Save the World or Ship the Product: The Climate Tech Reckoning
Silicon Valley has discovered the climate crisis. Now comes the part where we find out if that's good news or not.
The planet does not care about your product roadmap, your Series B timeline, or the fact that your pitch deck has a slide that says "10x impact by 2030." It operates on its own schedule, which is considerably less flexible than yours.
Yesterday we unpacked quantum computing and its cheerful promise to eventually break every encryption system currently protecting your bank account, your medical records, and roughly the entire digital civilisation we've spent fifty years building. You're welcome for the sleep. Today we are switching from the crisis in your computer to the crisis outside your window — or rather, increasingly inside it, given what summers have become. Climate technology: the sector that combines the ambition of the moonshot, the complexity of geopolitics, the patience requirements of geology, and the funding appetites of Silicon Valley. What could go wrong?
A lot, as it turns out. But also, possibly, quite a lot right. The honest answer — the one that nobody in either camp, optimist or doomer, wants to sit with — is that we genuinely don't know yet. What we do know is that the money is arriving, the technologies are maturing, and the gap between the press releases and the reality is wide enough to drive a container ship through. Several, in fact.
Let's go through what's actually happening.
01 — The Fusion Situation
Fusion energy has been thirty years away for seventy years. This is the joke. It's also increasingly, and somewhat awkwardly, becoming less funny — because for the first time in the history of the technology, several things are happening simultaneously that suggest the timeline might actually be compressing.
In December 2022, the National Ignition Facility in California achieved ignition — meaning they got more energy out of a fusion reaction than they put into the laser system that triggered it. This was, legitimately, a historic moment. The headlines were ecstatic. The caveats were, predictably, buried. The "energy in" calculation only counted the laser energy, not the energy required to power the laser itself, which was roughly one hundred times larger. Net energy gain, in the sense that matters for a power plant? Still not there. Still significant. Still a genuine milestone. Both things are true.
Private fusion investment has surged dramatically over the past five years. Companies like Commonwealth Fusion Systems, TAE Technologies, Helion, and a dozen others are pursuing approaches that range from conventional tokamaks to exotic configurations that would have seemed fanciful in a physics department a decade ago. Microsoft has signed a power purchase agreement with Helion for fusion electricity by 2028. This is either a visionary bet or a hedge that nobody actually expects to pay out. Possibly both.
Fusion promises clean, essentially limitless energy with no long-lived radioactive waste. It also promises this every decade, reliably, and has done so since Eisenhower was president.
The realistic picture: commercial fusion power before 2035 is unlikely. Before 2045 is possible. The question of whether it arrives in time to matter for the worst climate scenarios is a separate and considerably grimmer calculation. Fusion is real. Fusion is progress. Fusion is not a solution to a problem that needs solving in the next ten years.
02 — Sucking Carbon Out of the Sky
Direct air capture — the technology that does what it sounds like, pulling CO₂ directly from the atmosphere — is the category that produces the most interesting collision between genuine scientific validity and absolutely staggering cost problems.
The technology works. Climeworks, the Swiss company that operates the world's largest direct air capture plant in Iceland, is genuinely removing carbon dioxide from the air. The plant, called Mammoth, has a capacity of 36,000 tonnes of CO₂ per year. This sounds substantial until you note that global annual emissions are currently running at approximately 37 billion tonnes. Mammoth captures roughly one millionth of annual emissions. At a cost, depending on who's estimating, of somewhere between $600 and $1,000 per tonne.
The IPCC — the international scientific body whose job is to assess climate research — has concluded that meeting global temperature targets almost certainly requires removing CO₂ from the atmosphere, not just stopping new emissions. This is the uncomfortable context in which direct air capture exists. We need it. It currently costs too much. The price needs to fall by roughly ninety percent for it to be economically deployable at the scale required. The question is whether learning curves and scale effects can deliver that reduction fast enough, or whether this is a technology that will arrive in the 2040s having missed the most critical window.
For comparison: solar panels followed a learning curve that reduced cost by about ninety percent over twenty-five years. Direct air capture advocates point to this as the template. Critics note that solar benefitted from manufacturing simplicity that chemical carbon capture does not share. Both are correct. This argument will continue for at least a decade while the planet continues warming.
03 — Green Hydrogen: The Element of Disappointment
Hydrogen is the most abundant element in the universe. It is also, depending on how you make it, either a genuinely clean fuel or an elaborate way of using fossil fuels while feeling virtuous about it.
Green hydrogen — made by using renewable electricity to split water — is the version everyone wants. It can be used to decarbonise industries like steel production, shipping, and aviation that are extremely difficult to electrify directly. In principle, it is an elegant solution to a real problem. In practice, it is ferociously inefficient, expensive to store, difficult to transport, and prone to leaking through almost any container you try to put it in, because the hydrogen molecule is small enough to escape through materials that hold every other gas perfectly well.
The efficiency problem is fundamental: using renewable electricity to make green hydrogen and then burning that hydrogen to generate electricity gives you back roughly thirty to forty percent of the energy you started with. Using that same electricity directly gives you back eighty to ninety percent. For applications where direct electrification is possible, green hydrogen is a significantly worse option. The applications where it's better — long-haul shipping, industrial heat, aviation — are real and important. They are also a minority of the total decarbonisation problem.
Green hydrogen is not the universal clean fuel the press releases suggest. It is a specialised solution for specific hard problems, dressed up in rhetoric that implies it is something considerably more general.
The investment is enormous — hundreds of billions pledged globally by governments and corporations. The delivered projects, as of 2026, are a fraction of what was announced. This is a pattern in energy infrastructure. It is also a pattern that climate timelines cannot comfortably accommodate indefinitely.
04 — Vertical Farming and the Food System
Agriculture accounts for roughly a quarter of global greenhouse gas emissions. It also uses about seventy percent of the world's fresh water. It requires vast amounts of land, much of which was formerly forest. It is, in climate terms, a significant and awkward problem, because people have shown a persistent and inconvenient tendency to continue eating.
Vertical farming — growing crops in stacked indoor layers under artificial lighting — promises to address several of these problems simultaneously. Dramatically reduced water use. No pesticides. No weather dependency. Year-round production. Location flexibility. The ability to grow food in places that cannot conventionally support agriculture. These are genuine advantages.
The problem is energy. Crops grown in vertical farms require artificial lighting for every hour of photosynthesis. This is energy-intensive in a way that sunlight, which is free, is not. Early vertical farm companies — Plenty, AppHarvest, AeroFarms — raised enormous sums and ran into profitability walls that turned out to be, in some cases, insurmountable. AppHarvest went bankrupt. AeroFarms went bankrupt. Plenty pivoted. The cheerful projections turned out not to account adequately for electricity costs in a period when electricity costs rose sharply.
What remains is not a dead sector but a refined one. Vertical farming works well for high-value, high-water-content crops: leafy greens, herbs, strawberries. It works poorly for calorie-dense staples like wheat, rice, and corn, which are the crops that actually feed the world. As a supplement to conventional agriculture for premium produce, it is viable. As a replacement for the global food system, it is, at current energy economics, not.
05 — The Grid Is the Thing
Solar and wind power are, at this point, the cheapest form of new electricity generation in most of the world. This is not a projection or a hope. It is the current reality. The cost reductions over the past fifteen years have been more dramatic than almost any forecast predicted. This is genuinely good news, and it deserves to be stated plainly in a world where genuinely good news about the climate is not abundant.
The problem is not generation. The problem is the grid.
Renewable energy is intermittent. The sun does not shine at night. The wind does not blow on still days. Managing a power grid that depends heavily on intermittent sources requires either storage — batteries, at scale, which remain expensive and resource-intensive — or grid infrastructure that can move power from where it's being generated to where it's needed, across long distances, in real time. Most grids were not designed for this. Most grids are old. Upgrading them is slow, expensive, subject to regulatory processes that operate on timescales that make geological change look brisk, and reliant on supply chains for specialised equipment that are currently strained.
Smart grid technology — using software, sensors, and AI to manage power flows dynamically — is real, improving, and necessary. It is also not sufficient on its own. The physical infrastructure problem is a physical infrastructure problem. Software cannot substitute for copper wire and transformer capacity that does not exist.
The bottleneck in the energy transition, increasingly, is not the cost of renewable generation. It is the speed at which the infrastructure connecting that generation to consumers can be built. This is a planning problem, a regulatory problem, a supply chain problem, and a political problem, all simultaneously. It is the least exciting version of the climate conversation and probably the most important one.
06 — The Silicon Valley Problem
There is a particular Silicon Valley mode of engaging with large, complex, long-duration problems that does not always transfer well to the physical world. It involves moving fast, disrupting incumbents, achieving hockey-stick growth curves, and delivering returns to investors on a seven-to-ten-year horizon. It is a model that worked extraordinarily well for software. Software scales frictionlessly. Software doesn't require mining, permitting, construction, or forty-year asset depreciation schedules.
Climate technology is not software. Fusion reactors are not apps. Carbon capture plants are not platforms. A green hydrogen electrolyser does not benefit from network effects in the way that a social media platform does. The skills, timelines, and mental models that produced Google and Uber and Stripe are not automatically the right ones for producing the next generation of power infrastructure.
This is not an argument that technology cannot address the climate crisis. It plainly can, and in many cases is. It is an argument that the particular culture of Silicon Valley — the speed, the disruption mentality, the allergy to slow-burn infrastructure investment, the preference for elegant software solutions over messy physical ones — has limits when applied to problems that are stubbornly, irreducibly physical.
The most important climate technologies of the next thirty years will probably not be invented in a Palo Alto garage and scaled by a twenty-six-year-old with a laptop. They will be built by engineers, financed by patient capital, permitted by governments, constructed by workers, and operated for decades. The narrative around them needs to accommodate that reality, rather than forcing them into a venture capital story that may not fit.
Tomorrow we are going somewhere considerably more intimate than planetary-scale energy systems. We are heading inside the human body — specifically, into its genetic code. CRISPR therapeutics and the age of genetic engineering: what it can do, what it cannot do, who decides, and whether we are ready for the answers to questions we're only now learning to ask. See you then.
Switched On is a daily technology series covering the ideas, systems, and arguments shaping the digital world. Opinionated. Witty. Occasionally wrong. Always worth the argument.
