The period of easily creating billion-dollar companies using low-cost software is over.
In the last ten years, creating a billion-dollar company required cloud hosting services, a simple user interface, and a lot of salespeople. Now those things don't matter. The costs of computing power are causing pricing compression on all traditional software. Venture capital is fleeing basic applications and heading toward all of the hardware.
Of all the global venture capital funds, 36% has shifted to deep tech startups focused on what we now refer to as hard science and the latest advancements in engineering.
A deep tech startup does not just build another app on existing networks. It invents entirely new physical technology. These companies rely on fundamental breakthroughs in science and engineering, requiring long research cycles and heavy upfront capital to solve problems that were previously impossible to fix.
In 2026, to accurately determine the market's direction, you will need to consider hardware materials and physics. Simply acquiring users is no longer enough for success in this new economy; rather, you must focus on making meaningful advances in science. The market is rewarding advances in the fields of advanced packaging, photonics, quantum computing, synthetic biology and inter-connectivity among aerospace networks.
How Microchip Makers are Making More Money Than Software Companies
Historically, it was accepted wisdom that software generated higher margins than hardware, and that hardware was a low-margin trap for investors and manufacturers. By 2026, this concept was completely inverted based on financial results.
The S&P Semiconductor Index registered a 15% increase between the end of 2025 and the start of 2026. The same time frame saw software/services revenues falling by more than 24%. Because of the high costs involved with running complex AI models today, the profitability of software is compressed. At the same time, the companies producing the physical computing equipment have at least doubled their margins.
In 2025, the level of private investment in the US for computing equipment was $280 billion on a run-rate basis. This was almost double that of 2023... And most importantly, this investment has been directed away from the front end and into the infrastructure.
Understanding Technical Risk vs Market Risk
Engineering can be thought of in two ways: in terms of market risk and technical risk. A standard startup has designed its technology from existing technologies and has solved an existing market problem, so their risk is limited to that.
Conversely, a deep tech startup has invented an entirely new technology and has solved a market problem that was not previously possible to solve (or even existed). Therefore, a deep tech startup's risk is majorly due to the unknowns related to whether it can actually develop the new technology.
Standard startups are typically focused on short-term R&D cycles as the underlying technology has been established as being viable, whereas deep science startups are engaged in long-term R&D cycles and require significant capital investment up front to support the lengthy development process. Once the technology becomes viable, however, the intellectual property surrounding the technology provides a protected market for the company.
With $200 billion per year of global investment in AI, AI accounted for 61% of total venture capital in 2025. Upon further investigation, however, the majority of the funding for AI has been invested in a very limited number of types of AI, with most concentrated in two categories (high-speed memory and advanced packaging).
Therefore, founders raising capital in this environment must not use traditional metrics for determining company growth or success (such as active users), but rather be assessed based on their development or evolution through "technical gates. The higher the number of successful technical gates that have been passed, the greater the valuation of the startup. The distribution of capital takes place over multiple steps and is closely associated with individual technical milestones.
Building a Minimum Viable Technology Instead of a Simple Product
Small software companies, as an example, have developed minimum viable product (MVP) to quick test the market on their products. An example of an advanced engineer's technology would be the inability to develop a minimum viable technology (MVT) as an MVP. One cannot build or operate half a nuclear reactor. This leads teams on advanced engineering to focus on Minimum Viable Technology.
Minimum viable technology (MVT) represents proof that the underlying scientific principal functions in an operational environment. The first engineering (Nuclear Reactor; Electrical Generation) milestone, is developed to remove the highest risk element of the technology, before moving to operational scale.
Investing in either Dealroom or Preqin allows investors to evaluate and analyse technical milestones and to map technical dependence against valuation multiples. Once a team at these deep tech startups has achieved their baseline technology milestone, they will gain access to Series A and Series B funding and funding for the manufacturing infrastructure.
Selling High-Tech Equipment to the Military and Businesses
Historically, the defense marketplace has been one of the most rapidly growing capital markets. In 2025, government defense venture funding reached $21.4 billion from approximately 306 individual transactions, an increase of almost triple from the level of funding received in 2023.
The procurement cycles of government agencies are notoriously long, and a sale to a military agency may take five years to ten years. To manage the cash flow challenges associated with long procurement cycles, the best defence technology companies develop products that have both immediate commercial potential and meet military screens.
A dual-use (or dual-purpose) company is a commercial entity that designs and develops technology with both immediate commercial applications and military screens.
Of the hundreds of defence-related companies that received funding in 2025, 755 operated as dual-use companies and only 284 operated as purely defence-related. Agricultural firms could buy drone systems to keep an eye on how their crops are growing, all the while using the system to start selling their technology to the military for future contracts.
This would generate revenue from the agriculture business while allowing the company to keep its investors happy and keep the company afloat during the long wait for approval from military agencies to procure drones.
Where the New High-Tech Hubs are Growing
Innovation is currently strengthening geographically, with approximately $690 billion in advanced engineering across the combined European networks.
Recently, the city of Paris has surpassed London as the largest hub for deep tech startups in Europe, primarily due to extensive government funding and an abundance of engineering talent in the city. However, capital investment remains unevenly spread throughout Europe, with 70% of the European mega rounds being financed by investors from outside of Europe.
Why American Companies Buy Up European Inventions
Europe has the potential for high-caliber academic/scientific outputs, however, European academic institutions will not see the financial returns that will accrue to the United States. Currently, it is shown that, on average, 75% of the exit value of European-based academic companies is gained by U.S.-based acquirers.
There are few European IPOs in the advanced engineering space. The most frequent routes to exit are consolidation via acquisition. Corporations will often purchase startups for their technology (IP) and for access to the engineering teams.
Over the past 12 months, Synopsys purchased Ansys for $35 billion; Nvidia purchased the inference technologies of Groq for $20 billion; Skyworks purchased the assets of Qorvo for $9.8 billion.
Large corporations use small startup companies as a source of externalised research and development (R&D). They allow the venture capital market to absorb the initial technology risk until they purchase the successful startup at a very high price.
How Big Businesses Pick and Test New Science Partners
For corporate innovation leaders, vendor selection from a hard science vendor is a highly complex process. A software subscription is not typically what you purchase; rather, you are selecting a technological solution that is unproven and needs to be thoroughly integrated into your existing physical supply chain.
Finding the reality of a product among the sea of marketing is a difficult task for all of us. Accelerators such as Hello Tomorrow Challenge screen thousands of submissions every year; in 2026, 100 companies were selected from more than 4,800 submissions across 108 countries; even with that level of filtering by accelerators, corporate buyers must develop detailed internal guidelines to evaluate a vendor.
The technical evaluation of a new technology vendor when running a pilot program must include several specific technical measurements, as follows:
- Time and Complexity of Integration: Track exact time (in hours) required to integrate the new technology into existing legacy systems.
- Ownership of Data: Clearly delineate who owns the data created in the pilot.
- Regulatory Compliance: Verify that the hardware meets current regulations regarding export and is compliant with regional medical and structural regulations.
- Performance Metrics: Document the improvement in speed, cost, or accuracy of the new technology, as compared to the existing solution.
Without comprehensive service level agreements (SLA) associated with the above elements, there is no chance the pilot project will advance past the initial phase. Enterprise teams don't have the luxury of serving as free testing labs for pre-commercial companies.
Final Words About Deep Tech Startups
The major business reality for the infrastructure industry in 2026 is that capital will be tied to the infrastructure itself. The massive shift in venture funding toward advanced science-based technology (which currently provides 36% of the world’s deployment), is a permanent change in the structure of how companies operate. Due to processing costs, software profitability is becoming increasingly constrained, necessitating an increase in gross margin for both semiconductor and hardware products. Companies are no longer investing in the traditional software application; they are investing in the physical and quantum architectural elements necessary to enable large-scale computing and military responsiveness on a worldwide scale.
The absolute precision required by enterprise buyers and investors will define success in this industry. The minimum acceptable technology product must be mapped to its respective funding points, and a dual-use revenue model must exist in order to address protracted procurement cycles. A thorough analysis of the intellectual property associated with a product must occur prior to acquisition. Since the marketplace will continue to see a substantial number of acquisitions—primarily by large U.S. corporations—which will be assimilating and consolidating the finest technical assets from around the globe; those who understand the economics of hardware transformation will capture the considerable enterprise value created over the next 10 years.