How Can IP Analytics Unlock True Value in AI-Driven Deep Tech Transactions?

If intangible assets account for over 60% of global M&A value, why do most AI deals still treat intellectual property (IP) as if it were static software? 

WIPO and OECD data show that the intensity of intangibles in AI, semiconductors, and advanced software consistently surpasses that in traditional technology sectors. Yet, transaction valuation practices often reduce intellectual property to mere ownership confirmation rather than considering its economic behaviour. In AI-focused deals, this gap is increasingly significant, impacting price, risk allocation, and post-deal value realisation. 

Consequently, IP analytics has evolved into a strategic valuation discipline that intersects legal enforceability, technical architecture, and financial durability. For skilled dealmakers, the crucial question is not whether IP exists, but whether it can maintain its advantage in the face of scale, regulation, and competitive replication. 

Reframing IP from inventory to system 

The most significant shift in IP analytics is conceptual in nature. High-value AI companies do not possess discrete IP assets; they operate interconnected IP systems where patents, trade secrets, models, and data rights reinforce one another. 

Patent analytics has shifted its emphasis from filing volume to structural relevance. Citation-weighted influence, claim dependency graphs, and technology adjacency mapping are now standard in advanced diligence. EPO research indicates that patents located at technology convergence points generate significantly higher licensing and defensive value. In AI, these convergence points increasingly lie between compute, model optimisation, and application-layer integration. 

Trade secrets introduce an additional layer of complexity. Unlike patents, their value relies on organisational execution rather than registries. Advanced analytics evaluate whether proprietary knowledge is embedded in automated pipelines, version-controlled systems, and governance processes, or concentrated in individual personnel. OECD surveys confirm that most AI firms primarily depend on trade secrets for core differentiation, making enforceability a function of operational discipline rather than legal form. 

AI models as depreciating economic assets

 

AI models defy traditional IP categorisation, but they cannot be overlooked in valuation. Their economic value can be assessed through performance persistence, retraining cost trajectories, and reliance on exclusive inputs. 

Research from MIT Sloan shows that models trained on proprietary or longitudinal datasets exhibit significantly slower performance decay compared to those trained on widely available data. This extends their usable economic life and directly impacts forecasted margins. IP analytics increasingly incorporates these decay curves into valuation models, treating models as assets with observable depreciation profiles rather than static code. 

Significantly, the model value cannot be separated from the data rights. Training permissions, downstream reuse rights, and geographic portability now influence both regulatory exposure and economic optionality. In AI-intensive transactions, data contracts often hold more significance than model architecture itself. 

Where IP analytics actually changes deal outcomes

 

Rather than scattering examples throughout, it is more illustrative to examine how different organisations demonstrate the same underlying valuation logic. 

NVIDIA Corporation provides a clear example of patent-centric system design. Its disclosed IP strategy encompasses hardware architecture, parallel computing frameworks, and AI-specific instruction sets. This cross-layer positioning increases dependency across the ecosystem and supports long-term pricing power. Analyst assessments consistently attribute a significant portion of NVIDIA’s enterprise value to the centrality of its architectural IP, rather than its product cycles. 

At a different scale, Darktrace plc’s investor disclosures highlight that its competitive advantage stems from proprietary models trained on cumulative enterprise telemetry. The defensibility lies less in algorithm novelty and more in data network effects and retraining economics. In this context, IP analytics prioritises data rights, customer consent structures, and model retraining cost curves. 

The Salesforce acquisition of Slack Technologies further illustrates the valuation of trade-secret-heavy assets. Public transaction documentation emphasises proprietary deployment workflows, enterprise security logic, and integration frameworks, which are primarily protected through confidentiality and operational embedding. These assets significantly influenced integration assumptions and synergy valuation despite limited patent exposure. 

Ultimately, Google’s acquisition of Fitbit showcases how data rights can influence or redefine the value of IP. Regulatory commitments regarding data usage directly impacted the economic scope of AI-driven health initiatives. In valuation terms, enforceable restrictions altered expected return profiles despite unchanged technical capability. 

Across these examples, while scale varies, the analytic principle remains consistent: IP value is determined by enforceability under real operating conditions, not formal classification. 

Embedding IP analytics into valuation mechanics

 

Leading acquirers now integrate IP analytics directly into pricing models. Patent influence scores inform revenue durability assumptions. The strength of trade secrets affects discount rates and indemnity structures. Model longevity impacts earn-outs and deferred consideration. 

McKinsey’s analysis of technology M&A indicates that transactions where IP analytics shaped valuation inputs achieved significantly higher post-merger value realisation, especially in AI-heavy sectors. The causal factor is not merely improved diligence documentation, but better alignment between IP behaviour and financial forecasts. 

This integration necessitates collaboration among IP counsel, data scientists, engineers, and finance leaders. More importantly, it requires executives to interpret IP as a dynamic economic system rather than a compliance checklist. 

Conclusion: Valuation failure is now an IP failure

 

In deep tech and AI-intensive deals, valuation errors increasingly arise from misunderstandings of intellectual property dynamics. Overpaying is not the primary risk; rather, it is misjudging how patents, trade secrets, models, and data rights interact under scale. 

When applied rigorously, IP analytics reveals where advantages are sustainable, where they diminish, and where regulation or replication can erode value faster than financial models predict. As AI continues to concentrate value in intangible systems, dealmakers who view IP as static documentation will consistently misprice risk. 

The future of deep tech valuation belongs to those who can perceive IP not as a mere asset list, but as an operating system for long-term economic performance.