Skip to content
Briefings are running a touch slower this week while we rebuild the foundations.See roadmap
Data Centres: Boom and Backlash
10JUN

OpenAI cuts compute target by $800bn

4 min read
10:06UTC

The Wall Street Journal reported on Tuesday 28 April that OpenAI now targets roughly $600bn in compute spend through 2030, down from the $1.4 trillion Stargate nameplate, and has dropped or paused flagship sites in Texas, England and Norway.

IndustryDeveloping
Key takeaway

OpenAI's compute commitment has fallen 57% while hyperscaler capex has risen, leaving supply out of step with its anchor tenant.

OpenAI told investors in February that its compute commitment through 2030 now runs to roughly $600bn, down from the $1.4 trillion nameplate that anchored the Stargate programme, according to a Wall Street Journal report published on Tuesday 28 April 1. Alongside the revised number, the company declined its option on Crusoe's Abilene, Texas flagship lease, formalised the suspension of the Cobalt Park site in North Tyneside that surfaced in early April , reconfigured plans for Narvik in Norway, and lost several senior Stargate executives in mid-April. Tom's Hardware reported that OpenAI now treats Stargate as 'an umbrella term' and prefers leased capacity to owned build.

Stargate was launched as a joint venture with SoftBank and Oracle to anchor a generational AI build-out, with the $1.4 trillion figure cited as the project's expected lifetime cost; the new number is a 57% trim against that nameplate. The WSJ put OpenAI's 2026 cash burn at $25bn against a $30bn revenue target 2, leaving a $5bn margin that constrains any return to the original trajectory regardless of compute supply. The decline on the Crusoe Abilene option releases purpose-built inventory into a secondary market at the same moment Crusoe's first 1.2 GW phase has reached operational status .

The pivot to leased capacity transfers asset risk rather than removing it. Where OpenAI would have held land, transformers and substations on its own balance sheet, those positions now sit with Microsoft, CoreWeave and colocation operators including Equinix. The redirection of demand toward leased colocation is consistent with the suspension pattern: Cobalt Park, Narvik and Abilene were the three sites at which OpenAI was carrying the highest level of first-party development risk. Crusoe's Abilene inventory now enters a market where the price of declined hyperscale leases will set the floor for Q3 negotiations.

Cobalt Park's suspension sharpens the UK read-across more than the Texas or Norway moves. The North Tyneside pause lands against the industrial electricity gap OpenAI itself flagged in March, with UK rates roughly quadruple US and Nordic levels , the same gap the Gate 2 reforms are now trying to close with explicit discounts for AI Growth Zones . The trimmed compute target gives every operator, landlord and lender priced into the original $1.4 trillion a fresh anchor to reprice against.

Deep Analysis

In plain English

OpenAI, the company behind ChatGPT, told investors it now plans to spend roughly $600 billion on computer infrastructure by 2030, down from the $1.4 trillion figure it had announced. That is a cut of about $800 billion. Instead of building its own data centres, it now wants to rent capacity from companies that already run them. Think of it like deciding to rent a warehouse rather than building your own: cheaper upfront, but you pay rent forever. The companies that build and own those warehouses, such as Equinix, are now looking like the people who will benefit most from this change.

Deep Analysis
Root Causes

OpenAI's revenue runs on monthly subscriptions and API token volume, which grow quarter by quarter but cannot be committed five years forward. Hyperscaler and first-party AI campus capex runs on ten-year depreciation schedules and board-level commitments that require long-horizon demand certainty. The gap between the two horizons is the structural driver: OpenAI could only have honoured the $1.4 trillion trajectory if its 2030 revenue were contractually committed today, which it is not.

The Crusoe, Cobalt Park and Narvik retreats share a second structural cause: all three were sites where OpenAI was carrying the highest level of first-party development risk (land, planning, and substation commitments) rather than capacity reservations against existing infrastructure. First-party development risk cannot be unwound in a quarter; leased compute can.

What could happen next?
  • Consequence

    Crusoe's Abilene inventory, and similar declined leases, enters the secondary market, compressing US flagship-site prices through Q3 2026.

    Short term · 0.8
  • Risk

    Colocation operators absorbing OpenAI's redirected demand face margin pressure if OpenAI's $5bn cash headroom forces below-market multi-year agreements.

    Medium term · 0.65
  • Opportunity

    Equinix and Digital Realty are positioned as primary beneficiaries if OpenAI's lease pivot is the leading edge of a broader AI-lab shift away from first-party campus ownership.

    Medium term · 0.7
  • Precedent

    The revision sets the first public benchmark for how a major AI lab can contract out of a headline infrastructure commitment without a formal filing, using investor briefings rather than SEC disclosure.

    Long term · 0.75
First Reported In

Update #3 · OpenAI cuts $800bn; rivals double down

The Next Web· 16 May 2026
Read original
Different Perspectives
Data-centre developers and hyperscale operators
Data-centre developers and hyperscale operators
Hill County's rescission on 4 June, seven days after RCM Hill filed a $100 million taking suit, shows the Fifth Amendment threat is faster than any appellate route: the damages clock runs from the day the moratorium passed. Rural counties with no large-load permitting framework face a litigation bill that can exceed their entire annual budget.
Kenya and President Ruto
Kenya and President Ruto
Kenya's suspension of the $1 billion Microsoft-G42 Olkaria project in early May applies raw-capacity logic at national scale: President Ruto stated the full 1 GW build would mean switching off half the country against a 3 GW installed base. A single hyperscale campus can consume a third of a Sub-Saharan grid with no equivalent constraint in Europe.
Denmark and Energinet
Denmark and Energinet
Energinet's 27 May extension of its large-load connection pause, with a 60 GW queue against 7 GW peak demand, demolishes the assumption that surplus renewable generation is a relief valve for compute demand. Denmark has more wind than it can use and still cannot connect data centres, because transmission pace is the binding constraint.
France and EDF
France and EDF
EDF's repurposing of the Bouchain former power-station site for SoftBank's Phase 1 campus gives France a replicable siting instrument, a brownfield nuclear connection bypass, that no other G7 grid operator can match. France's Choose France summit on 30 May secured the boom's largest European bet without a connection-queue fight or community moratorium.
SoftBank Group
SoftBank Group
SoftBank's EUR 75 billion France commitment on 30 May anchors at EDF's Bouchain nuclear baseload, bypassing the UK's four-times-US electricity cost premium (cited by OpenAI as reason to pause Cobalt Park) and Germany's grid-queue delays. EDF's supply relationship is bilateral; SoftBank never enters the French connection queue.
US residential ratepayers and state regulators
US residential ratepayers and state regulators
Portland General Electric's 4 June tariff is the first evidence that PJM's cost-transfer warning to governors on 19 May can run in reverse: Oregon households get a 1.3% bill reduction as data centres absorb their grid costs. The 12 other states carrying active cost-attribution bills now have a filed tariff with actual numbers to cite.