LEAP's headline number has stopped growing, and its composition has changed

LEAP 2026 closed with organisers valuing the week's announcements, investments and agreements at nearly $15 billion. LEAP 2025 closed at more than $14.9 billion. LEAP 2024 closed at $13.4 billion. Set the three side by side and the first observation is the one nobody made during the week: the headline has stopped growing. That is not a criticism of the event, and it is not necessarily a signal of slowing investment. It is a reason to stop treating the annual total as the story. A number that rose 11 percent and then went flat is telling you something about how the number is constructed, not only about how much money exists. The construction is the problem. The closing figure combines at least four things that behave in entirely different ways. There is planned multi-year capital expenditure, of which the clearest example is Amazon Web Services putting more than $5.3 billion behind a Saudi cloud region opening in December. There are capacity commitments denominated in megawatts rather than money, like the AMD, Cisco and HUMAIN plan to add up to 250 megawatts from 2027 on the way to as much as a gigawatt by 2030. There are target fund sizes that have not been raised, such as the SAR675 million Axelerated Solutions and BIM Capital vehicle, which still needs Capital Market Authority clearance and is targeting a first close in the fourth quarter. And there is primary equity that actually moved into companies. Only the last of those is money that changed hands during the week. On the disclosed day-three figures, that was about $126.15 million across nine rounds. Against a headline of nearly $15 billion, primary startup capital is under one percent. The daily arithmetic does not reconcile either. Organisers put day two at more than $2.5 billion and day three at $857 million, while several outlets described day one alone as carrying more than $15 billion. Those cannot all be added together and still land at the closing figure. The most defensible reading is that the closing number is a cumulative event total dominated by two day-one announcements, and that the daily releases were counting on a different basis. Organisers did not publish a category breakdown, so the split across AI infrastructure, data centres, cloud, skills and startups that the closing statement names cannot be reconstructed from published figures. What can be reconstructed is the concentration, and it is severe. Take the disclosed infrastructure items alone. AWS at more than $5.3 billion. Al Moammar Information Systems at $1.2 billion to reach 192 megawatts of operational capacity. NHC Innovation at $880 million for data centres in the Khuzam Digital Valley, targeting 65 megawatts by 2033. Those three total roughly $7.4 billion, or about half the headline, from three announcements. Add the AMD, Cisco and HUMAIN partnership and the HUMAIN and DataVolt work at Oxagon, neither of which carries a published dollar figure, and the great majority of the week's value sits with fewer than ten counterparties, one of which, HUMAIN, appears in most of them. This is the composition change worth arguing about. In earlier editions the total was assembled from a wider spread of announcements. In 2026 it is essentially a small number of very large infrastructure commitments, with a long tail of product launches, memoranda and partnerships that add credibility but very little value to the total. A flat headline with more concentration underneath it is a different market from a flat headline with the same spread. There is a second change, harder to see in the money and visible in the units. Increasingly the significant commitments are not denominated in dollars at all. They are denominated in megawatts and in dates. Up to 250 megawatts starting in 2027. As much as a gigawatt by 2030. One hundred megawatts of a 360 megawatt phase available in 2028. Fifty megawatts in an AI Zone by 2028. A company can raise the dollar figure attached to a project by revising its own capital plan. It cannot revise a grid connection date. The megawatt numbers are the harder currency, and they are the ones a reader should trust more. The investor attendance figures point the same way. Ahead of the event, organisers said more than 1,000 investors from over 900 firms managing $14.3 trillion were coming. The closing statement put it at 1,397 investors from 1,027 firms across 72 countries representing $18.3 trillion. Both are real measures of convening power. Neither is a measure of deployment, and the gap between $18.3 trillion of assets in the room and $126 million of primary equity announced is the honest summary of what a trade show does. It is a place where relationships start, not where cheques clear. None of this makes the number dishonest. It makes it a category error to compare it with a venture funding total, or to read a flat year as a slowdown. What would make the figure genuinely useful is a published breakdown separating capital deployed, contracted capital expenditure, capaci