Deep Tech and the Commercialisation Models That Actually Work

The term was popularised by Boston Consulting Group and Hello Tomorrow, who defined deep tech as ventures built on substantial scientific or engineering breakthroughs — as distinct from ventures that repackage existing technology into a new business model. Their research frames deep tech as the fourth wave of innovation, following the first two industrial revolutions, the era of corporate labs like IBM and Xerox PARC, and the venture-backed Silicon Valley model of digital and biotech startups built through the 1980s and 90s.

Three characteristics separate deep tech from conventional innovation:

• Technical risk precedes market risk. A fintech app risks whether customers will adopt it. A gene-editing platform, a fusion reactor, or a new battery chemistry first has to face the risk that the underlying science works at all.
• The moat is hard to copy. Deep tech is typically protected by patents, trade secrets, or tacit know-how accumulated over years of lab work — not by a first-mover advantage in a crowded app store.
• Time horizons are measured in years, not sprints. Aggregate global private investment in deep tech categories tracked by BCG and Hello Tomorrow soared by more than 20% a year from 2015 through 2018 to reach almost $18 billion, and by BCG's own later estimate, investment more than quadrupled from $15 billion in 2016 to over $60 billion in 2020 — capital that is patient by necessity, because the underlying technology takes years to mature.

Three sectors, three different commercialisation clocks

Deep tech's abstractions get more useful once grounded in real outcomes. Three sectors — healthcare, agriculture, and energy — show three very different paths from lab to market, and three different lessons about what commercialisation actually requires.

Moderna's mRNA platform — a decade of unglamorous groundwork before a five-day sprint Moderna's COVID-19 vaccine looked like a five-day miracle: the company took a coronavirus vaccine from sequencing to human trials in 63 days. That speed was only possible because of what came before. Co-founder Noubar Afeyan has been explicit that the more than a decade of prior experience developing mRNA as a programmable "bioplatform" — a system that could be quickly redeployed and tailored across diseases rather than built bespoke each time — is what let Moderna move so fast once the pandemic hit. As CEO Stéphane Bancel put it, the company and its investors had spent heavily on the underlying technology for a decade, investing more than $2 billion over ten years before that capital ever produced a marketable product.

Pivot Bio — engineering microbes to replace a century-old chemical process Pivot Bio spent more than a decade studying how nitrogen-fixing microbes work in nature before developing the first commercial nitrogen-fixing microbes for row crops, engineering soil bacteria to convert atmospheric nitrogen into a form corn plants can use directly at the root — a direct challenge to the synthetic fertiliser industry built on the century-old Haber-Bosch process. The commercial proof followed the science: in 2023, U.S. farmers used Pivot Bio's products on over 5 million acres of corn, up from just over 3 million acres in 2022, and a University of Kentucky study found that replacing 40 pounds of synthetic nitrogen with the company's microbial product increased corn yields by an average of 11 bushels per acre.

Commonwealth Fusion Systems — selling the outcome before the reactor exists CFS, an MIT spin-out founded in 2018 to commercialise tokamak-based nuclear fusion using high-temperature superconducting magnets, has not yet produced net energy — first plasma and scientific breakeven are now targeted for 2027, and the technology risk remains real. What makes it a commercialisation case study rather than a pure science story is the financing model: Google and Eni have already signed power purchase agreements covering more than half the output of CFS's planned commercial plant in Virginia, and CFS became the first fusion company to submit a grid interconnection application to PJM, the largest U.S. wholesale electricity market. In July 2026 alone, the company raised $1 billion from pension and sovereign wealth funds, bringing total capital raised to $4 billion — roughly 30% of all private investment in fusion globally, and a signal that patient institutional capital now treats fusion as a real infrastructure asset class rather than a speculative bet.

The lesson across all three is the same: none of these companies commercialised the science by waiting for it to be finished and then looking for a buyer. Moderna built a platform, not a product, so it could redeploy fast once a market emerged. Pivot Bio proved yield economics on real farms before scaling acreage. CFS sold contracted demand years before the reactor produced a single watt. In every case, the commercial model was designed in parallel with the science — not bolted on afterwards.

Why researchers need to understand the problem before the solution

Deep tech's core failure mode is not technical — it is a failure of orientation. A researcher who starts from "what can this discovery do" rather than "what does an industry urgently need" tends to produce elegant technology in search of a market, rather than a market-ready answer to a real constraint. Three disciplines separate the two:

1. Real industry problems, not adjacent ones

Nitrogen loss to the environment, vaccine platform speed, and grid decarbonisation are all problems industries were actively paying to solve before the deep tech arrived. The research didn't create demand — it met demand that already existed, with a fundamentally better mechanism.

2. Market needs, not just technical elegance

A more nitrogen-efficient microbe only matters commercially if it's priced, packaged, and applied the way a farmer already works — as a seed coating or furrow-applied liquid, not a lab protocol. Deep tech that ignores the operational reality of its buyer stays a paper without an origin story.

3. Scalability, designed in from day one

A platform that can be redeployed against new targets (Moderna's mRNA backbone) scales faster than a bespoke solution built once. Researchers who ask "can this be repeated at 10x, 100x scale" before publication — not after — build technology that survives contact with a market.

The Malaysia challenge: strong science, weak conversion

Malaysia's innovation policy has the right ambition on paper. The New Industrial Master Plan 2030, unveiled on 1 September 2023, targets a 6.5% annual GDP boost in manufacturing and a RM587.5 billion contribution to national GDP by 2030, structured around 21 strategies and 62 action plans across four missions. MOSTI's National Technology and Innovation Sandbox has run since 2020 and has supported over 220 projects, commercialised 77 technologies, and generated RM157 million in value — real, if modest, proof that structured pathways move research to market faster than ad hoc technology transfer.

_{Sources: WIPO Global Innovation Index 2025 (Malaysia country profile); MOSTI/MRANTI, via Malay Mail}

That input-output gap is the single most important number in Malaysian innovation policy. Malaysia scores well on research spending, STEM graduate output, and patent filings — but far weaker on turning those inputs into commercial outcomes. Independent analysis of the 2025 ranking points to the same root cause repeated across Malaysian institutions: universities generate intellectual property that is evaluated for academic prestige — rankings, publication counts, faculty promotion — rather than for industry fit, so patents get filed but rarely licensed, and start-up formation from university research stays rare relative to research volume.

Four commercialisation approaches that close the gap

None of these is a novel invention — each is already working somewhere, in the case studies above or in Malaysia's own emerging sandbox model. What's needed is applying them earlier in the research cycle, not after a technology is already built.

01 — Problem-first research selection Start the grant proposal with the buyer, not the hypothesis Before committing lab years to a research question, require a documented industry pain point and a named category of potential first customer. This is the discipline that let Pivot Bio and Moderna commercialise fast once the moment arrived — the market question was answered years before the product was.

02 — Platform over single-product framing Build the reusable core, not just the first application Moderna's mRNA platform could be redeployed against Zika, MERS, and then COVID-19 within weeks because the underlying system — not just one vaccine — was the asset. Researchers who architect for redeployment dramatically compress the second and third commercialisation cycles.

03 — Milestone-based, patient capital Fund the science in stages tied to technical proof, not fixed timelines CFS's pension-fund-backed raise and pre-sold power purchase agreements show institutional capital is willing to underwrite long technical horizons — if the milestones and offtake are structured transparently. Malaysian funds and grant bodies can adopt the same staged logic rather than one-shot annual grants.

04 — Structured sandboxes with named industry partners Pair every research track with a real operator from day one MOSTI's National Technology and Innovation Sandbox already does this across agritech, biotechnology, robotics, and AI — running sector-specific pilots with named regulators and adopters, not generic accelerator mentorship. Scaling this model and tying university tenure and grant renewal to sandbox participation would directly address the GII's input-output gap.

What builders and policymakers should take from this

For researchers: the question "does this work?" is necessary but not sufficient. Pair it, from the earliest grant proposal, with "who has this problem today, and what would make them switch?" Technical elegance without an identified buyer is a paper, not a venture.

For founders and R&D leads: design for redeployment. A platform that can be repointed at a second and third market compresses your entire commercialisation timeline — the difference between Moderna's 63-day sprint and a decade-long bespoke development cycle was architecture, not luck.

For Malaysian policymakers and university leadership: the GII's 30th-in-inputs, 39th-in-outputs split is the clearest evidence available that more grant funding alone will not close the gap. What moves the needle is tying research incentives — tenure, renewal, institutional ranking — to demonstrated industry engagement, and scaling structured sandbox mechanisms like NTIS from pilot to default pathway.