Alphabet, Microsoft and Amazon signal bigger 2027 spend as Naver lands $9bn from Brookfield and Aramco takes 49% of HUMAIN
Alphabet and Amazon said capital spending will rise again in 2027 and Microsoft blamed delayed GPU capacity from third parties for lighter spend, while Naver signed up to $9bn of financing from Brookfield for its Sejong AI factory and Aramco closed a $2.6bn deal for 49% of HUMAIN.
The week in 30 seconds
- North America: Alphabet and Amazon both told investors that capital spending will rise again in 2027, while Microsoft said delayed deliveries from GPU-as-a-service providers held its quarterly spend below forecasts.
- Asia Pacific: Naver signed a financing agreement with Brookfield for up to $9bn, the week's biggest money story, completing funding for a $10bn plan to take its Sejong campus to a 200MW AI factory.
- Middle East & GCC: Aramco completed the purchase of a 49% stake in HUMAIN from PIF, with about SAR 9.9bn ($2.6bn) in cash plus AI, compute and digital assets.
The headline story
The three largest cloud providers reported results on consecutive days, and all three described demand that is still running ahead of supply. Alphabet held its 2026 capital expenditure guidance at $195bn to $205bn after spending $51.36bn in the third quarter, and its finance chief said capex would increase significantly in 2027, with details due on the next call (CNBC). Amazon kept its 2026 forecast at about $220bn and said it expects to spend meaningfully more in 2027, as AWS revenue grew 39% to $45.98bn (CNBC). Andy Jassy put the AWS backlog at $578bn on the call (Motley Fool transcript).
Microsoft spent $45.2bn including finance leases, below analyst expectations, and Amy Hood said some third-party capacity due in the quarter had slipped, pointing to GPU-as-a-service providers; those deliveries have moved into the second half of its fiscal year and she expects spending above $50bn this quarter (CNBC). Its commercial backlog grew 73% to $677bn (Motley Fool transcript).
The message for the supply chain is that the constraint has moved from money to delivery. Microsoft's capex fell short not because it chose to spend less, but because leased capacity from neoclouds did not arrive on time.
Talent signal: capacity planning, third-party supply and data centre delivery leaders at the hyperscalers in Seattle, Mountain View and Northern Virginia, and operations chiefs at neoclouds who can hand over GPU capacity on the dates they have promised.
AI infrastructure watch
The GPU and AI factory moves of the week, and the leadership they need. More on our AI infrastructure page.
Naver moves GPUs off balance sheet with Brookfield
A Brookfield-led vehicle will own the Nvidia GPUs and the power, cooling and network assets at GAK Sejong, while a Naver subsidiary sells the compute as a service; capacity is due to grow from 55MW in the first half of 2027 to 200MW in 2028, with about 100,000 GPUs (Baxtel).
Talent signal: GPU-as-a-service commercial and operations leaders in Seoul, and asset managers who can run GPU fleets as an infrastructure asset class.
FuriosaAI raises $620m for inference chips
The South Korean chip designer's Series D values it at about $2.9bn after the round, with Korea's National Growth Fund putting in $280m; the money will expand production of its RNGD accelerator and fund a third-generation chip (FuriosaAI via Business Wire).
Talent signal: supply chain and go-to-market leaders for non-Nvidia silicon in Seoul and Silicon Valley.
Envision claims the first 800V DC power system at an AI campus
Envision says the system, built on solid-state transformers and battery storage, is running at its Galaxy campus in Ulanqab, Inner Mongolia; it did not disclose the system's size or customer (Baxtel).
Talent signal: electrical engineering leaders with DC power distribution experience, as rack densities push designs beyond AC.
South Korea budgets $15.8bn for AI data centres and chips
President Lee Jae Myung's record 2027 budget allocates KRW 21.3 trillion to physical AI, AI data centres and semiconductors, plus KRW 4.7 trillion for a sovereign AI model (W.Media).
Talent signal: public programme and sovereign AI delivery leaders in Seoul and Sejong.
North America
Google named as the user behind a 1,200MW Goochland campus
Goochland County confirmed on 28 October that Google is considering the campus on about 1,000 acres of West Creek Business Park in Virginia, where data centres are permitted by right; due diligence is continuing (Baxtel).
Talent signal: site development and Dominion Energy interface leaders in the Richmond market.
Prometheus Hyperscale exits its Casper campus
The developer withdrew from the planned 1,250MW campus near Casper, Wyoming before filing a permit application; the landowner says a new, unnamed developer will take the project forward (Baxtel).
Talent signal: land, zoning and community engagement leaders who can carry gigawatt sites through local approval.
FirstEnergy's data centre pipeline reaches 26GW
The utility has 7.9GW of contracted large-load demand and is negotiating about 4GW more, and it expects data centres to lift its transmission plan by at least $4.75bn (Baxtel).
Talent signal: utility and large-load contracting leaders across Ohio, Pennsylvania and West Virginia.
EdgeCore completes a $521m green ABS
The securitisation is backed by the 36MW SV01 building in Santa Clara, leased to a hyperscale tenant, and will refinance construction loans and fund new development (Blackridge Research).
Talent signal: capital markets and treasury leaders who can repeat securitisations across a stabilised portfolio.
Europe
UK Budget sets out a Data Centre Covenant
HM Treasury's covenant asks strategic data centres to agree local benefit plans, design for a PUE of 1.3 or better and use closed-loop water, and in return offers priority grid connections through a new Ministerial Control Centre (GOV.UK).
Talent signal: UK policy, planning and community investment leaders who can turn covenant tests into faster connections.
TPG reportedly in talks to buy Ark Data Centres
The Financial Times reported that TPG is negotiating to buy the UK operator from Elliott at about $5bn; Ark runs 27 data centres in the UK and Belgium (IREI).
Talent signal: CEO, CFO and integration leaders for a UK platform likely to change hands.
Telehouse and KDDI close the East India Dock purchase
The more than £220m deal for the London Docklands campus comes with consent for a 376,000 sq ft data centre, and Telehouse says 65 MVA of power has been secured (Baxtel).
Talent signal: London construction and high-density design leaders.
Middle East & GCC
Aramco completes its 49% stake in HUMAIN
PIF keeps 51% of the AI company, which targets 3GW of data centre capacity by 2030; Aramco's contribution includes about SAR 9.9bn ($2.6bn) in cash during 2026 (W.Media).
Talent signal: data centre delivery and integration leaders in Riyadh and Dhahran as two state champions combine teams.
EQT seeks joint AI and data centre deals with Gulf sovereign investors
EQT's chair told The National the firm is in talks with UAE and Saudi investors and plans to grow in the region through its portfolio company EdgeConneX (The National).
Talent signal: Gulf country managers and development heads for European platforms entering the region.
Oracle keeps new Gulf data centre locations private
An Oracle executive said the company has added new centres in the UAE and Saudi Arabia and plans more, but will not disclose locations because of the regional security situation (The National).
Talent signal: physical security and resilience leaders for Gulf cloud regions.
Adnoc adopts OpenAI's Frontier platform through G42
The agreement, one of four signed between Abu Dhabi entities and OpenAI, will run on inference infrastructure based in the UAE (The National).
Talent signal: enterprise AI deployment and inference operations leaders in Abu Dhabi.
Asia Pacific
Hyundai Motor Group plans a Saemangeum AI data centre
Group affiliates approved about 5.8 trillion won for a data centre with more than 50,000 GPUs in Gunsan, with construction targeted from March 2027 and services from early 2030 (Baxtel).
Talent signal: project directors and GPU procurement leaders for a new Korean AI platform.
STT GDC reportedly seeks a S$3.1bn loan
Bloomberg reported that the operator, now owned by KKR and Singtel, is in talks with four banks for a three-year loan (W.Media).
Talent signal: group treasury and debt capital markets leaders in Singapore.
Macquarie agrees to buy control of Gabia and KINX
MAIF4 will buy the major shareholders' stakes in the Korean IT infrastructure group and its data centre subsidiary, then offer the same prices to minority holders, subject to approvals (Macquarie).
Talent signal: portfolio operating partners and Seoul-based leadership for infrastructure fund ownership.
L&T's Vyoma.AI breaks ground on 52MW in Chennai
The direct-to-chip liquid-cooled building expands an existing 30MW facility at Kancheepuram, with Together AI as the announced AI factory customer (Baxtel).
Talent signal: liquid cooling commissioning and GPU operations leaders in Tamil Nadu.
New money
Funding rounds, financings, IPOs, M&A and investment pledges announced this week.
| Who | What | Amount | Region | Source |
|---|---|---|---|---|
| Naver and Brookfield | Financing for GAK Sejong AI factory | Up to $9bn | Asia Pacific | Baxtel |
| TPG | Acquisition of Ark Data Centres (reported, in talks) | About $5bn | Europe | IREI |
| Aramco | 49% stake in HUMAIN, cash element plus assets | SAR 9.9bn ($2.6bn) | Middle East & GCC | W.Media |
| STT GDC | Syndicated loan (reported, in talks) | S$3.1bn (US$2.4bn) | Asia Pacific | W.Media |
| FuriosaAI | Series D | $620m | AI infrastructure | FuriosaAI via Business Wire |
| EdgeCore | Green ABS backed by SV01, Santa Clara | $521m | North America | Blackridge Research |
| Telehouse and KDDI | East India Dock campus purchase, London | Over £220m | Europe | Baxtel |
New capacity
| Project | Region | Size | Source |
|---|---|---|---|
| Google, West Creek, Goochland County, Virginia (under consideration) | North America | 1,200MW | Baxtel |
| Naver, GAK Sejong AI factory, South Korea | Asia Pacific | 200MW | Baxtel |
| SAIGONTEL consortium, Tay Ninh, Vietnam (proposed) | Asia Pacific | 200MW | W.Media |
| Hyundai Motor Group, Saemangeum, Gunsan | Asia Pacific | 100MW | Baxtel |
| Vyoma.AI (L&T), Kancheepuram, Chennai | Asia Pacific | 52MW | Baxtel |
| Telehouse, East India Dock, London | Europe | 65 MVA secured | Baxtel |
Leadership moves
- Jason Burnett, Chief Operations Officer at AIB Data Centers, reporting to CEO Jerry Tang and overseeing construction, procurement and operations; he was previously COO of Hivelocity (AIB). A sign that developers converting power into contracted AI capacity now need delivery operators as much as dealmakers.
The T4X view
The cheques are no longer the bottleneck: Alphabet and Amazon will spend more next year, a Brookfield-led vehicle will own Naver's GPUs and Aramco is putting cash into HUMAIN. What slipped this quarter was delivery, as Microsoft's leased GPU capacity arrived late. The leaders in shortest supply are those who hand over powered, working capacity on the promised date, and boards should hire for that record before the next funding round.