The Investment Dilemma: The “Must-Play” Game of the Era.
AI has emerged as the defining strategic priority and a primary target for capital allocation across industries. From pharmaceuticals to retail, companies are aggressively deploying Generative AI and automation at a velocity that outstrips any technological shift of the past two decades. Yet, despite the massive influx of capital and talent, companies that can demonstrate tangible financial returns on their balance sheets remain a minority.
A 2024 survey by MIT Sloan Management Review and BCG reveals a stark reality: only 11% of companies report achieving ‘significant financial benefits’ from their AI initiatives. Conversely, over 70% remain in the pilot or limited adoption stages. The implication is clear—while the scale of investment is expanding, the tangible returns remain largely unquantified.
Despite this, the investment frenzy continues. According to Stanford HAI (2025), global private AI investment hit $252.3 billion in 2024, up 44.5% year-over-year, with adoption rates jumping from 55% to 78%. Regardless of unclear near-term ROI, remaining absent from the AI race is simply not an option for any major player.
Underlying this surge is a fundamental paradox: If immediate returns are scarce, why is AI the one strategic arena where absence is simply not an option?
To decode this phenomenon, we must examine two theoretical frameworks that illuminate the essence of the lag between technology and returns: the Productivity J-Curve and the General Purpose Technology (GPT) Investment Cycle.
The Challenge of GPT: Navigating the J-Curve Valley for Long-Term Compounding
In every General Purpose Technology (GPT) revolution—from electricity and computing to the internet and AI—early investors have navigated the same phenomenon: a short-term dip in productivity and lagging financial returns. Yet, as technology permeates and organizations restructure, both productivity and profit curves eventually accelerate upward.
This is the essence of the ‘Productivity J-Curve’ (Brynjolfsson, Rock, & Syverson, 2019). Early adoption triggers heavy hidden costs—such as process redesign and data governance—that drag down short-term metrics. Yet, as integration deepens, the curve inflects upward, delivering value that vastly exceeds the capital committed.
AI currently sits in this trough. Enterprises are still learning how to orchestrate AI-human decision-making synergy, transform knowledge assets into proprietary language models, and establish enterprise-wide frameworks via AI agents. In other words, what appears to be a financial loss today is actually the construction of deep foundational infrastructure for future ‘Organizational AI Transformation.
History is poised to repeat itself :
- The semiconductor industry, following early losses in the 1970s and 80s, transitioned into an era of hyper-profitability driven by Moore’s Law, but only after process standardization and the division of labor had fully matured.
- Similarly, the Internet industry, in the wake of the 2000 dot-com bubble burst, left behind critical infrastructure and ingrained user habits—foundations that ultimately gave rise to giants like Google, Amazon, and Facebook.
Today, AI stands at a similar dawn, mirroring the early days of electricity and the internet.True, substantial profitability will not materialize during the initial adoption phase. Instead, it will unfold fully only after technology stabilizes, complementary assets mature, and a robust ecosystem is established.
From Rational Bets to Ecosystem Strategy: Why Pharma Giants Can’t Miss Out
Corporations and VCs are not chasing trends blindly; they are placing rational strategic bets. Drawing on Bowman & Hurry’s (1993) Real Options Theory, early investment serves as a ‘call option’—an entry ticket into a future where uncertainty is high but potential rewards are exponential. The value of AI is so immense that the risk of ‘missing out’ far outweighs the risk of ‘investing wrong.’ This mindset is particularly evident in the biotech and pharmaceutical sectors.
According to McKinsey (2024), over 70% of global pharmaceutical companies have integrated AI into R&D and process control.
- Moderna leveraged AI for mRNA design as early as 2016, slashing its preclinical development cycle by approximately 50% (Moderna, 2023).
- Insilico Medicine leveraged generative AI to design an anti-fibrotic drug, compressing the timeline from target identification to Phase I trials to just 18 months—one-third of the traditional duration.
Insilico Medicine leveraged GenAI to cut the drug discovery process to just 18 months—a third of the usual time. Though short-term revenue is not yet visible, these investments are building essential AI-driven competitive barriers.
Capital markets are following suit. With global AI funding hitting $109.1 billion in 2024 (Stanford HAI, 2025), giants like a16z and Sequoia are launching dedicated AI funds. Their bet is on the entire ecosystem’s maturity—from drug discovery to cloud computing—committing to a long-term cycle of 7 to 10 years.
Microsoft CEO Satya Nadella has observed that ‘this generation of AI will reshape every software category and every business.’ Kai-Fu Lee likens AI to the ‘New Electricity’—initially costly with delayed returns, yet once ubiquitous, every industry will be compelled to ‘plug in.’ NVIDIA’s Jensen Huang goes further, asserting: ‘AI is not merely a product; it is the engine of the modern industrial revolution.
For enterprises and investors, the return on AI is not measured in immediate revenue, but in the establishment of future structural positioning. Victory in this war will not go to those who profit first, but to those who complete their strategic deployment while navigating the trough.
[Founder’s View] The Prerequisite for Crossing the J-Curve Hinges on “Complementary Assets and Knowledge Fusion.
Success in AI investment hinges on the ability to ‘weather the trough and deepen integration.’ Here, ‘knowledge’ encompasses both existing global domain knowledge and unique internal expertise.
The ‘fusion of knowledge and complementary assets’ entails synthesizing experience, data, behaviors, and documentation—internal and external—using Generative AI and Deep Learning to construct proprietary enterprise models. These models are then operationalized via AI agents across all domains to accelerate every aspect of business operations.
As Brynjolfsson (2019) demonstrates, productivity experiences a non-linear surge only when technology, processes, and organization resonate. Companies that successfully cross this trough typically possess two prerequisites:
- Complementary Asset Integration
Successful enterprises invest simultaneously in comprehensive platforms that encompass data collection, analysis, modeling, knowledge fusion, talent development, and quality management. By embedding AI directly into these platforms, they position AI as the core of the process rather than a mere peripheral tool. This aligns with McKinsey’s (2024) finding: companies that integrate AI into their core workflows achieve productivity gains 3.5 times greater than those relying on point solutions. - Knowledge Integration :
The true value of AI stems from modeling, learning, exploration, and application that bridge internal and external organizational boundaries. Roche, for example, established an AI Center of Excellence (CoE) and a cross-functional Knowledge Graph platform. This infrastructure transforms R&D data into decision intelligence, boosting efficiency in early-stage clinical trials by 40%. By converting ‘tacit knowledge’ into explicit, shareable data structures, enterprises can successfully escape the pilot trap.
The key to AI investment lies not in ‘how fast you run,’ but in ‘how deep you integrate.’ The winners will be those who fine-tune technology and digital data from their knowledge systems into proprietary corporate language models, and transform experimental data and processes into deep learning models. These are the enterprises that will be the first to emerge from the trough of the Productivity J-Curve.
Conclusion: AI Transformation is a Long-Term Endeavor Redefining the CDMO’s Knowledge Base and Profit Profile.
The wave of AI investment has shifted from a technology-driven focus to the reshaping of organizational and industrial structures. The current phase of “high investment, low return” represents the most critical incubation period in the development of General Purpose Technologies (GPT). The real challenge for enterprises is not how to achieve quick results, but how to build a knowledge ecosystem capable of unlocking sustained value.
For pharmaceutical CDMOs, AI is not merely an investment in algorithms, computing power, or automation; it is an innovation centered on knowledge and intelligent pharmaceutical manufacturing. For investors, AI is no longer just a market theme, but a structural shift in long-term profitability. As Kai-Fu Lee and Jensen Huang remind us: “AI is the momentum; it is the engine
Therefore, AI transformation is not a mere technological buzzword, but a long-term strategic campaign to reinvent profitability through innovation. Only those enterprises and investors who persist through the downturn to continuously build core knowledge advantages will successfully traverse the trough of the J-Curve, reaping the long-term compounding dividends unleashed by this innovation.
References
Bowman, E. H., & Hurry, D. (1993). Strategy through the option lens: An integrated view of resource investments and the incremental-choice process. Academy of Management Review, 18(4), 760–782.
Brynjolfsson, E., Rock, D., & Syverson, C. (2019). The productivity J-curve: How intangibles complement general purpose technologies. American Economic Journal: Macroeconomics, 11(1), 333–372.
Huang, J. (2024). NVIDIA GTC keynote: The industrial revolution of AI. NVIDIA Corporation.
Li, K. F. (2023). AI 2041: Ten Visions for Our Future. CITIC Press Group. [李開復 (2023)。《AI 2041：預見未來二十年》。北京：中信出版社。]
McKinsey & Company. (2024). The state of AI in 2024: GenAI adoption accelerates. McKinsey Global Institute.
MIT Sloan Management Review & Boston Consulting Group. (2024). Achieving individual and organizational value with AI.
Moderna. (2023). Annual report 2023. Moderna Inc.
Nadella, S. (2023). Microsoft annual shareholders letter. Microsoft Corporation.
Stanford Institute for Human-Centered Artificial Intelligence. (2025). AI Index Report 2025. Stanford University.
World Economic Forum. (2024). AI governance frameworks for industry adoption.
Michael Han
Founder of therapiAI, is dedicated to the frontiers of PharmaTech. In this column, he explores global ADC markets, CDMO digital transformation, and AI innovation. Follow us for expert insights and actionable strategies to navigate the future of healthcare.
