This piece says pharma's R&D returns are near zero—most drugs never recoup costs, and the industry relies on a few blockbusters like Humira ($20B/year). New therapies (CAR-T, gene therapy) are so complex that Bristol-Myers Squibb's CAR-T unit has as many employees as treated patients (4,000 each). AI helps more with paperwork than drug discovery. Key picks: Novo Nordisk—went from a boring insulin maker to the 15th largest company via GLP-1 drugs; Bristol-Myers Squibb (Celgene)—its CAR-T complexity shows the scaling challenge; Humira (AbbVie)—a $20B blockbuster that keeps the system running.
Alex Telford discussed innovation in the pharmaceutical industry on the Invest Like the Best program. The core view is that drug development efficiency urgently needs improvement, with future directions including breakthroughs in gene therapy and AI applications. He analyzed the complex process from
Alex Telford is the founder of Convoke, a platform that simplifies drug development and commercialization processes; he has been writing a biotech industry blog since 2019. The main thread of this episode: the pharmaceutical industry is at an inflection point where "low-hanging fruit is exhausted" and "next-generation therapies are emerging," with innovation efficiency and regulatory balance as its core contradictions. The most weighty judgment of the entire episode: Alex Telford argues that the actual R&D return rate in the pharmaceutical industry is near zero, most companies do not make money, and the entire industry relies on a handful of "lottery-style" blockbuster drugs to cover the massive cost of failures—and this structure is being challenged by the increasing complexity of new drug modalities (e.g., cell therapies, gene therapies).
Alex Telford argues that the history of the pharmaceutical industry is a cyclical history of "low-hanging fruit being picked first, then new tools opening up new frontiers."
> "If you recall, Janssen had a gut feeling that the compound pethidine could be improved to make a better drug—but he didn't know exactly how to improve it optimally. He was just trial and error." — In other words, Janssen's "intuition-driven trial and error" was the core model of early discovery.
Alex Telford believes that the most impactful technologies over the next 10–20 years are already in their infancy, and share a common trait: "the drug itself is a complex process," rather than a traditional oral tablet.
Four emerging categories (ordered by complexity and industrial maturity):
| Category | Mechanism | Representative Cases/Data | Key Challenges |
|---|---|---|---|
| Monoclonal Antibodies | Artificially produced antibodies that bind to specific proteins to inhibit or kill targets | Humira (~$20 billion annual revenue, anti-TNF-α inflammatory drug); Keytruda (currently the world's best-selling drug) | Manufacturing process is relatively mature (cultured in bioreactors) |
| Gene Therapy | Uses viral vectors to deliver a functional copy of a gene into cells to compensate for a defective gene | Zolgensma (treats spinal muscular atrophy; severely affected children typically die before age 1, now can be nearly cured) | Delivery technology (viral vectors) is the core bottleneck; only a few successful commercial cases |
| Cell Therapy (CAR-T) | Immune cells are extracted from the patient, genetically edited, and reinfused to attack cancer cells | After Bristol-Myers Squibb (百时美施贵宝) acquired Celgene, the number of employees (~4,000) equals the number of patients treated (~4,000) | Manufacturing is extremely complex: collect cells → air freight → gene editing → air freight back for infusion, all must be completed in a very short time |
| Radiopharmaceuticals | Targeting element + radionuclide to selectively kill cancer cells | Several companies have made large-scale acquisitions in recent years | Short half-life (e.g., 7 days); manufacturing → delivery must be completed within days, otherwise the drug becomes ineffective |
> "These complex therapies are the moats that pharmaceutical companies are building—much harder to replicate than traditional pills. But their delivery itself is an extremely complex product-as-process." — In other words, the innovation direction shifts from "molecular design" to "system design and delivery."
Alex Telford 对 AI 在制药领域的应用持「短期悲观、长期乐观」态度，且认为 AI 在流程自动化上的价值大于在药物发现本身。
> 「发现药物方面的能力，并不是由我们设计分子的能力所限制的——它更受限于下游的临床开发和测试，以及我们为批准上市所需收集的信息。」——意即真正瓶颈在临床验证端，而非分子设计端。
Alex Telford argues that the biggest challenge for regulators is balancing 'keeping harmful drugs off the market' against 'missing the chance to save lives due to overly slow approvals' — and that different diseases require different standards of evidence.
Alex Telford argues that the pharmaceutical industry is essentially a 'lottery-like' economy, with revenue following an extreme Pareto distribution, and that this structure is being threatened by companies' tendency to 'prioritize accounting over science.'
| Indicator | Data |
|---|---|
| Proportion of approved drugs that cannot recover average development costs | 55% |
| Share of global pharmaceutical revenue from blockbuster drugs (annual revenue >$10 billion) | 30-40% |
| Number of active blockbuster drugs globally | Approximately 170 |
| Multiple of US drug prices (net) vs. Europe | Approximately 2x |
| US share of global pharmaceutical market revenue | ~40% (within 10 years of new drug launch, ~60%) |
| Pharmaceutical company net profit margin (after all costs) | Approximately 10-20% |
| Target | Guest Attitude | Key Data |
|---|---|---|
| Novo Nordisk | Not explicitly stated (mentioned as a positive example) | Once considered a "boring European insulin company," later rose to become the world's 15th largest company due to GLP-1 drugs; GLP-1 R&D spanned decades |
| Bristol-Myers Squibb (acquisition of Celgene) | Risk warning (manufacturing complexity) | In CAR-T business, number of employees (~4,000) ≈ number of treated patients (~4,000), reflecting process complexity and labor intensity |
| Humira (AbbVie) | Not explicitly stated (used as a representative blockbuster) | Annual revenue ~$20 billion, anti-TNF-α antibody, listed as a typical "super winner" |
| Keytruda (Merck) | Not explicitly stated (used as the best-selling drug) | Currently the world's best-selling drug, monoclonal antibody |
| Zolgensma (Novartis) | Not explicitly stated (used as an early success of gene therapy) | Treats spinal muscular atrophy, severely affected children usually die before age 1, now can be nearly cured |
1. "The actual R&D return rate in the pharmaceutical industry is close to zero, and most companies are not profitable" (Alex Telford) — Industry revenue follows an extreme Pareto distribution: 55% of drugs fail to recover their average development cost, while 30-40% of revenue comes from only about 170 blockbuster drugs.
2. "Drug discovery is not limited to designing molecules, but to downstream clinical development" (Alex Telford) — Even if AI perfectly designs a molecule, it still requires 10 years of clinical trials for validation. The short-term value of AI lies more in process automation (e.g., patient screening, regulatory document preparation) than in molecular design.
3. "The next wave of innovation shifts from 'pills' to 'processes' — CAR-T has as many employees as patients" (Alex Telford) — In Bristol Myers Squibb's CAR-T business, 4,000 employees serve 4,000 patients, illustrating that complex therapies are becoming a new moat for pharmaceutical companies, but the difficulty of industrialization has steeply increased.
4. "If you cut down the tallest ears of grain, you destroy the entire system" (Alex Telford) — U.S. drug prices are 2 times the global average, but high prices are a necessary condition for the extreme revenue distribution: retaining the super winners (e.g., Humira with annual revenue of $20 billion) is essential to sustain the industry's willingness to invest.
5. "Until we find surrogate endpoints for aging, we will not get drugs that extend lifespan" (Alex Telford) — Surrogate endpoints (e.g., tumor shrinkage as a surrogate for survival) can dramatically shorten development timelines, but chronic complex diseases (such as aging) lack reliable surrogate endpoints, which is a bottleneck in R&D.
6. "The biggest long-term risk is not overly strict regulation, but companies shifting from 'science-driven' to 'accounting-driven'" (Alex Telford) — Companies tend to invest in hard-to-replicate complex therapies (radiopharmaceuticals, CAR-T) to extend patent lifecycles, rather than betting on uncertain basic research — "Science that takes 20-30 years of trial and error to yield results is being squeezed by the logic of quarterly earnings."
7. "The true value of accelerated approval for HIV was not those drugs themselves, but that it kickstarted the iterative path" (Alex Telford) — Early AZT/DDI were not highly effective, but by generating initial revenue and market, they incentivized subsequent iterations, ultimately leading to highly effective therapies. This reflects the industry logic of "first get the wheels turning."
8. "Eroom's law (the reverse of Moore's law) is real — the pace of learning is too slow" (Alex Telford) — The iteration cycle in pharma is in stark contrast with the tech industry: tech can complete A/B testing in days, while pharma clinical trials take 10 years. Any technology that accelerates the feedback loop (e.g., automated trial-and-error, digital twins) is highly valuable.