AI Pharmaceutical: The Decade-Long, Billion-Dollar Dilemma is Being Rewritten Line by Line by Code | MatwingsVenus™（晓鹜™）

Preface

The destiny of innovative drugs should not be written only as "burning money."

In the early spring of 2026, Matwings Technology announced the completion of over 200 million yuan in Series A financing, jointly led by China Petroleum Kunlun Capital and Shanghai Future Industry Fund. Capital cast a vote of trust with real money—AI is reshaping the underlying logic of drug development.

However, amidst the capital boom, turning technical potential into industrial strength remains a severe challenge. As Matwings Technology CTO Liu Hao insightfully pointed out: "AI design is only the first step; the real barrier lies in 'scaling up implementation.'"

Liu Hao, CTO of Matwings Technology

Traditional drug development has a harsh 'double ten rule' — ten years, one billion dollars, and a success rate of less than 10%. Behind these numbers lie the youth of countless researchers and the sunk costs of massive capital. Today, AI is attempting to rewrite this formula, and the key battleground of this transformation is precisely hidden within lines of code and sequences of proteins.

The Underlying Game of a Trillion-Dollar Market

To understand the value of AI in drug development, one must first see the size of the market. According to iiMedia Research, the global AI drug development market grew from $790 million in 2021 to $2.41 billion in 2025 and is expected to reach $2.99 billion in 2026. Although the Chinese market started later, its growth rate is astonishing — rising from 80 million yuan in 2020 to over 600 million yuan in 2025, a more than sevenfold increase in five years.

Some organizations even predict that generative AI has the potential to bring $60 to $110 billion in annual value to the pharmaceutical industry overall.

However, above these optimistic projections, a sword always hangs.

Integrating and deeply merging dry and wet experimental data is currently the most challenging core difficulty and is still the focus of intensive research. ‘Dry experiments’ refer to AI computations and predictions; ‘wet experiments’ are the actual biological validations conducted in laboratories. The gap between the two is precisely the final hurdle for AI drug development to move from 'looking promising' to 'truly useful.'

This is also why when a company dares to confront this hurdle directly and presents a solution, the entire industry takes notice.

The Breakthrough Pioneer Starting from 'Big Zero Bay'

Science advances

Minhang, Shanghai, a technological innovation hub known as the 'Big Zero Bay.' In 2021, an elite team from Shanghai Jiao Tong University founded a company here called Matwings Technology, focusing on AI-driven protein research and development.

Unlike many AI pharmaceutical companies that remain in the conceptual stage, Matwings Technology has a clear and pragmatic commercialization path. Its co-founder and CTO Liu Hao completed his systematic studies from undergraduate to PhD at Shanghai Jiao Tong University and chose the interdisciplinary field of 'AI bioengineering.' Under his leadership, the team developed a general artificial intelligence platform for protein function prediction and has developed dozens of protein products for more than 30 domestic and international companies, covering multiple fields including biomedicine, bioenergy, food processing, and medical aesthetics, with over 10 products already industrialized.

A highly persuasive example is Matwings Technology's collaboration with GenSci Pharmaceuticals. In just four months, they modified an alkali-resistant protein, increasing its alkali resistance fourfold and doubling its lifespan under extreme conditions with a pH of 13-14. This successfully achieved industrial-scale production of 5,000 liters, becoming the world's first industrialized case of a protein designed by a large AI model and saving the company over ten million yuan annually.

This is precisely the most compelling narrative of AI in pharmaceuticals—not lofty papers and patents floating in the clouds, but tangible benefits flowing on real production lines.

MatwingsVenus™ (Xiao Wu™): When Protein R&D Becomes a "Conversation"

In April 2026, the AI-driven protein R&D platform company Matwings Technology officially released the world's first conversational protein R&D intelligent agent: MatwingsVenus™ (Xiao Wu™). The name of this platform is meaningful: 'Xiao' represents insight and understanding, and 'Wu' is derived from the phrase 'the rosy clouds and the solitary wild duck fly together,' symbolizing tranquility and far-reaching vision. 'Matwings' suggests giving wings to the world of materials—allowing protein design to truly enter ordinary laboratories.

So, what exactly is MatwingsVenus™ (Xiao Wu™)?

In simple terms, it is an agent-centered, one-stop protein R&D platform that supports retrieval of hundreds of millions of real-labeled protein data points, integrating over 200 protein design tools, more than 50 platform-certified experts, and over 30 domain experts' optimized skills.

What truly sets it apart from traditional tools is its core capability of a 'conversational wet-dry closed-loop.'

Users only need to input the task goal in natural language—for example, 'design a heat-resistant industrial enzyme'—and the system can automatically break down the task, coordinate the corresponding design, prediction, analysis, and screening capabilities, completing the entire workflow from literature research to protein design, and then to automated wet experiments.

This is the so-called 'closed loop': after the AI agent completes protein design, the platform uses its self-built communication mechanism to import the results into the plasmid ordering and experimental scheduling process, automatically connecting subsequent experimental tasks, driving robots to complete sample preparation, protein purification, and functional testing, and finally feeding experimental results back to the next round of AI design, forming an iterative cycle of 'computation-driven wet experiments, wet experiments feeding back computation.'

'Design is verification, verification is iteration'—these eight words precisely capture the core challenge of integrating dry and wet experimental data.

To use a vivid metaphor: in the past, protein research and development was like a long 'handcrafted workshop' style creation, where scientists repeatedly experimented in the lab; Matwings Venus™ (XiaoWu™), however, is like equipping every researcher with a super intelligent assistant and a shared automated laboratory, transforming R&D capabilities that used to be accessible only to large enterprises and institutes into infrastructure that individuals can utilize.

Hong Liang, founder of Matwings Technology

Hong Liang, the founder of Matwings Technology, described his vision in a conversation: “Riding on the rise of 'crayfish,' we launched this intelligent system… It is not just a simple model, but a one-stop platform. You can search literature, patents, and market information to find the general direction you want to take, and then use our design technology and protein models to design the functional protein/enzyme you want. Once designed, you can directly call our robots to conduct experiments for you.”

Practical Validation: From 'Possible' to 'Feasible'

No matter how good a concept is, it ultimately needs results to speak for itself.

In the field of in vitro diagnostics, testing for pancreatitis requires a key material called 'maltose heptasaccharide.' For a long time, this raw material has been highly dependent on imports, with market prices as high as hundreds of thousands of yuan per kilogram, not only driving up testing costs but also forming an invisible supply chain barrier.

The key to breaking this barrier lies in modifying the enzyme that catalyzes its synthesis—but that is easier said than done. Under the traditional model, training an expert capable of independently modifying proteins takes 5 years, and the trial-and-error period to successfully modify a protein takes another 5 years. However, in this project, the enzyme modification design was completed by a physics PhD with no prior biology background. Without consulting any related literature, he simply input the target sequences provided by the client into the AI large model and successfully obtained enzymes that met the requirements.

This contrast is highly illustrative: once AI large models accumulate 9 billion cross-disciplinary protein sequence knowledge entries, the professional threshold for protein design is being redefined—'cross-disciplinary background' is no longer a barrier and might even become a source of innovative perspectives.

The results are equally solid. The performance of the modified tool enzyme was greatly improved, reducing the production cost of 'maltose heptasaccharide' from hundreds of thousands of yuan per kilogram to tens of thousands of yuan per kilogram, instantly breaking import dependence. Currently, based on the MatwingsVenus™ (Xiaowu™) platform, Matwings Technology has extended the AI design boundary to multiple industrial fields—from optimizing industrial enzyme thermal stability and catalytic activity, to achieving severalfold increases in alkalinity tolerance and activity surpassing the best comparable products of international leading companies. What traditionally would take years in R&D can now be completed in just a few months, and nearly 10 products have already been commercialized.

From the directed evolution of tool enzymes to multi-parameter parallel optimization of industrial proteins, MatwingsVenus™ (Xiaowu™) has completed the full process of AI design, experimental verification, and iterative results in multiple real projects—indicating that this platform has moved beyond the conceptual level and truly transformed the 'conversational wet-dry loop' into practical R&D capability.

One-Person Laboratory: The Future of Democratized Pharmaceuticals

One-person lab

Returning to the question at the beginning of the article: What exactly can AI bring to drug research and development? The answer may lie in the detail mentioned earlier — the enzyme used to successfully redesign a pancreatitis detection tool was modified by a physics PhD with no biological background. He did not review literature or rely on traditional experience; he merely input the target sequence into AI and completed protein modification that previously required '5 years of cultivation and 5 years of trial and error.' This is precisely the future represented by MatwingsVenus™ (Xiaowu™): when AI’s predictive capabilities are deeply integrated with automated execution capabilities, the 'one-person laboratory' model is moving from concept to reality. Individual researchers or small teams can start protein research and development using AI platforms; complex R&D capabilities that were once exclusive to large institutions are transforming into a 'shared laboratory'-style public service. Of course, this does not mean that drug development will become 'simple.' As Liu Hao, CTO of Matwings Technology, said: 'The value of technology lies not in breakthroughs in the lab, but in its ability to solve real-world problems.' Strict verification of Phase III clinical data, gradual implementation of regulatory frameworks, and the continuous accumulation of high-quality training data — AI drug development still faces numerous hurdles. But the direction is clear: AI is pushing drug development from a 'handcrafted workshop' to an 'engineered platform,' from an 'elite privilege' to an 'inclusive capability.' In this sense, MatwingsVenus™ (Xiaowu™) is not just a product; it is more like a signal — announcing the arrival of a 'personal innovation era' in protein R&D. When everyone can call on shared research resources through conversational AI and quickly turn their ideas into real products, the technological revolution that began in laboratories will truly reach the depths of industry. The ten-year, billion-dollar dilemma will not disappear overnight. But at least, with the support of AI, the 'birth certificates' of next-generation innovative drugs may feature fewer tragedies and more composure.