Key features of Deep Tech

Deep Tech is better thought of as a horizontal rather than a vertical—covering a number of sectors (e.g. medicine, biotech, energy, carbon capture, and material science), and technologies (e.g. machine learning, quantum, spatial computing, robotics, computer vision), and so on.

Here are some key features that encapsulate ‘Deep Tech’: 

• Breakthroughs rooted in science or engineering—usually the result of extensive research and development, emerging from labs or universities

  • e.g. Monolith AI (FF18) emerged as a result of founder Dr Richard Ahlfeld’s PhD research at Imperial College, exploring uncertainty quantification in manufacturing and engineering. This formed the basis for Monolith’s core AI platform allowing engineers to fast-track product validation

• Technology first—often only once the technology has been developed will founders identify an exact problem for it to address, rather than the other way around (‘solution in search of a problem’) 

  • e.g. Jeevan (FF23) also emerged from a university science lab, the result of Dr Arup SenGupta’s research at Lehigh University. Their core HIX Technology (a nanotechnology resin platform) was initially developed and patented as a water purification technology, commercialised and scaled widely through Drinkwell, before realising its valuable potential for direct air capture (DAC)

Visualisation of Jeevan's direct air capture technology

• Enabling technology—something that is a building block for other technologies, rather than an application technology that is restricted to one particular purpose

  • e.g. NVIDIA is one of the most successful examples of this, as a foundation for numerous advanced technologies. While their GPUs were initially manufactured with the gaming industry in mind, their application for training AI models has seen the business sky-rocket 

• Often blends hardware and software—not always the case, but deep tech usually involves combination of software/digital technology alongside a physical, real world element

  • e.g. Wayve has developed groundbreaking architecture for autonomous driving, sold as a software (their AI Driver operating system) that powers hardware (AVs) 

The challenges

The distinct challenges around deep tech can be categorised into four clear categories. 

Talent

The personnel and expertise required to build in deep tech creates a much higher barrier to entry. Take Peptone (FF18), a biotech company that has built an AI engine for protein drug discovery. Founder Kamil Tamiola, himself a PhD in biomolecular science, has built a team with deep expertise not just in their specific field of disordered proteins but in pharmaceuticals, AI, and biotech. Not only is finding this hyper-specific talent challenging, it requires considerable energy (and significant funding) to persuade them to join. 

For this very reason, however, deep tech has considerably lower competition risk. Technologies are fundamentally harder to build, meaning that assembling the right talent and overcoming technological risk gives deep tech an inherent defensibility that digital tech doesn’t have. 

Funding

For many investors, funding deep tech requires a large leap of faith. Not only is deep tech far more resource and capital intensive (requiring greater upfront investment) than digital tech, it comes with higher technical risk and longer payback periods. 

This sentiment has certainly changed given the success of some of the world’s biggest deep techs—but it is still often a risk too far for your typical investor. Founders are better off engaging with seasoned deep tech investors who better understand these risks, timelines, and dynamics.

Data does suggest that this funding discrepancy narrows over time, showing that deep tech companies take the same amount of time to reach $1m in revenue (approximately 2 years) and just 11% more capital to reach $10m in revenue. 

One caution is the funding ‘valley of death’: the point at which you need considerable upfront investment to start manufacturing a physical product (or with AI, training your model). There are ways to navigate this—co-development with an industrial partner or existing manufacturers, or even exploring a tech licensing model—but this still carries considerable risk. 

Product development

Product development is starkly different for deep tech. Firstly, much more work is required on discovery. If you are starting with a technology, rather than a problem, this requires considerable effort from the start to narrow down your focus.

Unlike the usual ‘build, test, measure, learn’ cycles that you go through when building software, deep tech is more drawn out, with far more uncertainty. It is much more linear (represented through the Technology Readiness Levels metric), as opposed to typical cyclical testing and iteration processes. 

Reaching your point of MVP is much further along for deep tech than it is for digital tech. This is because it is much harder to optimise things once a product is in production, and even harder once a prototype is out there and needs to be recalled and improved. 

GTM

Once you’ve developed your product, GTM is much more complex. When you are creating a new market, you have the benefit of novelty: but there is also no existing market and no comparable competitors. You can’t target existing users if there are none. 

This also reveals a bigger challenge around deep tech, that it is not built with a buyer in mind. Novel technologies are harder to articulate, especially when they have a highly technical foundation that requires translation into ‘human’ language. Deep tech is typically more than just an ‘optimisation of X’, requiring more education when selling to potential customers.