The Capital Radar, Issue 007

Every figure below was read from the Solandor database on 3 September 2026, and each record behind it names the public source it came from. Where something is not verified against a primary source, this issue says so rather than rounding the doubt away.

This issue is about the distance between announced and committed. It also corrects the largest error we have made so far.

In brief

• Two separate corrections took our "seeking finance" figure to zero. We cleared USD 11.3 billion of capital figures that no source supported, then separated 321 financing records from the project catalogue.
• 553 GW is on record as not yet operating. Only 98 GW is under construction. The rest is at concept or development, where a project is an intention rather than a commitment.
• Storage is the sharpest example: 89 GW at concept, 10 GW operating. Nine times more announced than built, and 83 GW of the announcements sit in India alone.

The correction

Until this week Solandor reported that 24 projects were seeking a combined USD 18.72 billion. That figure is now zero. Two distinct errors produced it, and they are worth separating because they are different mistakes.

The first was capital figures no source supported. On 2 September we cleared the capital-sought field on 32 records, removing USD 11.3 billion. Each had a number attached that no public document backed as a live requirement. Adani Green Energy is a listed company that self-funds. Dholera is a state solar park allocated by SECI tender, not a project canvassing for capital. Qurayyat has a signed 25 year power purchase agreement, which is the opposite of an open raise. In each case a plausible figure had been attached to a project that was not asking for money.

The second was a whole class of record in the wrong table. On 3 September we separated 321 entries that came from the World Bank projects API, the IATI datastore and the Green Climate Fund open data library. All three publish financing activity: a lender committing money to a programme. We had been reading the committed amount as capital sought by a project, which sounds close enough and is wrong in the way that matters. A disbursement by a development bank is not a project looking for money. It is a project that already found it.

Those 321 records still exist and still have their own pages. They are no longer counted as dealflow, and the catalogue now separates the two kinds of record structurally so the confusion cannot recur.

That second fix also resolved four things we had reported separately over recent weeks: sovereign lenders appearing as developers, commitments read as capital sought, 189 records naming a funder in the developer field, and a COVID working capital facility listed as a clean energy project. All four were one error wearing different clothes.

So Solandor currently lists no project with a live raise. That is a worse headline than the one it replaces, and it is the first one we can stand behind.

553 GW that is not being built

Strip out everything already operating and the catalogue holds 1,516 projects, 553 GW, that are not yet generating.

The instinct is to read that as a pipeline. It is better read as a queue with a very wide mouth:

• Concept: 573 projects, 235 GW
• Development: 620 projects, 219 GW
• Construction: 322 projects, 98 GW

Only the last line is money committed. A project at concept has been announced and may never proceed. A project under construction has sponsors, contractors and a financial close behind it. Presenting the sum as one number invites a conclusion that none of the underlying records support, which is why Solandor keeps the stages apart and never totals them into a headline.

Eighteen per cent of what is not yet built is actually being built.

Storage is where the gap is widest

Across the region the catalogue holds 131 GW of storage. Its distribution is unlike any other technology:

• Concept: 72 projects, 89 GW
• Development: 23 projects, 18 GW
• Construction: 13 projects, 14 GW
• Operating: 15 projects, 10 GW

Nine times more storage has been announced than has been built. And the announcements are concentrated to an unusual degree: 83 of the 89 GW at concept stage is in India, across 61 projects, most of them pumped hydro rather than batteries.

Two readings are available and this issue does not choose between them. Either India is early in a genuine build-out of long-duration storage and the operating figure follows in a decade, or the announcement pipeline is running well ahead of what will be financed. The database cannot tell you which. What it can tell you is that anyone quoting a regional storage figure is, at present, mostly quoting Indian intentions.

Solar is fourth

A smaller finding, offered because it contradicts the received view.

By installed and planned capacity across the Gulf, South Asia and South East Asia:

• Hydro: 1,385 projects, 343 GW
• Wind: 1,911 projects, 306 GW
• Storage: 123 projects, 132 GW
• Solar: 1,098 projects, 90 GW

Solar leads on nothing here except attention. Wind has nearly twice as many projects. Hydro carries more capacity than any other technology, largely because Himalayan and Mekong schemes are old, enormous and long since built.

Two cautions. Capacity and project count answer different questions, and a 5 GW hydro scheme and a 5 MW rooftop array each count as one project. And this is the record as published, not as complete: coverage of distributed solar is weaker than coverage of utility-scale plant everywhere, because small assets are less consistently registered. Solar's true project count is certainly higher than 1,098. Its share of regional capacity is not obviously higher.

What changed on the site

Three things, all of which exist to let you check the work rather than take it on trust.

Every source now publishes its own licence, per source, rather than in a paragraph. Ten of the 23 are openly licensed.

Those ten are downloadable whole, free and without an account: 5,139 of the 5,177 projects, at solandor.co/data. The 38 that are missing are named on the page. Their sources are public but do not license redistribution, and public and redistributable are not the same permission. There is no investor export at all, for the same reason.

And every change to the catalogue is now logged week by week at solandor.co/changes, written at the moment a record enters or leaves rather than reconstructed afterwards from a timestamp.

Method, and what is not verified

Every record names its source and links to it. Nothing in the catalogue is generated by a language model. Where a source does not publish a tariff, a CAPEX or a return, the field stays empty and is reported as missing rather than filled in.

Capacity figures should be read as a floor. Several of the upstream datasets are periodic snapshots rather than live feeds, so a figure is the last published value, not today's.

Contact details are never published. Introductions are requested through us and made with both sides' knowledge.

Information only. Not financial advice. Solandor is not a broker, holds no funds or securities, and charges no fee contingent on any transaction.