The November 2023 guidance from the FDA on computational modeling wasn't exactly beach reading. Dense, technical, focused on verification standards—the kind of document that lands with a bureaucratic thud. But buried in its pages was something industry insiders had been waiting for: tacit acknowledgment that virtual replicas of human physiology had graduated from academic curiosity to something regulators felt compelled to police.
By the time the European Medicines Agency finalized ICH M15 on Model-Informed Drug Development in February of this year, the message was unmistakable. Digital twins—the sector's preferred term for these computational proxies of patients, organs, and entire hospital systems—had acquired regulatory scaffolding. Whether the market was ready for them was another question entirely.
The numbers suggest someone thinks so. Grand View Research valued the global healthcare digital twins market at $902.6 million in 2024, projecting it will hit $3.55 billion by 2030—a 25.9% compound annual growth rate. Precedence Research, in a December 2025 forecast, sees $9.05 billion by 2034. Research and Markets pegs 2025 at $1.76 billion, this year at $2.02 billion.
The spread between these estimates is wide enough to drive a truck through, which is typical for nascent markets where vendor hype outpaces validated use cases. What matters isn't the precision of any single forecast. It's the consensus: computational models of patients and populations are migrating from PowerPoint decks to procurement budgets.
The Regulatory Machinery Wakes Up
The FDA's 2023 guidance aligned industry practice with ASME V&V 40, the verification and validation standard that engineers have been using for decades in aerospace and other high-stakes domains. In May 2025, the agency released a draft framework for assessing AI models in drug and biologic submissions, noting—almost in passing—that more than 500 submissions containing AI components had arrived since 2016.
That figure alone tells you regulators are playing catch-up.
Europe moved in parallel, though with its characteristic preference for committees and reflection papers. The EMA issued a qualification opinion in 2025 on an AI-based histology methodology for NASH and MASH. In November, it hosted a workshop on external controls—regulator-speak for synthetic comparator arms and digital twin-augmented evidence. These aren't fringe experiments. They're the EMA sketching the boundaries of what it will and won't accept.
Unlearn, a San Francisco-based company building what it calls "prognostic digital twins" for clinical trials, received a draft qualification opinion from the EMA back in May 2022 for its PROCOVA framework. This February, it announced a partnership with VectorY Therapeutics to deploy digital twins as exploratory external comparators in the PIONEER-ALS Phase 1/2 trial. Exploratory, mind you. Not primary endpoints. The industry is still walking carefully.
The Avicenna Alliance's 2024 book Toward Good Simulation Practice and a National Academies report the same year offer something resembling a community-driven to-do list: credibility standards, uncertainty quantification, multiscale modeling, interoperability. In other words, the infrastructure builders are trying to standardize before regulation forces their hand. Whether they'll succeed is less clear.
Data Plumbing Gets an Overhaul
Regulatory clarity is necessary but not sufficient. You can't build a patient twin without data, and healthcare data—fragmented across EHRs, imaging systems, wearables, and lab databases—has historically been a mess.
That's changing, albeit unevenly. The CMS Interoperability and Prior Authorization Final Rule, finalized January 17, 2024, mandates that impacted payers implement FHIR-based APIs by January 1, 2027. ONC's HTI-1 rule, effective March 11, 2024, requires algorithm transparency for decision support tools and pushes adoption of USCDI v3 and US Core by this year. TEFCA, which designated its first Qualified Health Information Networks (QHINs) in December 2023, planned FHIR exchange pilots for 2025.
These aren't headline-grabbing moves. They're plumbing. But plumbing matters when you're trying to build computational models that ingest real-time physiological data, imaging streams, and longitudinal health records.
AWS HealthLake added FHIR features and launched in the EU in June 2025. Microsoft has been communicating EHR-scale capabilities on Azure throughout the past two years, citing an IDC survey from March 2024 showing 79% of organizations using AI in healthcare. NVIDIA's Clara, Holoscan, and IGX platforms underpin edge compute for streaming data and real-time simulation in clinical settings.
The stack is consolidating, which is what happens when an emerging market starts to look like an actual market.
The European Commission's Virtual Human Twins Initiative, launched in 2023, moved forward in 2025 with procurement for an integration and validation platform—a roughly €100 million investment signaling that Europe intends to build shared repositories and audit trails, not just fund isolated pilots. That's infrastructure thinking, and it suggests public money will shape how models get built, validated, and shared across borders.
Hospital Ops: The Fast Movers
Of all the digital twin use cases, hospital operations have moved fastest—perhaps because the stakes are lower and the value proposition is immediate. Simulate bed capacity, emergency department flow, perioperative schedules under different demand scenarios, and you can optimize staffing and throughput without touching a patient.
GE HealthCare's Digital Twin for hospital operations, detailed in research articles from October 2025 and February 2025, models bed capacity, ED boarding times, and perioperative flow. Siemens Healthineers introduced its ActExcell Operational Twin at RSNA 2025. These aren't departmental skunkworks projects. They're enterprise systems designed to run "what-if" scenarios across entire facilities.
The American Hospital Association's 2026 Environmental Scan flagged digital twins for healthcare systems as a key trend hospital leaders are tracking—which is consultant-speak for "we're spending money on this."
Physiological twins are taking a narrower path, focusing on device planning and therapy optimization in cardiology. Dassault Systèmes announced the next phase of its Living Heart Project in February 2025, emphasizing AI-powered virtual twins for device R&D and regulatory testing, built on a five-year collaboration with the FDA. FEops, acquired by Materialise in July 2024, received FDA 510(k) clearance in November 2025 for HEARTguide, a patient-specific simulation app for left atrial appendage occlusion planning. HeartFlow announced FDA clearance for its next-generation plaque analysis in September 2025.
These aren't visions. They're cleared software applications generating predictions about individual patients' anatomies and hemodynamics, integrated into clinical workflows. The regulatory machinery exists to evaluate them, and the reimbursement architecture is beginning to accommodate the underlying data streams—Medicare's remote patient monitoring codes, clarified in the 2024 Physician Fee Schedule, incentivize continuous physiologic data collection that feeds personalized models.
The gap between operational twins (widely deployed) and physiological twins (narrowly deployed) is a function of risk, not technology. Simulate a hospital incorrectly and you waste money. Simulate a heart incorrectly and you kill someone.
The Trial Augmentation Gambit
Clinical trials represent perhaps the most economically consequential use case for digital twins, and here the infrastructure remains uneven—though progress is accelerating.
Unlearn's February announcement with VectorY marks a deployment of AI-generated digital twins as exploratory external comparators in ALS, a disease notorious for small patient populations and high placebo response variability. The company's PROCOVA-MMRM methodology, detailed in an April 2024 preprint, models individual patient trajectories and generates synthetic control arms. If it works, it could shrink trial sizes, reduce placebo exposure, and accelerate timelines.
If it works. Peer-reviewed validation at scale remains sparse.
Aitia announced a partnership with Orion in September 2024 to create what it calls Gemini Digital Twins in oncology, framing its approach as causal AI integrating multi-omic data to de-risk discovery and simulate trial designs. VeriSIM Life continues advancing its BIOiSIM platform for translational biosimulation. These are vendor claims, and the industry has learned—sometimes painfully—to wait for independent validation before declaring victory.
The EMA's 2025 workshop on external controls and ongoing work on a reflection paper suggest regulators are moving toward formal guidance rather than reactive review. The FDA's May 2025 draft framework for AI models in drug submissions emphasizes risk-based credibility—acknowledging that different contexts demand different validation thresholds. That flexibility, if codified, could accelerate adoption. Or it could open the door to poorly validated models dressed up as evidence.
Regulators are threading a needle: enable innovation without enabling garbage.
Vendor Claims Meet Market Skepticism
Twin Health, a consumer-clinical metabolic health platform, raised $53 million in a Series E round at a $950 million valuation in August 2025. The company's Whole-Body Digital Twin integrates more than 3,000 daily data points—continuous glucose monitoring, activity trackers, nutrition logs—to generate personalized intervention recommendations.
Conference posters at EASD 2024 referenced program outcomes. Large-scale, peer-reviewed randomized controlled trials? Not yet widely published.
That pattern repeats. FEops cites improved efficiency and outcomes for its structural heart planning tools, but those claims come from vendor-sponsored studies or proprietary case briefs. PrediSurge highlights AI-powered modules for EVAR and TAVI planning, noting a Medtronic collaboration and iSizing validation in January of this year. External corroboration is pending.
The infrastructure players—GE HealthCare, Siemens Healthineers, Dassault Systèmes—have regulatory clearances and installed bases, which lends credibility. But the clinical evidence for operational twins remains largely proprietary performance reports rather than published comparative effectiveness studies. The Digital Health Most Wired National Trends Report 2025, published in January, indicates AR/VR and digital twin modeling are moving beyond pilots for training, patient education, and operations. Adoption percentages weren't disclosed.
Translation: interest is high, but nobody's ready to share hard numbers.
What Actually Works (and What Doesn't)
For companies evaluating digital twin platforms, the infrastructure checklist is becoming specific and non-negotiable.
FHIR APIs are mandatory by 2027 under CMS rules. Any twin relying on EHR data must architect for US Core and USCDI v3. OMOP Common Data Model extensions, including imaging OMOP, provide analytic backbones for real-world evidence generation. AWS HealthLake and Microsoft Cloud for Healthcare offer FHIR-native cloud architectures. NVIDIA Clara provides edge inference for real-time imaging and streaming data.
Regulatory credibility hinges on ASME V&V 40 compliance for device applications and alignment with the FDA's emerging AI credibility framework for drug and biologic submissions. The Avicenna Alliance's Good Simulation Practice guidelines and MDIC/FDA symposium proceedings published in February offer implementation roadmaps. The Digital Twin Consortium launched testbeds in May 2025, including a GenAI track for healthcare twins, though these remain programmatic and early-stage.
The gaps, though, remain significant—sometimes alarmingly so.
Model credibility and uncertainty quantification for clinical decision support at the patient level are active research areas, not solved problems, according to the National Academies 2024 report. Data interoperability across EHRs, imaging, wearables, and omics remains uneven despite FHIR and OMOP progress, per a 2025 Frontiers systematic review. Regulatory clarity for AI-updated models in the real world is evolving; the FDA's 2025 request for public comment on real-world evaluation of AI-enabled devices underscores ongoing uncertainty about post-market performance and update governance.
In other words, the industry has figured out how to simulate hospital beds and model coronary arteries. It hasn't figured out how to validate continuously learning models that update themselves in production, or how to govern those updates when they're influencing treatment decisions.
Where the Money Meets the Math
Europe's Virtual Human Twins Initiative and the €100 million procurement for an integration platform suggest public investment will shape model repositories and validation infrastructure in ways that favor standardized data and model interfaces. Infrastructure providers offering audit trails, credibility documentation, and semantic alignment across data sources are positioning for that procurement environment, betting that Europe's regulatory culture will reward transparency and standardization over speed.
The U.S. market, characteristically, is betting on faster iteration and venture-backed scale. Twin Health's $950 million valuation reflects investor appetite, but also investor willingness to overlook the absence of peer-reviewed outcomes data—at least for now.
The industry is past the proof-of-concept phase in operational twins and select device planning applications. That much is clear. It is entering what might be called the credibility-at-scale phase for clinical trials and population health, where regulatory frameworks exist in draft, data infrastructure is mandated but not yet ubiquitous, and the gap between vendor claims and independent validation remains uncomfortably wide.
The $9 billion market forecast by 2034 assumes those drafts finalize, those mandates enforce, and that gap narrows. The timeline is compressing—regulators are moving, infrastructure is consolidating, reimbursement is adapting.
But it hasn't collapsed. Not yet.
