04/13/2025 What I Read Last Week
#DigitalInfrastructure news: Google Cloud, data center energy consumption, AI Infrastructure
Weekly Edition of curated news about Digital Infrastructure
[Link] Google Cloud Next 25. There was an incredible amount of news from this Vegas event… and more AI advancements than I have time to write about, let alone research and test.
GC CEO Thomas Kurian kicked off the event with a speech, where he said, “We expanded Google Cloudto 42 regions, including Sweden, Mexico and South Africa, and are rapidly expanding to countries like Malaysia, Thailand and Kuwait. We expanded our 2 million mile terrestrial and subsea fiber network by announcing new subsea cables like Umoja, Bosun and Proa.”
[Link] Collaboration with Tapestry (Alphabet incubated moonshot) and PJM to develop AI-driven data capabilities for a smarter, more reliable electricity system. Tapestry, using Google Cloud and DeepMind, will optimize power generation for the PJM electric grid, making electricity more reliable and affordable. This initiative aims to handle increasing electricity demand, streamline interconnection, and integrate various energy sources into the grid.
[Link] Google introduced its seventh-generation Tensor Processing Unit (TPU), Ironwood. The Ironwood computing system can connect up to 9,216 liquid-cooled chips linked with Inter-Chip Interconnect (ICI) networking spanning nearly 10 MW. Ironwood is a key part of Google Cloud AI Hypercomputer, a system that tightly integrates hardware and software for demanding AI tasks. For Google Cloud customers, Ironwood comes in two sizes based on AI workload demands: a 256 chip configuration and a 9,216 chip configuration.
[Link] Google's new Cloud WAN solution is a managed enterprise backbone that leverages Google's global network. It provides high-performance connectivity, cost savings, and enhanced security for geographically dispersed locations. Compared to the public internet, Cloud WAN offers up to 40% faster performance.
[Link] International Energy Agency report: Data center energy consumption set to double by 2030 to 945TWh. The report projects that renewable energy generation will grow by more than 450TWh to meet data center demand by 2035.
[Link] Opinion: The AI infrastructure race hits a political reality check. AI infrastructure development is facing political and regulatory hurdles due to increased demand for resources like land, electricity, and water. Communities, grid operators, and environmental advocates are raising concerns, while states re-evaluate the economic benefits of data centers.
[Link] Start Campus celebrated the official inauguration of SIN01, its first operational facility within the company’s 1.2-gigawatt (GW) SINES Data Campus in Portugal.
[Link] Rowan Digital Infrastructure, a sustainable hyperscale data center developer and portfolio company of Quinbrook Infrastructure Partners, announced it closed a further $1.24 billion of construction financing to build the second phase of a flagship hyperscale data center campus in Maryland.
[Link] MARA and NGON have fully energized a 25-megawatt (MW) data center initiative, converting previously flared natural gas into low-cost electricity. In its first five months of operations, the project reduced 29,300 metric tons of CO₂ equivalent emissions, comparable to removing 6,800 gasoline-powered vehicles from the road for a year.
[Link] Galaxy Digital is converting its West Texas cryptomine data center to host AI cloud provider CoreWeave, signing a 15-year lease agreement to deliver 133MW of Critical IT Load (200MW total).
[Link] To address the strain on the United States' electric grid caused by increasing electricity demand, President Trump created an executive order to focus on ensuring the grid's reliability, resilience, and security. It directs the Secretary of Energy to streamline processes for issuing orders under section 202(c) of the Federal Power Act, develop a uniform methodology for analyzing reserve margins, and establish a process to identify and retain critical generation resources.
[Link] NEye Systems, a startup developing a new kind of networking chip for artificial intelligence data centers, has raised $58 million in venture financing in a round led by CapitalG, a growth-stage fund backed by Alphabet. California-based nEye is developing a chip that taps optical technology to send information between AI chips in the form of light rather than electrical signals, a field that chip giants such as Nvidia and startups alike are chasing because it could lower energy use for AI data centers.
Follow me on X for more links and news.
Coinbase is the easiest (and best) place to buy and sell cryptocurrency. Use my referral link to sign up.