AI-led drug discovery to create decade-long CDMO tailwind: Harini Dedhia

AI is a utility for pharma companies, delivering drug discovery in a shorter time frame with far higher chances of success.

Disclaimer: This article is written by Harini Dedhia, fund manager at Tamohara Investment Managers. Views expressed are her own. Readers' discretion is advised.

 Every now and then, a major theme grabs the attention of most investors. The persistent tailwinds make it impossible and foolhardy to ignore these themes. A big chunk of capital goes towards businesses that are eventually deemed utilities. Another portion goes towards the short-term cyclical winners that typically see a surge in demand as they enable the infrastructure buildout for the said “utility”. Initially, a small portion, and eventually the largest capital, ends up chasing those that utilise the rails built by the utility and build businesses enabled by those- the structural wealth builders. Let’s take the example of the internet. From Netscape to Firefox - none of the browsers that brought the internet to users made money. Microsoft broke the model by offering Internet Explorer free of cost on every installation of Windows to make their OS far more attractive to personal and enterprise buyers. In the internet era, the providers became the ultimate utility. No one today expects to pay anything for internet browsers. The most sustainable wealth creators of the internet era were those that harnessed the power of this utility: e-commerce (Amazon), digital ads (Google and Meta), digital content (Netflix). There was a short duration when those that laid the tracks for the internet proliferation- telecom providers, fibre optics manufacturers, compute power manufacturers- made wealth. Timing was, however, key for them. 

 

 The framework holds true for railroads as well. The crazy expansion in the late 1800s that resulted in creating the tycoons of the Gilded Age,  very soon found themselves bankrupt. Railroads in the US were deemed a utility, with the government mandating ‘rural coverage’ and regulating fares for a certain number of routes. All but one railroad tycoon survived the ‘utility-fication’ of the business. Steel companies, however, made cyclical and structural wealth as a result of railroads. Not only were they used in building modern tracks, but the network itself also enabled the industrialisation of the USA.  

Check - TOP GAINERS NSE | TOP LOSERS NSE

 The all-pervasive theme of the decade is AI. The consensus is wrong in judging India as an anti-AI play simply because we don’t possess a presence in the dominant LLM supply chain. AI requires buildout of data centers and supporting power infrastructure. There are a multitude of companies with export qualification catering to this cyclical demand- think transformers, electrical bushings, heat exchangers, etc. 

 There is, however, an industry whose course alters with AI - pharmaceuticals. AI will single handedly bring down the cost and the time to bring drugs to the market.   

 Source: Novartis, 2025 New drugs take 10-15 years to get approved and about $1.5-$2bn to get it past the USFDA approval line. The real cost of discovery is however much higher as there are a number of scientists that work on chemical and biological entities that never make it to clinical trials. The gap between >10,000 compounds to less than 250 commencing Phase 1 trial, takes another $3-4bn of large pharma research budgets. The true cost of discovery per drug therefore resides somewhere between $4-6bn. AI changes this. 

 - AlphaFold has mapped over 220mn protein structures, i.e. all protein structures known to man. Genes are proteins. AI has therefore enabled biomarker mapping

 - Instead of wet lab tests, AI simulations can turn genes on and off in a digital cell to see the impact of the target. This enables narrowing down projects for pharmaceutical companies and their researchers to bet on. Novartis narrowed down on 5 potential drug candidates to cure an autosomal polycystic kidney disorder by using AI to mine vast amounts of scientific literature, past clinical trial data and human gene structures, and then applying lab experiments to a shortlist of 12 candidates. This was all done within a year by a single team unlike the pre-AI era which would have stretched this project for a decade+ - Phase 1 trials which test for safety markers of the drugs, have seen AI simulations modelling the trials prior to wet labs resulting is over 25% lesser failures additionally

 AI is therefore a utility for pharma companies delivering drug discovery in a shorter time frame with far higher chances of success. The true cost of drug discovery might now sit between $2.5-$3bn vs. $4.5-6bn. This reduction in costs will spur more drugs to market, freed up time bandwidth for researchers will also increase the number of drugs to market. 

 CDMO (contract development and manufacturing) companies in India will enable this global revolution in pharma. The pandemic helped our companies prove their manufacturing prowess and speed of execution. The investment in further capacities and capabilities have made them candidates for sustained wealth creation. 

 This is the structural tailwind that will sustain for a decade. The winners within might rotate, the sector stays.