Capital Is Not Scarce. It Is Highly Selective.

Capital is not scarce. It is highly selective.

Two accounts of energy transition finance circulate, and both are wrong. The first holds that capital has retreated and that projects would be built if the money were there. The second holds that capital is abundant and that deployment is only a matter of time. The 2026 data supports neither, because the aggregate figures that generate both readings describe different things and are routinely quoted as though they described one.

What the 2026 record actually shows is a market in which the total pool is very large, new commitments into it have slowed to the weakest first half on record, existing commitments are being drawn down at roughly four times the rate they are replaced, and no manager that went to market cut a target or abandoned a raise. Capital is not scarce. It is being rationed, and it is being rationed toward a specific kind of asset that most announced projects are not.

The constraint is not the supply of capital. It is the frequency with which Final Investment Decisions actually result in capital being received.

The clearest measurement of that conversion failure is hydrogen, where the units are physical rather than financial. The announced project pipeline for 2030 has shrunk by 10 million tonnes to 27 million tonnes. Projects that are operational, under construction or past Final Investment Decision account for 4.3 million tonnes of it. Of all the jurisdictions with a 2030 target, two are on track: the Netherlands and China.

This paper sets out what 2026 data shows about that conversion layer, why the conversion rate differs so sharply between jurisdictions, and what that difference is. The answer is not ambition, funding envelopes or political will, each of which is broadly comparable across the major markets. It is whether the support instrument creates an enforceable obligation on an identifiable counterparty — something a credit committee can take security over, or sue on. Where it does, plants get financed. Where it does not, they do not, however large the grant.

Every figure below was checked at its publishing source. All statements of position are as at 25 July 2026, in a year with five months still to run.

The weakest first half on record, and deployment running four times faster

Infrastructure fundraising in the first half of 2026 reached $40.8 billion — the weakest first half on record, below even the $71.6 billion raised in the first half of 2024. Read alone, that number supports the retreat thesis.

It does not survive contact with the deployment data. Preqin recorded $23 billion raised across 17 fund closes in the first quarter of 2026 against $85 billion of deal value, a ratio of nearly four to one. Infrastructure dry powder has fallen to 23% of assets under management. Capital is leaving the pool considerably faster than it is entering it, which is the signature of a market deploying hard, not one in retreat.

Nor did the managers who went to market struggle. Seven energy transition and next-generation infrastructure vehicles reached a final close in the seven months to late July 2026. Of the four that published a fund-level target, three cleared it and one fell short when measured on fund commitments alone. None cut a target or abandoned a raise. The other three report at programme or strategy level and published no target against which the raise can be measured.

†  Reported at programme or strategy level rather than as a single closed-ended fund. Partners Group's figure blends a fund with bespoke mandates investing alongside it; RGREEN INVEST reports INFRAGREEN V at strategy level. Neither published a target, so neither can be described as oversubscribed.
‡  Includes co-investment. Taaleri reached €630 million including €74 million of co-investment; on fund commitments alone it raised €556 million against a €600 million target, which is below target. Vesper's figure is total assets under management including co-investment vehicles against an €800 million fund target. Lime Rock closed at $640 million against a $650 million hard cap.

Those markers matter more than their size suggests, which is why we have set them in the table and not beneath it. Read at face value the set appears uniformly oversubscribed. Read properly, three of the seven closes cannot carry that description at all — and a paper whose central argument is that the market conflates targets, raises and commitments cannot afford to do it in its own evidence.

The aggregate picture is consistent. Global energy investment is expected to reach $3.4 trillion in 2026, a 5% rise on 2025, with around $2.2 trillion going collectively to renewables, nuclear, grids, storage, low-emissions fuels, efficiency and electrification, against some $1.2 trillion to oil, natural gas and coal — close to two dollars of clean investment for every fossil dollar. Networks alone are expected to absorb around $550 billion, up nearly 20% year-on-year, and renewable power around $665 billion, of which $365 billion goes to solar, a billion dollars a day.

Around three-quarters of anticipated 2026 energy investment is effectively locked in by decisions already taken. That is the number that should trouble a sponsor still seeking capital, because it means this year's allocation was largely settled before this year began, and the assets that captured it were the ones already bankable.

There is no dispute about which segments those were. Grids, data centres and utility-scale solar and wind in mature markets absorbed the capital. The segments that generated the announcement rhetoric did not. The next section measures the size of the gap.

Where the capital stops

Hydrogen is where the conversion problem is most cleanly measured, because the gap between what was announced and what has been financed is published annually in physical units by a single authoritative source.

Production is growing. Low-emissions hydrogen output rose 20% in 2025 to reach almost 1 million tonnes, and 2026 is expected to bring another record year, taking low-emissions hydrogen above 1% of global hydrogen production for the first time. On its own terms the technology is working.

The pipeline behind it is going the other way. The announced project pipeline for 2030 has contracted by 10 million tonnes, to 27 million tonnes, driven by cancellations and by projects slipping beyond 2030. Against that 27 million tonnes, committed production — projects operational, under construction or past Final Investment Decision — grew 3% to 4.3 million tonnes.

The IEA puts a deadline on the remainder. Committed production could rise above 6 million tonnes if projects with a strong chance of operating by 2030 take Final Investment Decision during 2026 or 2027. More than 100 gigawatts of announced electrolysis capacity will lose any prospect of operating by 2030 if investment decisions are not taken before the end of 2027.

Every other national hydrogen target now depends on projects that have not been financed and, on current evidence, will not be.

Both exceptions are jurisdictions in which the principal purchaser of hydrogen is either under a statutory obligation or under state direction — the Netherlands sits inside the European blending and sub-mandate framework described later in this paper, and Chinese demand is concentrated in state-owned industry. That is consistent with the argument this paper makes, and we do not push it further than that: two observations do not establish causation, and we found no 2026 source that isolates the mechanism in either case. It is the obvious question for further work.

The offtake data explains why, and it is the same finding from a different angle. New offtake agreements for low-emissions hydrogen reached 1.7 million tonnes per annum in 2025, level with 2024. As at the first quarter of 2026, more than 0.3 million tonnes had been contracted. The committed column is built from contracted volume, not announced capacity.

This is the layer at which the capital described in the previous section fails to arrive. It is not that the funds are empty. It is that a fund with a fiduciary duty and a defined deployment window cannot underwrite a project whose revenue is indicative. The money and the projects exist in the same market and do not meet.

What else could explain this

An argument built from a single year of data owes the reader the alternatives. Three explanations compete with the one advanced here, and each would be preferred to it if the evidence supported it.

This is the rate cycle, not a structural change

The most serious objection. The weakest first half on record is exactly what a tighter allocation environment produces, and explaining it requires no thesis about contracted revenue. If that were the whole account, however, two things would follow. Deployment would slow alongside fundraising, because allocators short of capital deploy less of it. And funds in market would miss their targets, because a constrained LP base does not fill books.

Neither happened. Deal value ran at roughly four times new fundraising in the first quarter, and dry powder fell as a share of assets under management, which is deployment accelerating relative to inflow rather than slowing with it. Of the four 2026 closes with a published fund-level target, three cleared it. A market short of capital does not fill books above target. Macro conditions plainly explain part of the fundraising softness; they do not explain a market that is deploying hard into some assets while a documented pipeline of others contracts by a quarter.

Allocators are retreating to the big names

A fair reading of the same fund data is that limited partners under pressure retreat to the managers they already know, and that the funds which closed did so on the strength of who was running them rather than what they intended to buy. Partners Group closing above $15 billion is at least as consistent with that as with anything argued here.

Two things argue against it. Emerging managers with a first-time fund in market were tracked as collectively targeting $14.4 billion, which is not what a market that has stopped backing anyone without a track record looks like. And the objection does not actually compete with the thesis so much as restate it one layer up: an allocator concentrating into managers with predictable deployment is rationing capital toward whatever is most contractible, which is the same behaviour this paper observes at project level. Both can be true at once. Either way the money is choosing, which is the point.

Hydrogen is a bad example

Conceded, and it should be stated plainly rather than defended. Hydrogen is the most extreme conversion failure in the transition, not the typical one. It is used here because it is the only segment where the gap between announcement and financing is published annually in physical units by a single authoritative source, which makes it measurable in a way that most sectors are not.

The argument does not rest on it. The Australian evidence later in this paper turns on biomethane and liquid fuels, the jurisdictional comparison turns on aviation fuel, and the fund data turns on infrastructure generally. Hydrogen is the clearest illustration of the mechanism. It is not the proof of it.

What would prove this wrong

Stated so a reader can hold this paper to it. If a material number of first-of-a-kind low carbon fuel projects reach Final Investment Decision during the remainder of 2026 without a contracted offtake, a mandated purchaser or a revenue certainty instrument behind them, the enforceability argument is wrong and should be discarded. If Britain awards its first revenue certainty contracts and the projects behind them do not move to financial close, the argument is also wrong. We will publish either way.

Five stages, one number

Part of the reason the conversion problem is persistently misdiagnosed as a funding problem is that the language used to describe capital does not distinguish between five materially different states. A target is not a raise. A first close is not a final close. A fund commitment is not a capital call. A capital call is not a drawn project loan. Each compression is defensible on its own. Together they systematically overstate how much capital a given project can actually draw, and when.

The most instructive example of the problem is one that has now been abandoned by the institution that created it. The $130 trillion figure attributed to the Glasgow Financial Alliance for Net Zero, launched in 2021, described the balance-sheet assets of member institutions rather than any capital earmarked for transition assets. As at mid-2026 GFANZ operates as a Principals Group with regional networks and thematic workstreams, publishing case studies on mobilising private capital, including a second blended finance study with British International Investment and BCG in June 2026. It no longer advances a capital figure of any kind. The number was not disproved; it was retired.

The Net-Zero Banking Alliance followed a similar trajectory, ceasing operations as a formal alliance in October 2025 and persisting into 2026 only as a guidance resource hosted by UNEP FI, with no successor alliance, and member banks reverting to their own targets. Neither tells a developer whether a plant can be financed, which was always the difficulty with reading a balance-sheet total as capital.

The practical consequence for a sponsor is straightforward. A headline number describing an aggregate stage-one or stage-three quantity carries no information about the availability of capital at stage five for a given asset. The only questions that carry information are what the revenue is contracted at, who is obliged to pay it, and what a lender can foreclose on if they do not.

Ambition is not the variable. Enforceability is.

Four jurisdictions have built support architectures for low carbon liquid fuels. Their stated ambitions are broadly comparable and, adjusted for the size of each fuel market, so are their funding envelopes. Their outcomes are not comparable at all. The variable is what a credit committee can do with the instrument.

A lender does not underwrite ambition, and cannot take security over a policy objective. It underwrites a claim against an identifiable party. Rank them by what can actually be enforced and they stop looking like four versions of one idea. They are different instruments.

The European Union created a buyer. The United Kingdom created a price. The United States created a cash flow. Australia created a subsidy for the seller. It created no buyer.

That distinction is the whole of it. Every jurisdiction that has financed a low carbon fuels plant did so by manufacturing a counterparty — someone who must buy, or must pay the difference. A grant reduces the quantum of capital a project requires. It does not create the revenue that services the capital that remains. It is the revenue that stops projects, not the quantum.

The United States adds a second lesson: an instrument can be materially weakened without being repealed, and the change is easy to misread. The 45Z credit survived the 2025 reconciliation act. What did not survive was the sustainable aviation fuel uplift, repealed for fuel produced after 31 December 2025, which leaves SAF earning the same credit as any other transport fuel rather than the enhanced rate the sector modelled against. Restrictions on foreign entities and foreign feedstocks were added on three separate commencement dates, and the implementing regulations were only proposed on 4 February 2026, with a public hearing at the end of May.

Federal lending shows the same pattern. The Loan Programs Office now operates as the Office of Energy Dominance Financing, and it has continued to write very large cheques: a $26.5 billion package to Southern Company subsidiaries in February 2026, the largest loan in the Department's history, and $17.5 billion of conditional commitments in June for the nuclear supply chain. In January 2026, however, the Department restructured, revised or eliminated more than $83 billion of prior loans and conditional commitments, of which roughly $9.5 billion of wind and solar was cancelled outright.

We do not have a published total for federal energy lending commitments across 2026, and we are not going to net one set of announcements against another to manufacture one. What the disclosed figures do establish is that the composition has changed decisively. The single largest commitment of the year went to a regulated utility with a rate base, the second to the nuclear supply chain, and the cancellations fell on wind and solar. Whatever the aggregate proves to be, the technologies that federal credit now favours are not the technologies it favoured two years ago.

A project modelled in 2024 on the pre-existing US architecture is not financeable on the same assumptions in 2026, and a sponsor who has not re-run the case since the reconciliation act is carrying a revenue line that no longer exists.

An instrument that exists is not an instrument that is operating

The United Kingdom has the best-designed instrument of the four. The Sustainable Aviation Fuel Act 2026 empowers the Secretary of State to direct a designated counterparty to offer a producer a revenue certainty contract, under which the counterparty pays the producer where an agreed strike price exceeds the market reference price, and the producer pays the counterparty where the reverse holds. The Act permits designation only of a company all of whose shares are held by a Minister of the Crown. It received Royal Assent on 5 March 2026, and every provision except section 1 commenced that day; section 1, the power to direct an offer, followed two months later on 5 May.

It is the only one of the four architectures that addresses the problem this paper describes head-on. It converts an indicative price into a contracted one, held by a counterparty a lender can name and sue.

Two things have not happened. No counterparty has been designated. And no contract has been offered, tendered or awarded.

The Department for Transport published its contract allocation strategy on 13 July 2026, twelve days before the date of this paper. It sets out a competitive process, which the Act itself does not provide for — the statute contains only a bare power to direct an offer. The first tranche is reserved for UK projects using non-HEFA technologies and feedstocks, up to a combined 230,000 tonnes per annum, together with a small number of first-of-a-kind projects. The indicative timetable runs as follows.

A developer pricing a British project off the Act today is therefore not pricing a revenue certainty contract. It is pricing an entitlement to apply, in a window that opens in roughly six months. The award itself is two and a half years out, against draft criteria, from a counterparty that does not yet exist. A credit committee cannot advance against that, and would be wrong to.

Legislation is a necessary condition for bankability, not a sufficient one. What a lender advances against is never the statute. It is the contract executed under it.

None of this is criticism of the British mechanism, which is the most carefully constructed of the four and will very probably work. It is a caution about reading a legislative achievement as a financing one. The gap between Royal Assent and the first executed contract is, on the government's own published timetable, thirty-one months.

This is the single most useful correction available to a sponsor building a financing case in any of the four jurisdictions. The relevant question is never whether a support mechanism has been announced, legislated or funded. It is whether a specific, named counterparty is currently under an obligation that a specific project can point to — and if not, what the shortest path to creating one is.

What closed, what did not, and the difference between them

Australia's low carbon liquid fuel ambition is not smaller than anyone else's. The A$1.1 billion Cleaner Fuels Program is, per dollar of national fuel demand, comparable with the instruments operating in Europe and the United States. ARENA was named its delivery agency on 13 May 2026, with production-linked incentives running over ten years and aimed explicitly at projects approaching Final Investment Decision. A further A$250 million sits open under the Future Made in Australia Innovation Fund's low carbon liquid fuels stream, and on 21 July 2026 ARENA made its first award from it: up to A$32 million to HAMR Energy for front-end engineering on forestry-residue-to-fuels projects at Portland in Victoria and Gillman in South Australia.

Deals do close in Australia. Two bioenergy projects have reached a Final Investment Decision or started construction so far in 2026. Both are instructive, and neither produces a litre of liquid fuel.

The pattern is the point. What has reached decision so far is biomethane and renewable gas — products with an existing domestic market, an identifiable purchaser and, in Delorean's case, grant funding covering close to half the capital cost. What has not reached decision, at any point in the year to date, is liquid fuels, where there is no purchase obligation on anyone, and where the demand measure that would create one was described in the Budget in the future tense and has not been introduced as a bill.

Australia can finance a bioenergy plant. It has financed two this year. Both had a buyer; neither makes a liquid fuel.

Delorean is the clearer case because the numbers are disclosed. A$30.5 million of grant funding against A$62.1 million of capital cost is roughly half the project, and the Final Investment Decision remains conditional on closing the balance. Grant capital did not make that project bankable on its own. It cut the amount the revenue had to carry, and the biomethane offtake carried what was left. Remove the offtake and no quantum of grant would have carried it.

The disadvantage compounds at the capital layer, and this is the part that is rarely priced. Every energy transition infrastructure fund reaching a final close in the first seven months of 2026 carries a European or global mandate. Not one of them is Asia-Pacific-dedicated. An Australian project seeking capital from a globally mandated fund is bidding for the same dollar as a European project whose offtake is backed by a statutory supply obligation. On identical engineering, identical sponsors and identical returns, the European asset wins on revenue quality alone.

That frames the Australian liquid fuels question precisely. It is not whether the money exists, and it is not whether the government is committed — A$1.1 billion and A$250 million answer both. It is that a A$600 million alcohol-to-jet plant in Townsville has no buyer obliged to take its output at a known price, and until one exists, the engineering will keep getting funded and the plant will never get built.

What actually finances a plant

If the constraint is enforceability rather than capital, the implication for a sponsor is that the order of work is different from the one most development programmes follow. The common sequence is to secure funding, then use the funding to complete engineering, then use the completed engineering to negotiate offtake. On 2026 evidence that sequence does not close, because the capital being sought at step one is being asked to underwrite a revenue line that does not yet exist.

The inverse sequence is the one that closes.

The reason this ordering works in 2026 specifically is that step five has become easy and step one has become hard. There is more committed, undeployed infrastructure capital than there are bankable assets to absorb it, which is precisely what a deal-to-fundraising ratio of nearly four to one describes. A project arriving at that market with an enforceable revenue contract is not competing for scarce money. It is a scarce asset competing for capital that has nothing enforceable to buy, and it will be priced accordingly.

The corollary is uncomfortable for sponsors who have already been to market. A project that has been declined was, in most cases, not declined on its engineering or its sponsor. It was declined because the revenue underneath it could not be enforced, and nothing about a second approach will change that unless the first stage is addressed. A prior decline rarely closes the door for good. But the door does not reopen on a better presentation of the same file.

Method and sources

Every figure in this paper was measured, published or reported between 1 January and 25 July 2026. Figures from earlier periods were excluded even where they were the best available, and no earlier figure has been carried forward, adjusted or presented as current. Where a source measures an earlier year but was published in 2026, the paper says which year is measured.

Every figure was checked against the publishing source itself rather than against secondary coverage of it. Where a figure could not be traced to a 2026 source it was removed rather than softened, and derived ratios are not presented as published figures.

This paper draws on no confidential information and no client material. Every figure is traceable to a published 2026 source named above.