Deloitte’s 2025 Health Tech Industry Accounting Guide, now in its sixth edition, is the most comprehensive map available of how this complexity is being navigated by professional firms. Drawing on ASC 350-40, ASC 985-20, ASC 606, and ASC 340-40, the guide establishes a framework that anyone running, advising, or investing in a health tech company should understand at a working level. This article distills the most consequential parts of that framework — and explains why they matter for the operational and commercial realities of the sector.

For most health tech companies, the single most important accounting decision is which of two standards governs their software development costs: ASC 350-40 for internal-use software, or ASC 985-20 for software to be sold or marketed externally. The choice is not optional. It is dictated by whether there is a substantive plan to market the software externally, and the threshold for what counts as substantive is high — selection of marketing channels, identified promotional and billing infrastructure, support activities, and a plan that is at least reasonably possible to implement.

The practical impact of this choice is enormous. Under ASC 350-40, capitalization begins when the preliminary project stage is complete and the application development stage starts — typically much earlier in the development cycle. Under ASC 985-20, capitalization cannot begin until technological feasibility is established, which requires either a completed detail program design or a working model. For most software products, this happens far later in the timeline, meaning a much smaller pool of costs ends up on the balance sheet.

The accounting framework for software costs was largely written more than 20 years ago for a world of waterfall development — long, structured project plans with clearly delineated stages. Modern health tech companies don’t work that way. Sprints last two to three weeks, requirements evolve continuously, and the same team may move through preliminary planning, application development, and post-implementation activities within the span of a single Monday-to-Friday cycle.

No part of health tech accounting is moving faster than the treatment of generative AI development costs. The Deloitte guide dedicates substantial coverage to how foundation models, fine-tuning, prompt engineering, adversarial training, and data acquisition costs should be treated under existing standards — and the analysis reveals just how poorly suited the existing framework is to the realities of building AI applications.

For health tech companies developing AI-powered diagnostic, clinical decision support, or workflow automation tools, the most consequential issues are data acquisition costs, ongoing fine-tuning expenditures, and compute infrastructure. The general rule: data acquired from third parties for use in software with alternative future uses should be recognized as a separate intangible asset under ASC 350-30. Data acquired for a specific software project with no alternative future use, however, can be capitalized as part of the internal-use software asset under ASC 350-40 — but only if incurred during the application development stage. If the data is used to develop technological feasibility itself, it falls under ASC 730-10 and must be expensed as R&D.

ASC 606 introduced a five-step revenue recognition model: identify the contract, identify performance obligations, determine the transaction price, allocate the transaction price, and recognize revenue when or as obligations are satisfied. For health tech companies bundling smart devices, embedded firmware, cloud services, professional services, and post-contract support into single arrangements, every step requires significant judgment.

The hardest question is usually performance obligation identification. A health tech company selling a wearable patient monitor with embedded firmware, telemetry to a cloud platform, and physician-facing analytics has to decide whether these are one obligation or several. The answer turns on whether the device and the service are highly interdependent or highly interrelated — whether the customer’s intended benefit can be obtained from one without the other. If the cloud service is essential to the device’s functionality (transformative rather than additive), the entire bundle may be a single performance obligation recognized over the service period. If the device works on its own and the cloud service is optional or replaceable, multiple obligations exist with revenue recognized at different points in time.

As health systems transition from on-premise software to SaaS deployments, an entire category of contract modifications has emerged that the original revenue standard didn’t directly anticipate. A hospital purchases a five-year on-premise term license for a clinical workflow platform; two years in, the vendor wants to migrate the customer to a hosted version. The accounting answer is genuinely contested. Deloitte presents two views — the material right model (preferred) and the right of return model (acceptable) — and the timing of revenue recognition can differ materially depending on which is selected.

For health tech CFOs, the practical implication is that two companies with economically identical contracts can report substantially different revenue patterns depending on the policy they elect. Investors and acquirers reading financial statements should pay close attention to disclosed revenue recognition policies for cloud transitions, because the comparability across companies is genuinely limited.

Mental health and wellness has become one of the largest single subsectors within health tech, with investment patterns that illustrate the broader dynamics of the industry. Pre-pandemic activity hovered below $1 billion annually. The 2020 lockdowns triggered a tripling of capital deployment to over $3 billion, peaking at $3.8 billion in 2021. The market correction of 2022 reduced overall financing but produced a record 105 deals — investors writing smaller checks across more companies. Activity declined further in 2023 before recovering in 2024.

Seven MHW companies achieved unicorn status in 2021. None did in 2022. Three more reached the threshold between 2023 and 2024. The clinics and outpatient services subsector — capital-intensive brick-and-mortar with significant compliance overhead — has consistently led deal volume, with more than 20 expansion-stage deals annually from 2020 through 2024. Substance use disorder treatment alone accounted for $13 billion in emergency department costs in 2017 and $35.3 billion in employer-sponsored health insurance payouts in 2018, illustrating both the scale of unmet need and the commercial opportunity.

ASC 340-40 requires capitalization of incremental costs of obtaining a contract — costs that would not have been incurred if the contract had not been obtained. The canonical example is sales commissions, but the application is more nuanced than it appears. Fixed employee salaries don’t qualify even if they’re partially based on sales projections. Legal and travel costs incurred during contract negotiation don’t qualify because they would have been incurred even if the contract fell through. But commissions paid to multiple employees on a single deal — the salesperson, the manager, the regional manager — can all qualify as incremental, as long as each commission is tied directly to contract execution rather than to broader performance metrics.

The amortization period also matters. For a SaaS company that pays commensurate commissions on renewals, the amortization period is the original contract term. For one that doesn’t pay renewal commissions — meaning the initial commission effectively bought a multi-year customer relationship — the amortization period is the estimated customer life, often five years or more. The difference can shift millions of dollars between operating expense and balance sheet.

For U.S.-based health tech companies, two tax provisions deserve specific attention. The R&D credit can offset federal income tax for qualified research activities, including software development that resolves technological uncertainty. Internal-use software faces a higher innovation threshold but can still qualify if it doesn’t simply automate back-office functions. The Section 250 deduction for foreign-derived intangible income (FDII) provides a permanent reduction in effective tax rate for income from sales or services to foreign customers — particularly relevant for health tech companies with international hospital system customers or telehealth offerings serving cross-border patients.

Both provisions reward proactive structuring. Companies that wait until tax filing season to think about them typically miss the documentation requirements needed to substantiate the deductions. The compounding effect of getting these wrong over multiple years can run into the millions for a mid-stage health tech company.

For hospital IT leaders evaluating vendor proposals, the accounting framework reveals important purchasing dynamics. Vendors with significant capitalized software balances are typically further along in their development lifecycle and more committed to ongoing investment in their products. Vendors offering aggressive cloud conversion incentives may be transitioning their revenue models in ways that affect long-term pricing power. Multi-element arrangements that bundle hardware, software, and services should be scrutinized for hidden cost allocations — particularly when the total contract value seems disproportionate to comparable point solutions.

The accounting also signals where vendor incentives may not align with hospital priorities. A vendor whose revenue is recognized over the SaaS service period has economic incentive to maintain product quality and customer satisfaction. A vendor whose revenue was recognized at point-in-time license delivery has weaker incentives to invest in ongoing improvements unless renewal economics are structured carefully.

They are structural. Fragmented data foundations, regulatory uncertainty, workforce capacity gaps, and governance blind spots are holding back what could be the most consequential shift in healthcare delivery since the adoption of electronic health records. The OECD’s response — a policy checklist organised into four pillars covering enablers, guardrails, engagement, and trustworthy deployment — is the first serious attempt at creating a cross-border framework for responsible AI scaling in health.

For healthcare technology companies, medical device manufacturers, and digital health startups, this report is more than a policy document. It is a roadmap for where procurement budgets, regulatory requirements, and institutional priorities are heading over the next three to five years.

The OECD finding that every member country uses AI in administration but almost none have scaled it clinically reveals the real bottleneck. Administrative AI — scheduling, billing, claims processing, resource allocation — operates on structured data within well-defined workflows. Clinical AI — diagnostic imaging, pathology analysis, treatment recommendation, drug interaction prediction — operates on messy, unstructured, highly sensitive patient data within workflows where errors have direct human consequences.

The infrastructure requirements are fundamentally different. Administrative AI can run on standard cloud compute. Clinical AI at national scale requires GPU-accelerated inference, real-time processing of imaging data, federated learning architectures that keep patient data within institutional boundaries, and validation pipelines that meet regulatory standards in every jurisdiction where the system operates. The compute costs alone are significant — and growing.

The OECD checklist puts data foundations first for a reason. AI models are only as good as the data they’re trained on, and healthcare data is notoriously fragmented. Patient records sit in different formats across different systems in different institutions, often within the same city. Imaging data from one hospital may use different DICOM standards than the hospital across the street. Lab results, clinical notes, genomic data, and wearable sensor data all live in separate silos with different access controls and different levels of quality.

The emerging solution — country-led health data authorities that ensure data is findable, accessible, interoperable, and reusable (the FAIR principles) — is the right idea but will take years to implement. In the meantime, companies building AI for healthcare have to work with what exists: incomplete datasets, inconsistent formatting, and access processes that vary by institution, region, and country.

One aspect the OECD report touches on but doesn’t fully explore is the compute infrastructure required to scale AI in health. Training a diagnostic imaging model on millions of scans requires substantial GPU resources. Running that model in production across a national health system — processing thousands of scans per day with sub-second latency — requires even more. And when you add the emerging category of large language model applications in healthcare — clinical note summarisation, patient communication, literature analysis, clinical trial matching — the API costs start to compound rapidly.

Health systems and digital health companies are increasingly using AI APIs from providers like Anthropic, OpenAI, and the cloud-hosted AI services from Azure and Google Cloud. Many entered these platforms through startup grants, research partnerships, or promotional credit programmes — and now find themselves with credits allocated to one provider while their actual usage has shifted to another. If your organisation has unused Anthropic capacity from a research grant or pilot programme that didn’t scale, you can sell Anthropic credits through brokers who match sellers with buyers looking for discounted AI API access. It’s a practical way to recover value from credits that would otherwise expire — capital that can be redirected toward the AI workloads you’re actually running.

The OECD checklist identifies workforce capacity as a critical enabler — and the data supports it. Only 29% of member countries have established a national approach to improving AI use in the health workforce. The gap isn’t just about training clinicians to use AI tools. It’s about creating entirely new roles: clinical AI validators who can assess model outputs against medical evidence, health data engineers who can build compliant data pipelines, and AI ethics officers embedded within health institutions rather than technology companies.

The countries that move fastest on workforce development will have a structural advantage in AI adoption. Korea is already mandating AI education within health professional curricula. The UK’s Digital and Data Professional Capability Framework is proactively mapping the skills needed across both clinical and technical roles. These aren’t future plans — they’re being implemented now.