Veeva Systems: Infrastructure at the Edge of Pharma's AI Transformation

Drug development is one of the last major industries that AI has not visibly restructured. The reasons are structural, not accidental. Pharmaceutical development operates under a regulatory framework that treats every software system as a validated instrument, where changes require documented evidence of equivalence, and where the consequences of a documentation failure are measured in clinical holds, FDA warning letters, and consent decrees rather than customer support tickets. AI cannot be dropped into a drug development workflow the way it can be integrated into legal research or financial modelling. It must be validated, audited, and defensible to a regulator before it touches any data that informs a submission.

That friction has protected pharmaceutical software infrastructure from the disruption wave that has already reshaped other sectors. But friction is not immunity, and the signals that the transformation is beginning to arrive are accumulating.

The cost of bringing a drug to market currently averages approximately $2.6 billion and takes 12 to 14 years. Most of that cost is not molecular science. It is process: clinical trial execution, site monitoring, safety signal processing, regulatory document preparation, quality system maintenance. These are labour-intensive, document-heavy, rule-governed workflows that large language models and AI agents are specifically suited to compress. A safety narrative that takes a medical writer two days to draft could, with validated AI tooling, be generated in minutes from a structured safety database. A Trial Master File quality check that requires teams of monitors reviewing thousands of documents could be partially automated. A regulatory submission that requires months of cross-referencing across disparate systems could be assembled in weeks.

The question is not whether this transformation will happen. The FDA has been increasingly explicit in its guidance that it will accept AI-assisted analysis provided the underlying data is validated, the methodology is documented, and human oversight is maintained. The question for an investor is who captures the value when the transformation arrives, who loses it, and whether the companies that currently provide the infrastructure are positioned as beneficiaries or as casualties.

That distinction is not obvious, and anyone who claims it is has probably not examined both sides honestly.

The Picks and Shovels Logic and Its Limits

The investment case for pharmaceutical infrastructure companies rests on what is often called the picks and shovels argument. In any transformational wave, the companies that supply the enabling infrastructure to everyone participating tend to generate more reliable returns than the companies making directional bets on specific outcomes. In the 1849 California gold rush, the merchants who sold shovels and boots to miners made money regardless of which individual miner struck gold. The infrastructure provider's return depends on the activity continuing, not on any single bet paying off.

In pharmaceutical AI, the equivalent of the miner is the drug developer: a biotech or pharmaceutical company betting on a specific molecule in a specific indication navigating a specific regulatory pathway. That bet is binary by nature. The infrastructure layer, the companies providing validated data systems, workflow platforms, and auditable AI applications that every drug developer must use regardless of outcome, theoretically profits from the activity itself rather than from any single success.

This logic is genuinely compelling in the pharmaceutical context, more so than in most sectors, because the regulatory requirement for validated systems is non-negotiable. Every company developing drugs must use validated software. That creates a stable demand floor that does not exist in less regulated industries.

The limit of the picks and shovels argument, however, is that it assumes the infrastructure layer itself is not disrupted. The shovel merchant's business was safe because nothing replaced the shovel. If AI replaces significant portions of the pharmaceutical workflow that current infrastructure platforms were built to manage, the infrastructure itself faces disruption. The picks and shovels metaphor breaks down precisely at the moment it is most needed: when the nature of the work being done changes fundamentally. That is the tension at the centre of evaluating any pharmaceutical software company in 2026, and it applies directly to Veeva.

What Veeva Is

Veeva Systems is the dominant cloud software platform for the global pharmaceutical and biotech industry. It is not a drug company, not a contract research organisation, and not an AI drug discovery platform. It is the system of record for the most critical workflows in pharmaceutical development and commercialisation: the software in which clinical trial documentation lives, regulatory submissions are prepared, safety signals are processed, quality processes are tracked, and commercial sales forces manage their interactions with physicians.

The company operates across two major segments. Development Cloud, covering clinical, quality, safety, and regulatory applications on its Vault platform, now accounts for more than 55% of total revenue and is growing faster than Commercial. Commercial Cloud manages the interaction between pharmaceutical sales representatives and the healthcare systems they call on.

The regulatory context is what makes Veeva's business structurally different from most enterprise software. Every system touching clinical trial data, safety reports, or regulatory submission documents must be validated under 21 CFR Part 11, the FDA regulation governing electronic records and electronic signatures. Validation is not a software certification. It is an ongoing programme of testing, documentation, and change control that must be maintained for the life of the system. When a pharmaceutical company switches validated systems, it must run both in parallel during migration, maintain audit trail continuity across both, revalidate the new system, and document equivalence of every historical record. For a system holding complete documentation of a drug programme spanning a decade, this is not a migration project. It is a multi-year regulatory risk event.

The observable consequence: net revenue retention consistently above 100%, and no documented major customer defection from the Vault development platform in the company's history. As of fiscal 2026, Veeva serves 1,552 customers. This switching cost is real and has been demonstrated empirically, not just argued theoretically.

What Veeva holds as a result of this installed base is the most consequential fact for its AI positioning: decades of structured, validated, interconnected clinical trial documentation, regulatory submission records, safety reports, and quality data across essentially every major pharmaceutical company in the world.

The AI Question: Opportunity and Threat Are the Same Coin

This is where genuine analytical difficulty begins, because the AI story for Veeva is not simply bullish or simply bearish. It is the same set of facts read from two different angles, and which reading proves correct will determine whether the current price is an opportunity or a trap.

The opportunity reading: Veeva's data is the raw material for pharmaceutical AI in a way that no competitor can easily replicate. The corpus it holds is not just large. It is structured, validated, and relationally connected across the development lifecycle. A safety narrative in Vault is linked to the adverse event that generated it, which is linked to the clinical study in which it occurred, which is linked to the regulatory submission in which it was reported. That relational structure is what makes the data useful for AI rather than merely voluminous. No company building pharmaceutical AI from external data sources has this kind of connected, validated foundation.

Veeva has begun deploying AI agents on this data. Safety narrative drafting agents, Trial Master File quality review agents, regulatory submission formatting agents, and commercial content generation tools are in early deployment. The most significant near-term application is safety processing. Pharmacovigilance, the ongoing monitoring of drug safety signals across a product's commercial life, is one of the most labour-intensive and outsourced functions in the industry. Individual case safety reports must be processed, narratives drafted, and regulatory submissions filed to dozens of health authorities on specific timelines. The industry outsources much of this to contract research organisations at significant cost. If Vault Safety AI agents can process a meaningful proportion of this work at acceptable quality standards, the financial implication is not just incremental AI software revenue. It is that pharmaceutical companies may choose to internalise safety processing they currently outsource, expanding Vault Safety's addressable market in the process.

A pharmacist reading that argument can evaluate it in a way a generalist investor cannot. Whether AI-generated safety narratives meet the quality standards that pharmacovigilance professionals and regulators actually require is a scientific and regulatory question. It is not determined by Veeva's product roadmap. It is determined by whether the FDA and other health authorities accept AI-assisted case processing as the basis for compliant adverse event reporting. That acceptance is not guaranteed, and the timeline is not set by Veeva.

The threat reading starts from the same facts. Veeva's seat-based subscription model is priced on the assumption that pharmaceutical companies employ large teams of people managing these workflows. Safety teams, regulatory affairs groups, clinical operations functions, and medical writing departments all generate Veeva subscription revenue because they use the platform in their daily work. If AI agents reduce the headcount required to manage these workflows, the number of people using Veeva daily shrinks, and with it the natural basis for subscription expansion. A safety team currently requiring 50 people using Vault Safety could, with effective AI tooling, require 20. The revenue per user metric is irrelevant if the number of users is falling faster than the per-user price is rising.

This is not speculation. It is the structural pressure that has already compressed valuations across legal workflow software, accounting platforms, and knowledge work infrastructure more broadly. The question is whether Veeva's AI monetisation layer, incremental fees for agent access and new use cases AI opens up, outpaces the compression of seat-based demand. Management has not quantified either side of this equation in reported results. Vault AI revenue is not a disclosed line item. Customer adoption metrics for AI agents have not been reported with enough specificity to model the trajectory.

Goldman Sachs, which maintains a Sell rating on Veeva, has framed this concern explicitly: the more challenging growth outlook as core products mature, tighter CRM competition, and questions about Veeva's role in the broader AI stack rather than as a pure beneficiary of it. That concern is not irrational, and the fact that most sell-side analysts remain constructive does not make Goldman wrong.

The Commercial Cloud: A Genuine Battle With an Uncertain Outcome

The commercial segment is where the competitive pressure is most visible and where the AI disruption risk is most immediate, operating simultaneously through two distinct channels.

The first channel is platform competition. Salesforce, Veeva's original infrastructure partner and now its primary commercial rival, has aggressively pursued pharmaceutical customers with its Agentforce AI platform and has accumulated approximately 40 wins among pharmaceutical companies choosing it over Vault CRM. Veeva, meanwhile, counts 9 of the top 20 largest pharmaceutical companies committed to Vault CRM versus Salesforce's 3, with Veeva migrations already underway while Salesforce timelines extend into 2027 through 2029. The competitive picture is mixed rather than decisive in either direction.

The migration complexity compounds the competitive uncertainty. Veeva will not migrate customers' custom code, custom objects, or third-party integrations. Customers must handle these themselves, concentrating the execution risk on the customer's own IT resources. Smaller companies that have not begun planning face potential delays as enterprise capacity for large transformation programmes is finite. A backlog of competing migrations could create the kind of decision paralysis that benefits neither Veeva nor Salesforce.

The second channel is more structural and less discussed. If AI agents replace a meaningful portion of the pharmaceutical sales representative's function, the total addressable market for pharmaceutical commercial software shrinks regardless of which platform wins. The business case for large field forces has already been challenged in some therapeutic areas where prescribing patterns are consolidating. The commercial software market may be competing over a declining prize, and the platform competition between Veeva and Salesforce may be less important than the question of how large the commercial software market is in five years.

These two channels operate independently. Veeva could win the platform competition and still face headwinds from market size compression, or lose the platform competition in a market that turns out to be larger than expected.

The Competitive Landscape Without a Clear Winner

Veeva's competitive position differs significantly by segment, and conflating them produces a misleading overall picture.

In Development Cloud, the primary competitor is Medidata, owned by Dassault Systèmes, which holds a significant installed base in electronic data capture and clinical data management. Medidata has its own AI analytics platform in Acorn AI and its own clinical trial data assets accumulated across decades of deployments. The data moat argument for Veeva in clinical AI is therefore contested rather than exclusive. Medidata's data may be narrower in scope than Vault's cross-functional coverage, but it is deeper in clinical execution specifics that matter for AI-assisted trial design and site performance prediction. IQVIA, whose legal dispute with Veeva resolved during fiscal 2026, holds a real-world evidence and data analytics business that competes with Veeva Data Cloud ambitions in the commercial and safety analytics space. Oracle Health Sciences has largely ceded the modern cloud narrative but retains significant legacy installations that represent displacement opportunities for Veeva without guaranteeing Veeva captures them.

In Commercial Cloud, the Salesforce competition is real and the outcome is genuinely uncertain. Whether Vault CRM's pharma-specific compliance architecture outweighs Agentforce's general AI capabilities in the decision-making of pharmaceutical commercial leadership is not a question that has been settled, and it will not be settled by a single earnings call.

What the competitive landscape does not contain is a single overwhelming threat that would resolve the question quickly in either direction. The Development Cloud competition is slow-moving given regulatory switching costs. The Commercial Cloud competition is fast-moving but contested. Neither scenario produces a clear near-term answer.

The Financial Picture and What It Actually Shows

Fiscal year 2026 revenue reached $3.195 billion, up 16% year over year. Subscription revenues of $2.684 billion grew 17%. Non-GAAP operating margins reached 44%. Free cash flow of $1.39 billion represents a 43.5% cash flow margin. The balance sheet holds approximately $6 billion in net cash with no meaningful debt. A $2 billion share buyback programme is underway.

Fiscal 2027 guidance projects revenues of $3.585 to $3.600 billion, implying approximately 12.5% growth at the midpoint, a deceleration from the 16% delivered in fiscal 2026. Non-GAAP EPS growth is expected at approximately 8.8%. Management has guided to a $6 billion revenue run rate by 2030, implying a compound annual growth rate of approximately 13% from fiscal 2026 levels.

At $160 per share, the market capitalisation is approximately $26 billion. With $6 billion in net cash, the enterprise value is approximately $20 billion. Against fiscal 2027 revenue guidance of approximately $3.6 billion, the enterprise value to revenue multiple is approximately 5.5 times. On forward non-GAAP earnings near $8.80 per share, the forward earnings multiple is approximately 20 times. These multiples are near multi-year lows for Veeva, and well below the 40 to 50 times forward earnings the stock commanded at peak in 2021.

The stock has declined approximately 30% from the highs reached following the March 2026 earnings beat, despite the underlying business delivering results that exceeded guidance across every reported metric. The bear reading of this divergence is that the market is correctly identifying a structural deterioration in the growth outlook that the reported numbers do not yet reflect. The bull reading is that the market is mispricing a durable business by applying software-sector AI disruption fears indiscriminately to a platform protected by regulatory switching costs.

Both readings are internally consistent with the available data. The financial results do not yet resolve which one is correct.

Analyst opinion is divided in a way that is itself informative. Goldman Sachs maintains a Sell, pointing to maturing product cycles, CRM competition, and uncertainty about Veeva's role in the AI stack. Most other major banks, including TD Cowen, Piper Sandler, Stifel, Canaccord, RBC, and Barclays, maintain positive or overweight views citing execution quality and the development cloud moat. The analyst consensus leans bullish, but the dissenting voice is Goldman Sachs, which is not an institution that makes casual Sell calls on $26 billion market cap software companies.

What Would Make You Buy, What Would Make You Avoid

The case for buying rests on three conditions holding simultaneously. Development Cloud must continue growing at or above 15% through fiscal 2027 and beyond, demonstrating that the regulatory switching cost moat is not eroding under AI pressure. Vault AI agents must begin appearing as disclosed revenue or quantified adoption metrics by the end of fiscal 2027, demonstrating that Veeva is a beneficiary of AI rather than a platform whose headcount-based demand is quietly compressing. And the Vault CRM win rate among top-tier pharmaceutical companies must remain above 50%, demonstrating that the commercial cloud is navigating a transition rather than undergoing structural share loss.

If all three conditions hold, the current price at 5.5 times forward enterprise value to revenue, for a business generating 44% non-GAAP margins with high-retention subscription revenue and $6 billion in net cash, is likely to look attractive in retrospect. The Development Cloud alone, if valued at 7 times revenue on $2 billion in annual revenue and growing, implies an enterprise value near $14 billion for that segment. Against a total enterprise value of $20 billion for the entire business including Commercial Cloud, the market is attributing relatively little to commercial, data cloud ambitions, or AI optionality.

The case for avoiding rests on a different set of conditions. If AI adoption in pharmaceutical workflows compresses the headcount that generates subscription demand faster than new monetisation replaces it, the reported revenue growth will begin decelerating in ways that the current multiple does not account for. If Salesforce wins the commercial cloud battle among top-tier accounts, the 45% of revenue in that segment faces structural erosion. And if Vault AI agents fail to reach disclosed adoption within fiscal 2027, the patience required to hold through that uncertainty may not be rewarded within a reasonable investment horizon.

The honest position is that these conditions cannot be resolved from available public data. Vault AI revenue is undisclosed. Commercial win rates among the top 20 are partially visible but not complete. Development Cloud AI adoption is described qualitatively in earnings calls but not quantified. The investor is being asked to form a view about outcomes that the evidence has not yet determined.

Conclusion

Veeva at $160 is neither obviously cheap nor obviously expensive. The business generates $3.2 billion in revenue growing at 16% with margins above 44%, holds $6 billion in net cash, and retains essentially every customer in its most regulated segment. That is a genuinely high-quality business at a historically low multiple. The multiple is low for a reason: the market is uncertain whether Veeva is a beneficiary or a casualty of the pharmaceutical AI transformation, and the financial evidence to resolve that question does not yet exist.

The Development Cloud moat is real and empirically demonstrated. The Commercial Cloud is under genuine competitive pressure from Salesforce and from the structural possibility that pharmaceutical field force models change materially. The AI opportunity is real in principle and unquantified in practice. The AI threat is real in principle and unquantified in practice.

An investor who buys Veeva today is making a bet that the Development Cloud moat holds through the AI transition, that Vault AI agents monetize faster than seat-based demand compresses, and that the commercial competition resolves without destroying the segment's revenue contribution. None of those outcomes is improbable, but none is assured.

An investor who avoids Veeva today is accepting Goldman's read: that the growth outlook is structurally more challenging than the current consensus implies, that the role Veeva plays in the emerging pharmaceutical AI stack is not yet clear, and that the current multiple, though historically low, does not represent a sufficient margin of safety for the uncertainty involved.

The investment case is not resolved by the available evidence. It is resolved by a view about pharmaceutical AI adoption that requires exactly the kind of scientific and regulatory judgment that this analysis alone cannot supply.