👀In Case You Missed It
OpenAI signs a record $38B cloud deal with AWS — securing hundreds of thousands of Nvidia GB200/GB300 GPUs in the biggest cloud contract ever, reshaping hyperscale alliances and compute control.
Microsoft, Amazon, and Google post record AI infrastructure spend — with Alphabet’s first $100B quarter, Amazon’s fastest cloud growth in years, and Microsoft revenue up 18%, driving a collective $400B AI build-out sprint.
Google’s Cisco-powered fibre expansion cuts GPU cluster latency — new networking architecture optimised for multi-GPU scaling and lower interconnect lag across next-gen AI systems.
Space computing era begins — SpaceX confirms plans for orbital data centres as Starcloud-1 becomes the first satellite running an Nvidia H100 in orbit, marking the start of spaceborne accelerated compute.
Read more here
🧠 Cloud & Infrastructure Watch
Azure suffered a European region outage tied to a data-centre “thermal event”, with a brief AWS blip stateside, underscoring power and cooling fragility at scale.
Read MoreIBM begins thousands of layoffs while pivoting toward AI software and consulting, signalling a deeper enterprise shift in where value is accruing.
Read MorePower planning becomes front-page: a 1.8GW “Rio AI City” grid study advances, while Australia’s solar boom will mandate daily free-power windows from 2026.
Read MoreGoogle’s Project Suncatcher teams with Planet Labs to put TPUs in orbit, envisioning 1-km arrays of 81-satellite compute clusters as space data centres gather pace.
Read MoreMicrosoft outlines plans for a Johor, Malaysia cloud region (following its first region launch in-country earlier this year), extending SEA capacity.
Read MoreAmazon reportedly spends $700m on a Prince William County, VA site (Devlin Tech Park in Bristow) for future data-centre build-out.
Read MoreNebius lights up an Nvidia GPU cluster in London, adding UK capacity for AI training/inference.
Read MoreZoom plans to lease UK data-centre space to serve regulated sectors, signalling more compliance-grade capacity demand.
Read MoreExxonMobil is in talks to supply data centres with natural-gas power + carbon capture, as operators hunt lower-carbon baseloads.
Read MoreFermi wins preliminary approval for 6GW of natural-gas generation at an up to 11GW campus in Amarillo, Texas to feed AI loads.
Read MoreBT remains on track for 25m-premises fibre and signs a satellite connectivity deal with Starlink, bolstering long-haul/backhaul options.
Read More
💻 Hardware & Semiconductor Moves
Google unveils Ironwood TPUs and Axion CPUs; Anthropic lines up access to as many as one million chips, a marquee custom-silicon bet beyond GPUs.
Read MoreQuantum-adjacent stack expands as Quantinuum launches Helios for general-purpose hybrid quantum workloads.
Read MoreConnectivity backbone strengthens: BT nears its 25m-premises fibre goal; Starlink adds spectrum and airline partners as it crosses 8m customers.
AMD posts record data-centre revenue for Q3 ’25, underscoring unabated accelerator/CPU demand.
Read MoreDaikin acquires Chilldyne, expanding liquid-cooling hardware from cold plates to full system integration for high-density racks.
Read More
📰 News of the Day
Google’s Project Suncatcher: Orbital TPUs Take AI Compute to Space
Summary:
Google has unveiled Project Suncatcher, a partnership with Planet Labs to launch TPU-powered satellites into orbit, creating kilometre-scale arrays of 81 interconnected satellite compute clusters. The initiative aims to process AI workloads directly in space, marking the first move toward orbital data centres optimised for low-latency Earth observation and AI-driven analytics. It represents Google’s boldest infrastructure experiment yet, extending its AI compute footprint beyond terrestrial limits.
Why It Matters:
Project Suncatcher could rewrite the economics and physics of AI compute. By relocating processing above the atmosphere, Google bypasses constraints around land, water, and grid power while reducing latency for satellite data processing. It also strengthens Google’s vertical integration — controlling not just chips (TPUs) and data centres, but now orbital compute nodes. For hyperscalers, it opens a new race to space-based infrastructure sovereignty.
The Bigger Picture:
This marks the beginning of a post-earth compute era, as hyperscalers, space firms, and energy companies converge to sustain AI’s exponential demand. SpaceX is building orbital data centres, Nvidia-powered satellites are already in orbit, and now Google is exporting TPUs to space. The line between cloud regions and orbital regions is blurring — and whoever masters off-planet compute could define the next decade of AI power, connectivity, and data control.
