Patenting Strategy for photonics innovations enabling AI

The landscape of artificial intelligence (AI) is evolving at unprecedented speed, pulling along supporting technologies in its slipstream. Photonics is expected to serve as a key enabler for AI by providing technologies such as optical connectors and photonic integrated circuits that enable the fast, reliable data transfer required by AI workloads while improving the energy efficiency, driving down the cost-per-token.

 

Traditionally, many organisations have prioritised maintaining a robust patent portfolio due to the costs associated with developing and manufacturing photonic components. However, with the velocity of innovation driven with AI, many organisations are re-evaluating whether the traditional value of patents has diminished, and whether the pursuit of a robust patent portfolio is commercially viable.

 

Photonics as an AI Enabler

Photonics is seen as a key enabler for AI in various ways, from moving information more efficiently through the compute infrastructure that trains and serves models, to capturing real-world data that those models learn from and act on. Here we discuss some examples of where photonics can create leverage for AI.

 

Data-center connectivity and optical connectors

AI training and inference at scale are ultimately constrained by how quickly and how efficiently data can move between accelerators, memory, and storage. The use of traditional copper connections and the power cost of pushing multi‑hundred‑Gb/s signals off-package are becoming limiting system constraints.

 

Photonics can help relieve this bottleneck by enabling high-bandwidth, lower-loss optical links (within and between data centers) and by supporting integrated photonic components that can improve the density and energy efficiency of optical I/O.

 

As a clear example of this shift toward bringing optics closer to the switch and compute packages, NVIDIA [1] are developing a co-packaged optics platform for data center networking, utilising optical I/O as a path to higher bandwidth density and improved energy efficiency to help scale AI implementation.

 

Similarly, Intel [2] has demonstrated a fully integrated optical I/O “optical compute interconnect” (OCI) chiplet co-packaged with an Intel CPU and running live data, explicitly targeting AI/HPC infrastructure use cases.

 

These examples also illustrate a broader trend: optical interconnect is transitioning from “pluggable at the front panel” toward tighter integration at the package/board level, which creates new opportunities, e.g., photonic engines, coupling structures, packaging and test flows.

 

Edge sensing and data capture

Photonics can also help enable AI by providing the data that models learn from and act on. Photonics-based sensors (such as time-of-flight (ToF) and structured-light depth sensing, LiDAR, high-speed imaging, and spectroscopy) can provide richer, higher-dimensional inputs than conventional 2D imaging alone, improving both model training data quality and real-time perception in deployed systems.

 

In consumer devices, depth sensing has already become a mass-market example of photonics feeding on-device AI. ToF/structured-light modules (typically using near‑infrared emitters and dedicated receivers) are widely used for face authentication, portrait effects, and AR features, where the photonic subsystem generates a depth map that downstream algorithms interpret.

 

In automotive and robotics, LiDAR is a prominent case where photonics provides the core measurements and AI provides the perception and decision-making based on the core measurements. Here, rapid evolution is visible not only in sensor performance (e.g., range, resolution, frame rate, ambient robustness) but also in architectural choices (e.g., wavelength, scanning approach, solid-state integration, and sensor-fusion pipelines).

 

In industrial inspection and life sciences sectors, spectroscopy and hyperspectral imaging similarly translate optical signatures into features for AI classification (e.g., material identification, contamination detection, or biological markers), often driving coordination between the optical front-end design and the downstream model.

 

Why are some organisations de-prioritising patents?

Clearly there are multiple opportunities for patentable innovation within the photonics-for-AI sector. So why are some organisations turning away from pursuing expanded patent portfolios?

 

The answer I’ve been given by the majority of people I’ve spoken to about this issue is simple: Time.

 

Technologies are advancing so quickly that the commercial window for any single implementation can be shorter than the time it takes to obtain a patent protecting that implementation. The European Patent Office (EPO) estimates that the grant procedure takes about three to five years from the date an application is filed, but products may become obsolete within as little as 1-3 years – sometimes less than half the time it takes to get a patent granted.

 

Furthermore, most organisations have limited resources, both in terms of finances and time to invest in in drafting, filing, and prosecuting a patent application to grant. Many are forgoing the financial and time investment in favor of driving product development and to try and reap the rewards with being first to market.

 

So why bother patenting something that will be obsolete before a patent protecting it is granted, and may cost you more than it is worth?

 

Risks of not protecting short-lifetime innovations

It can be tempting to dismiss short-lifetime innovations as “not worth patenting”, particularly where a specific implementation may be obsolete before a patent is granted. However, choosing not to protect these innovations can create practical risks, particularly where value is captured through means that competitors can adopt quickly.

 

Even when a feature has a short commercial window, it can form a steppingstone into the next generation. Competitors who can quickly copy may use your solution to accelerate their own roadmap and potentially file follow-on patents around improvements, which may leave you defending freedom-to-operate while they build the longer-lived patent portfolio position.

 

Furthermore, many short-lifetime innovations take the form of incremental advances that make a particular product manufacturable, reliable, and cost-effective at scale. If these advances can be reverse engineered or inferred from compliance behavior, then without patent protection in place, competitors may replicate them with limited R&D cost, impacting on an organisation's differentiation and compressing margins in an already competitive field.

 

Additionally, patents on short-lifetime features can still have strategic value as leverage. For example:

 

• Negotiation leverage and cross-licensing: a small number of well-targeted filings can improve your position with suppliers, customers, and competitors, particularly where the market is focused on iterating on similar integration problems.
• Standards and ecosystem positioning: where interconnects and interfaces are standardised (or de facto standardised), having IP around implementation paths can influence design choices or strengthen your hand if others’ implementations converge on your approach.
• Investment, partnerships, and mergers & acquisitions: even early-stage or transitional IP can support diligence narratives by evidencing defensibility and a track record of inventiveness, which may matter when negotiating collaborations or acquisition terms.

 

Is There Still a Place for a Robust Patent Portfolio?

The answer to this question is nuanced.

 

Patent portfolios can remain valuable assets, but mainly when they concentrate on innovations that outlast fast product cycles. The goal is perhaps less of a “more patents” approach and more “the right coverage”, matched to how quickly standards, form factors and customer requirements are shifting.

 

Organisations can take several steps to keep patent coverage aligned with fast product cycles while still protecting long-term value, for example:

 

• Protect long-lifetime innovations by focusing claims on what customers qualify and rely on over multiple years to obtain patents that can stay relevant for longer periods.
• Prolong relevant patent families by using continuation/divisional applications wherever possible to keep claim scope aligned with the roadmap and regularly drop families that no longer map to products or standards.
• Plan for market changes by prioritising narrower scope, faster-to-grant filings (and selective use of acceleration mechanisms) and weigh patenting against cost and disclosure. Trade secrets may be better for short process windows or incremental integration choices.

Organisations should also maintain awareness of third-party rights within their sector and continue to monitor their own freedom to operate in view of those rights, which may change frequently as competitors seek to gain an advantage in a rapidly evolving field.

 

Conclusion: Finding the Balance

While the photonics sector supporting AI is marked by rapid innovation, it can be tempting to overlook building a patent portfolio due to the mismatch between the product lifetime and patenting process. However, a carefully maintained patent portfolio can still offer meaningful value.

 

By adopting strategies that balance applications targeting longer lifetime inventions with narrower-scope applications for a faster grant to keep up with the pace of technological change,  organisations can continue to protect their interests and capitalise on commercial opportunities.

 

Whatever patenting policy is chosen, a coherent strategy matters to safeguard critical innovations while remaining responsive to the constantly evolving driving force that is AI.

 

[1] https://nvidianews.nvidia.com/news/nvidia-spectrum-x-co-packaged-optics-networking-switches-ai-factories

[2] https://newsroom.intel.com/artificial-intelligence/intel-unveils-first-integrated-optical-io-chiplet

 

This briefing is for general information purposes only and should not be used as a substitute for legal advice relating to your particular circumstances. We can discuss specific issues and facts on an individual basis. Please note that the law may have changed since the day this was first published in June 2026.