesses prioritize industrial deployment over demonstration and as technology zeroes in on solving real-world problems, a new development era defined by artificial intelligence is taking shape – making every ordinary person able to enjoy and benefit from the technological dividends which will be the greatest blessing of the AI era.

Belunn Se, a special commentator on current affairs for CGTN, is a senior industry observer based in Shenzhen, Guangdong Province in China.