Liza Poliakova, University of Oxford
On a lava plain outside Reykjavik, a new piece of plumbing offers a glimpse of a lower-carbon future. Climeworks’ ‘Mammoth’, the world’s biggest direct-air-capture plant, silently removes CO2 from the air. Yet this engineering marvel, scaling to 36,000 tons yearly, secured just $650M in equity after an 18 year-long struggle. A locus classicus positive externality, this plant’s societal benefit vastly exceeds its private return. Meanwhile, carbon-accounting startup Watershed secured an over $1B valuation for software that tracks emissions it cannot reduce. Capital market failure is at the centre of the paradox which defines climate tech’s crisis: Silicon Valley’s obsession with scalable, asset-light SaaS has created a funding abyss for the heavy industries. Robert Smith, VP of Technology at Greensoil PropTech, laments whilst software identifies ‘energy and carbon emission reduction opportunities, it’s hardware that translates these insights into action’.
Silicon Valley’s misalignment with climate imperatives is pernicious and stark; industrial processes generate 34% of global emissions, but attract only 7% of the cash. Andrew Symes, CEO of company OXCUU which converts greenhouse gases into plastics and jet fuel, argued that this funding produces a second-order effect: talent allocation failure. Discussing the sector bias against hardware, Symes cites the perception of hardware as outdated compared to the ‘glamourised ‘digital’ economy’, as well as the drain in engineering talent as the brightest gravitate towards finance, AI and software. He predicts that breakthroughs that will drive our sustainable future will not be ‘born from code or finance alone but from the minds of engineers willing to work in the physical realm’. Amidst this misallocation of human capital, the Solow Swan growth model, which maintains that long-run growth depends on capital stock transformation and technological progress, remains as relevant as ever. We cannot software-optimise our way out of emissions from cement, steel, chemicals, and heavy transport, or overlook their seminal requirement in process innovation.
Another reason for this economic misallocation is fund structure itself, with investors favoring the scalability and predictability of software startups. VC funds, with typical 10-year lifespans, operate on the time value of money and a portfolio theory that demands a few ‘unicorn’ exits within a decade to deliver target return. Hardware, on the other hand, often takes over 15 years to bring return on investment. The risk profile of hardware tech is off-putting for many investors, with higher manufacturing, supply chain, and market risks compared to SaaS. Exit uncertainty is even worse. Plenty of buyers covet a SaaS app; few want a half-built fusion reactor. Silicon Valley Bank quantifies the divergent profitability: the median software company with over $50M revenue has a 30 percentage point higher profit margin compared to hardware in 2024.
In their appositely titled MIT working paper ‘Venture Capital and Cleantech: The wrong model for energy innovation’, Benjamin Gaddy et al. contend the ‘cleantech boom-and-bust’ has been instigated by the poor performance of investments in deep technology innovations. Aggregating hundreds of investments in their study, Gaddy et al. found that in Series A round investments in new materials and processes, along with investments in hardware integration companies, lost around $1.25 billion, whilst software was the only class of investment that rewarded investors, returning nearly $550 million on investment of over $150 million.
(Source: Venture Capital and Cleantech: The Wrong Model for Clean
Energy Innovation An MIT Energy Initiative Working Paper July 2016)
The consequences of this misallocation are materialising. Data centres epitomize this dark irony of skewed investment. Before the AI surge, data centres consumed only 1% of global electricity before the AI surge. Now fuelling the digital boom, projections currently estimate that data centers could account for 11.7% of U.S. power demand by 2030. Rather than enabling decarbonization, this software revolution risks fossil fuel lock-in, particularly in America, the country with the biggest AI power consumption, which is on track to build 80+ gas plants.
Some investors are adapting, but correcting this failure requires investors and governments to reshape the underlying economic incentives. Pioneering the disruption of traditional VC myopia are Lowercarbon Capital and Breakthrough Energy Ventures (BEV). Lowercarbon Capital, which funds portfolio companies removing over 1 million tonnes of CO2 yearly, has longer horizons, deeper technical diligence and accepts higher risk. Targeting gigaton-scale solutions, it has entered into long-horizon tech such as Renaissance Fusion Similarly, BEV, founded by Bill Gates in 2015, runs 20-year investment horizons with recalibrated risk-tolerance. These specialized funds essentially act as arbitrageurs, exploiting the market’s bias, operating with 20-year horizons, recalibrated risk tolerance and an explicit mandate to fund ‘negative NPV, positive Earth’. Both counter Silicon Valley’s ‘scale fast’ dogma by proving that hard tech can yield market-rate returns; market failure lies not in hardware economics per se, but in the mismatch between hardware timelines and fund structures.
Government policy is also evolving in the right direction. America’s Inflation Reduction Act, allocating $369 billion to climate provisions and decarbonization projects, offers DAC operators a tax credit of up to $180 per tonne of CO2 captured. This essentially functions as a Pigouvian subsidy, aiming to bridge the gap between private cost and social benefit. The EU Clean Industrial Deal first unveiled in February, promises to de-risk hydrogen investment and make electricity cheaper for manufacturers. The creation of an Industrial Decarbonisation Bank which ‘will complement the ETS price signal and help bridge the funding gap in both capital and operational expenditures’ has also been proposed, a top-down attempt to create demand-side pull for low-carbon industrial products.
But will the mainstream pools of capital really follow? New ‘hardware-friendly’ climate funds blending venture money with patient institutional cash are emerging, but scaling must accelerate if we are to reach EU goal of net-zero by 2050. Until finance flows to the engineers building in the factory floors, rather than the coders in Silicon Valley, ‘net zero’ will remain a ledger entry, not a livable future.