Why the end of AI is biopharma?
Regarding the future development direction of AI, the investment community has different opinions, some believe that it is the photovoltaic industry, some believe that it is the power industry, while NVIDIA believes that it is the field of biopharmaceuticals. In NVIDIA’s 2023 investment layout, in addition to the core algorithms and infrastructure construction in the field of AI, biopharmaceuticals have become a key area of its investment. Kimberly Powell, NVIDIA’s vice president of healthcare business, has made it clear: “If the computer-aided design industry can launch the first chip company with a market capitalization of $2 trillion, why can’t the computer-aided drug discovery industry spawn the next trillion-dollar market capitalization pharmaceutical company?”
It’s clear that NVIDIA is optimistic about the future of AI+BioPharma, and is aiming to create another NVIDIA-like success story. Why does NVIDIA have such high expectations for this field? The answer to this question requires a deep understanding of the core challenges and underlying logic of the pharmaceutical industry.
Anti-Moore’s Law
The rapid development of AI technology cannot be separated from the rapid progress of the integrated circuit industry. Gordon Moore, the founder of Intel, after years of in-depth study of the semiconductor industry, summarized the idea that the number of transistors on integrated circuits doubles about every 18 to 24 months. This is the famous Moore’s Law. Moore’s Law states that as technology continues to advance, processor performance will double approximately every two years and prices will drop to half of what they were. It is thanks to this continuous iteration and advancement of technology that computer technology has grown exponentially, enabling more and more amazing features.
However, unlike Moore’s Law in the semiconductor industry, trends in the biopharmaceutical industry show an “anti-Moore’s Law” phenomenon. For a long time, the development of innovative drugs has been known as the “law of two tens,” meaning that it usually takes ten years and costs one billion dollars to develop a new drug. However, these figures are now outdated, with the latest data showing that the average R&D cost of an innovative drug is around $2.6 billion globally, and the average R&D cycle is around 10.5 years. Long R&D cycles and high R&D costs make innovative drug development extremely risky, not only requiring large amounts of financial support, but also facing high failure rates. To make matters even more difficult, the difficulty of R&D continues to increase over time, as the cost of developing new drugs and targets rises and the return on investment diminishes as more drugs and targets are discovered.
In this anti-Moore’s Law environment, the threshold of innovative drug R&D is increasingly high and the investment risk is rising, making innovative drug R&D gradually become the preserve of large enterprises. However, the pharmaceutical industry is often driven by innovation at the margins, with many new technologies initially born in start-ups. It is only when these technologies are progressively validated that large companies begin to intervene and layout.
An apparent contradiction has developed between the industry’s laws of development and its driving forces, a highly unusual phenomenon that could lead to a bottleneck in the advancement of human medical technology, a situation no one would be happy to see.
The biopharmaceutical industry is in dire need of change, and the ever-increasing cost of innovative drug development provides room for such change. It is based on these two factors that NVIDIA strongly believes that AI technology has the potential to revolutionize the pharmaceutical industry.
Experience vs. intuition
The modern pharmaceutical industry is often compared to a walled city built on intuition. Although innovative drugs need to undergo rigorous and systematic clinical validation before they can be marketed, the discovery process is full of uncertainty. The entire drug discovery process resembles a funnel, with the number of successful candidate compounds progressively decreasing at each stage, from drug discovery, to preclinical validation, to clinical validation. Among them, the drug discovery stage is particularly difficult, as developers need to sift through countless compounds to identify about 10,000 potential candidates, and finally lock in the right compound after layers of screening. Discovering and identifying a particular target is not only serendipitous, but also complex and tedious to validate, and even more difficult to successfully localize to the right molecule.
Although generations of drug discovery efforts have standardized the process of drug discovery, when faced with a huge amount of compound data, R&D is still highly dependent on researchers’ intuition, and data can only provide limited help. The right direction of R&D is a prerequisite for success, and once the wrong route is taken, no matter how much effort is put into it, it may still be futile. This over-reliance on intuition is the root cause of the continued rise in traditional drug R&D costs.
To reduce R&D costs in the pharmaceutical industry, it is necessary to shift to a more data-driven R&D approach. Data is essentially accumulated experience, and its digitization does not mean the elimination of R&D failures, but rather that failures can be turned into a learning base for the next R&D. Through large-scale model training, AI can make the drug screening process faster and more accurate.
The development of innovative drugs can be likened to a Roguelike game; although each playthrough may seem random and the experience may be different each time, the difficulty of the next challenge can be gradually reduced through constant failure and data accumulation.
AI pharma is essentially a process of shifting from relying on expert intuition to relying on data feedback to find the optimal R&D path through continuous model training. This shift marks a shift from intuition to data, and a progress from emotion to reason. Especially in many disease areas that have not yet been conquered, relying on expert intuition may not be better than randomizing success, while continuous AI model trial and error is the most effective way to reduce the failure rate. ai pharma not only reduces r&d costs, but also significantly improves r&d efficiency.
Data resources are the most precious
Algorithms, arithmetic and databases constitute the three core elements of AI technology. In most AI application scenarios, algorithms are the key link, and although arithmetic power and databases are equally important, investors tend to pay more attention to the development of large model algorithms. However, in the biopharmaceutical field, the dominance of algorithms may not be so obvious. Unlike other fields, data resources in biopharmaceuticals are precious, often non-open-source, and are the core assets of major pharmaceutical companies. Both successful and failed results are obtained through high-cost clinical trials.
It is thus clear that in the AI pharma field, the database is the key to core competitiveness. Observing the popular AI pharmaceutical companies in China, many of them are transformed from CRO companies. Unlike traditional pharmaceutical companies, CRO companies have rich R&D experience, and although the R&D data belongs to Party A, they are able to accumulate a large amount of process data and methodology in the course of multiple R&D processes, which provides them with an advantage in constructing databases. Considering the non-open-source nature of biopharmaceutical data, the development of AI pharma could go in two directions. One is well-funded multinational corporations (MNCs), which have accumulated rich R&D experience and data for a long time and have begun to comprehensively lay out AI technology; the other is CROs mainly engaged in the transformation of AI pharma in China, which have a strong database construction capability, and what they lack is only the development of large model algorithms, and the problem of computational power can be solved by cooperating with tech companies such as AliCloud and TencentCloud The problem of computing power can be solved through cooperation with technology companies such as Ali Cloud and Tencent Cloud. The first model may be more difficult to realize in China because the domestic biopharmaceutical industry started late and lacks multinational companies that have been conducting innovative drug research and development for a long time.
In the next few decades, CRO companies will probably become the core assets of China’s AI pharma, while overseas it is mainly a competition between multinational companies, which are reluctant to open up their data to third parties, and even companies such as Nvidia can only participate in the biopharmaceutical field through investment. Currently, domestic AI pharmaceuticals are in the early stages of development, and can be roughly divided into three echelons. The first echelon includes companies that have been laying out AI pharmaceutical technology for many years, such as Chengdu Pilot, Hongbo Pharmaceuticals, Jingtai Technology, and Medicine Stone Technology, etc.; the second echelon is a company that has rich experience in R&D but has just started in the field of AI pharmaceuticals, such as WuXi Kantei, Medicilon, and Haoyuan Pharmaceuticals, etc.; and the third echelon is experienced but has not yet been laid out in depth in the field of AI by other CRO companies. In the AI pharmaceutical field, data is the most critical resource, and the value of database far exceeds algorithms and arithmetic power, which is why CRO companies can take the lead in the current domestic AI pharmaceutical industry.
Contact Us Now!
If you need Price, please fill in your contact information in the form below, we will usually contact you within 24 hours. You could also email me info@longchangchemical.com during working hours ( 8:30 am to 6:00 pm UTC+8 Mon.~Sat. ) or use the website live chat to get prompt reply.
| Compound Glucoamylase | 9032-08-0 |
| Pullulanase | 9075-68-7 |
| Xylanase | 37278-89-0 |
| Cellulase | 9012-54-8 |
| Naringinase | 9068-31-9 |
| β-Amylase | 9000-91-3 |
| Glucose oxidase | 9001-37-0 |
| alpha-Amylase | 9000-90-2 |
| Pectinase | 9032-75-1 |
| Peroxidase | 9003-99-0 |
| Lipase | 9001-62-1 |
| Catalase | 9001-05-2 |
| TANNASE | 9025-71-2 |
| Elastase | 39445-21-1 |
| Urease | 9002-13-5 |
| DEXTRANASE | 9025-70-1 |
| L-Lactic dehydrogenase | 9001-60-9 |
| Dehydrogenase malate | 9001-64-3 |
| Cholesterol oxidase | 9028-76-6 |