AI in Healthcare

Today's Stories

1. 1

   AI drug discovery hits data wall as models need negative results, not just wins

   AI is shifting pharmaceutical research from physical screening of molecular libraries to computational prediction of drug candidates, but the approach is exposing a critical bottleneck—most publicly available datasets contain only positive results, leaving AI models trained on incomplete information that lacks the failed experiments needed to avoid bias and improve reliability. Bringing a new drug to market takes 10–15 years and costs $1 billion（約1600億円） to $2.5 billion（約4000億円） with failure rates upward of 90%; AI is the industry's biggest bet to reduce timelines and improve success rates. However, without access to negative data (compounds that don't bind, failed experiments), models cannot be adequately trained to make accurate predictions, and fabricated or manipulated data—easier to create with generative AI—could have "potentially disastrous consequences" when used to train models.

   No drug discovered primarily through AI-driven design has yet received full FDA approval, though Belcher expects that to change within the next two to three years. The future state is fully autonomous labs that cycle through prediction, testing, and optimization while feeding results back into AI models—but this requires interoperable lab systems and structured, comprehensive datasets that most labs do not yet have.

2. 2

   Pfizer, Medtronic, Hinge Health offer AI healthcare exposure

   An investment screener identified three publicly traded companies—Pfizer (market cap US$142.5b), Medtronic (US$105.0b), and Hinge Health (US$5.9b)—as stocks with meaningful exposure to artificial intelligence applications in healthcare, spanning drug discovery, surgical robotics, and digital musculoskeletal care. Healthcare AI addresses a structural need across systems facing cost pressure, with potential to improve diagnostics, treatment decisions, and productivity. Pfizer's AI-powered R&D partnerships and drug pipeline, Medtronic's AI-guided surgery and robotics (Hugo system), and Hinge Health's motion-tracking digital platform and Enso wearable each represent different entry points to this theme for investors seeking long-term structural trends rather than short-term headlines.

   Pfizer's dividend sustainability (not fully covered by earnings or free cash flow), Medtronic's execution on AI initiatives amid diabetes spin-off and margin pressure, and Hinge Health's shift from devices to software-based care amid analyst guidance raises and index inclusions—each carries distinct risks that could affect growth and valuation going forward.

3. 3

   J&J hits $25.3bn Q2 sales, eyes $100bn milestone; Eli Lilly bets $3.8bn on psychedelics

   Johnson & Johnson reported Q2 2026 sales of US$25.3bn, up 6.6%, and raised its full-year guidance to US$101.1bn—putting it on track to exceed US$100bn in annual revenue for the first time. Separately, Eli Lilly agreed to acquire AtaiBeckley for up to US$3.8bn, marking its entry into psychedelic-based drug discovery for treatment-resistant depression. Bristol Myers Squibb is deploying a second NVIDIA-powered AI system that will deliver up to ten times greater performance per megawatt than its predecessor, positioning it as the life sciences industry's most powerful privately owned AI infrastructure. J&J's guidance upgrade underscores sustained momentum in its core pharmaceutical and medical device businesses at a scale few companies match. Eli Lilly's psychedelic bet signals confidence that novel mechanisms can address treatment-resistant depression, a condition that persists even after multiple standard therapies fail. BMS's AI infrastructure expansion directly supports drug discovery—a bottleneck in pharmaceutical R&D where computational power and efficiency are now central competitive levers.

   J&J's progress toward the US$100bn revenue threshold across 2026; whether AtaiBeckley's investigational psychedelic therapies advance through clinical development; and whether BMS's tenfold efficiency gain translates to faster candidate identification or reduced discovery timelines.

4. 4

   Intuitive Surgical unveils five-layer AI framework, demos live telesurgery

   Intuitive Surgical revealed a five-layer AI framework for its surgical platforms and demonstrated a live real-time telesurgical collaboration across distant facilities, showing how it plans to integrate AI into clinical workflows rather than as a standalone feature. The roadmap may influence how hospitals and surgeons adopt AI in surgery, shape partnerships, and affect investor views of Intuitive Surgical's long-term positioning in computer-assisted surgery—particularly as the stock has fallen 40.9% year to date and faces pressure.

   Hospital responses in terms of system placements, procedure volumes, and any disclosed AI-related service revenue in coming quarters; execution risks include regulatory comfort, data security, and surgeon adoption for AI-assisted and remote procedures.

5. 5

   Cigna deploys AI to scale care management program

   Cigna is using AI to expand its care management program, leveraging the technology to reach and support more patients across its member base. Care management programs help patients navigate complex health needs and reduce unnecessary medical costs. By using AI, Cigna can automate routine tasks and allocate human staff more efficiently to cases requiring personal attention, potentially improving outcomes while controlling expenses.

   The article does not specify a timeline, target enrollment, cost savings goal, or rollout regions, so concrete metrics on program scale and impact are not yet available.

6. 6

   Merck KGaA's AI Chief: Models Aren't the Advantage—Context and Trust Are

   Walid Mehanna, chief data and A.I. officer at Germany's Merck KGaA (the world's oldest pharmaceutical company, founded in the 17th century), is guiding the company's 62,000 employees through a three-layer A.I. strategy: everyday tools for productivity, embedded A.I. in core workflows like R&D and supply chains, and eventually product A.I. for drug discovery. The company deployed an internal platform called MyGPT in June 2023, then replaced it within a year by partnering with LangDock, a Berlin-based startup, to build a GDPR-compliant, model-agnostic system hosted in its own environment. Mehanna has shifted from believing the competitive edge lies in foundational models to holding that durable advantage comes from an organization's data, processes, workforce fluency, and customer trust—a view that challenges the assumption that the "best model" will win. For enterprises, this means the infrastructure and governance layer (not just the AI vendor) determines success; Merck KGaA's approach to privacy-protected A.I. (analyzing over 12 million internally-generated prompts at an aggregate level only, never monitoring individuals) shows how companies can move fast while meeting strict European regulation and labor relations requirements.

   Mehanna's Digital Ethics Advisory Panel, guided by five principles (autonomy, beneficence, non-maleficence, justice, and transparency), pushes teams to identify risks and design safeguards before launch rather than simply approving or rejecting projects. Teams now integrate data incrementally (process by process) rather than via a company-wide semantic layer, reflecting a matured view of how to scale A.I. across a large, regulated enterprise.