ReadingResearchRadarInvestment frameworkAI Company Primary Sources
Sign in / Sign up
← Back to reading
NewsTopic · AI数据中心电力交付瓶颈Material3 min readDisconfirmAI Editorial

EQT Global Chair Jean Salata: The Bottleneck for AI Data Centers Is Power Delivery, Not Contract Signing

Jean Salata, global chair of the Nordic private equity giant EQT which manages $400 billion in assets, referred to a large volume of signed but undelivered data center contracts as 'bragawatts' (bragging watts). He explained why EQT chose to hold $100 billion in energy assets and partner exclusively with hyperscale cloud service providers, rather than participating in new cloud credit models.

2026-11-02 20:44 GMT+8

In AI infrastructure investment, there is an easily overlooked distinction: signing a power contract and actually delivering electricity to the server rack are two entirely different things. In an interview with Semafor, Jean Salata, global chair of EQT, used a specific term to label the former—'bragawatts' (bragging watts)—and used it to explain why this Nordic private equity giant, managing $400 billion in assets, has chosen a path distinct from most of its peers during the AI boom.

Salata’s core judgment is that power and energy assets are the true limiting factors in current AI construction. EQT had previously invested in power assets and data centers geared toward cloud migration; at the time, these investments were unrelated to AI. By 2026, however, these assets suddenly became the bottleneck for AI infrastructure—'We suddenly find ourselves holding $100 billion in energy assets, which are precisely the real limiting factor for AI right now.' EQT currently partners only with hyperscale cloud service providers (hyperscalers), with 92% of its data center contract revenue coming from such clients.

This contrasts sharply with the model relied upon by much of the current AI infrastructure financing. Salata points out that many people have signed massive contracts with anyone—they call them bragawatts—but signing something and delivering it on time, with power connected, are two very different things. He believes sophisticated investors understand this distinction, but many in public markets are investing in things they truly do not understand. This concern is well-founded: in the fourth quarter of 2025, investors withdrew more than $7 billion from the largest private credit funds, including those managed by Blackstone, Apollo, Blue Owl, and Ares; Blue Owl was even forced to raise its quarterly redemption cap from 5% to 17% and borrow money to meet payouts.

Sign in / Sign up
ReadingResearchRadarInvestment frameworkAI Company Primary Sources

EQT’s overall stance on private credit is also noteworthy. Salata describes the credit business as a 'beta business,' arguing that its best-case scenario yields only an 8% return and is unrelated to EQT’s strengths in equity operational capabilities. EQT deliberately avoided chasing the path taken by US peers like Apollo and Blackstone, who expanded their asset management scale through private credit and insurance businesses.

However, Salata’s position comes with specific conditions. EQT was co-founded in 1994 by Sweden’s Wallenberg family, which remains the largest shareholder today. Its patient holding model resembles the long-hold structure that many private equity firms are attempting to transition toward. This means EQT can withstand longer asset realization cycles than typical ten-year closed-end funds. Salata himself acknowledges that the entire industry is making the same AI trade, and differentiation ultimately depends on execution rather than directional choice—'There are no patents in finance. We are all moving in the same direction, and it ultimately comes down to execution.'

For readers, this distinction has practical implications: when evaluating AI infrastructure investments, 'signed capacity' and 'delivered power' are two different dimensional metrics. Investors who own built-out power assets and have contracts with hyperscale cloud service providers face a risk structure different from those relying on credit exposure to the future revenue streams of new cloud companies. Salata used the term 'bragawatts' to name this execution gap, which has been obscured by market narratives.

Sources

  • semafor.com↗
  • evidenceinvestor.comEvidence Investor↗
Topic

AI数据中心电力交付瓶颈

【背景】北欧私募巨头EQT全球主席Jean Salata指出,AI数据中心的核心瓶颈已从资本签约转向物理层面的“电力交付”,他将大量已签约但未通电的项目称为“bragawatts”(吹牛瓦)。这一判断正被市场验证:2026年第一季度,美国约…

3 steps1 piecesUpdated 2026-11-03 00:05

How it got here

  1. 2026-09-29MENAFN报道指出,Vertiv以14.5亿美元收购微电网公司,Cipher Digital延长德州数据中心租约,显示企业正通过自建或锁定长期电力资源来应对瓶颈。menafn.com ↗
  2. 2026-10-25AInvest报道,微软因英国电网连接延迟至2030年代中期而取消Loughton数据中心项目;亚马逊因弗吉尼亚州当地反对取消4.27亿美元项目。ainvest.com ↗
  3. 2026-11-02Semafor报道,EQT全球主席Jean Salata提出“bragawatts”概念,强调电力交付而非合同签约是真正瓶颈,并解释EQT持有1000亿美元能源资产的策略。semafor.com ↗

What is still unsettled

  • FERC的加速审批指令能否在短期内实质性缩短电网互联排队时间?
  • 随着各州立法限制增多,未来两年美国数据中心项目的实际落地率是否会低于分析师预估的35%?
  • 自建微电网和现场发电技术能否成为超大规模云服务商解决电力瓶颈的主流方案?
Read the full topic →
Continue researching →