Humain's $2.5bn outside fund buys about a twentieth of the capacity it promised for 2030

The per-capita figure explains the roadmap better than the roadmap does. Alvarez & Marsal put Saudi data centre capacity at 12 watts per head against roughly 50 in the UAE and the United States ^[16], and 467 MW installed at 12 watts implies a denominator near 38.9 million people ^[1]; 38.9 million at 50 watts is 1.945 GW ^[2], which is within rounding of the 1.9 GW Humain says it will have by 2030 ^[13]. The 2030 plan, then, is parity with where the UAE already is, and the 6 GW promised for 2034 works out at about 154 watts per head ^[3], call it three times parity.

Getting there is a financing problem before it is an engineering one. The kingdom has put SR56.2bn, which official figures translate as $14.98bn, into data centres and digital infrastructure to reach 467 MW from 68 MW in 2021 ^[15], a blended $32m per MW ^[4]; the 1,433 MW still missing at 2030 ^[5] costs roughly $46bn at that rate ^[6], and the $2.5bn fund Bloomberg reported Humain assembling from global and local investors ^[10] is 5.4% of it ^[7]. The blended rate includes digital infrastructure that is not data halls, so treat it as an order of magnitude rather than a quote. The pace is the less forgiving number: 399 MW added in about five years is some 80 MW a year, while 1,433 MW by 2030 needs about 358 MW a year, four and a half times the delivered rate ^[8][9]. Humain has told Bloomberg that timeline holds despite the U.S.-Iran war ^[14].

Kurt Davis Jr of Alvarez & Marsal calls the listing a funding-discipline decision as much as a monetisation one, at a point when PIF has been explicit that portfolio companies must reduce their draw on the sovereign ^[5]. Against a fund that PIF puts at $900bn ^[4], $2.5bn is 0.28% ^[10], so the instruction is directional rather than material. And the same company is scaling a $10bn venture fund, possibly larger, by the end of 2026 ^[11], writing cheques only where the recipient commits to Saudi compute or Saudi headcount, because as Tareq Amin put it, "we don't do passive investments" ^[12]. That is capital pointed outward to buy occupancy for halls that do not exist yet, which pairs neatly with the A&M observation that hyperscalers entering new markets increasingly lease instead of build ^[17].

Mohammed Soliman of McLarty Associates reads the same fundraising through export controls: both G42 and Humain want to stay inside the American AI ecosystem, and outside capital, possibly an American majority, is how they keep the access, with Saudi Arabia lacking the A:5 designation and Humain preparing for a G42-style arrangement with Washington ^[9]. G42 disclosed last week that it has held preliminary talks with US companies over selling a majority stake ^[7], and its licence-free window for advanced chips runs to 2027 unless it changes structure ^[8]. On that reading, who writes the cheque matters more than what the cheque is priced at.

Both readings can hold, and the more interesting version is that the discipline directive sets the deadline while chip eligibility sets the structure, since an IPO in 2029 pays for nothing in 2027 and arrives, as Davis notes, mid-construction with most of the capex still ahead ^[6]. Two outcomes would break the discipline thesis. PIF writing another large primary cheque into Humain during the 2026-2028 build would make the sovereign-draw language rhetoric. A fast close of the $2.5bn with American strategic investors rather than the local and global institutions described would make it a structural manoeuvre wearing a financing label.