03/23/2025 What I Read Last Week
#DigitalInfrastructure news: NVIDIA, AI reference architectures, Gigawatt-scale data centers, quantum computing, Edge AI
Weekly Edition of curated news about Digital Infrastructure
NVIDIA (GTC 2025 event) or NVIDIA-adjacent news filled my news feed this past week. I’ll put an AI-generated summary at the end of this - because it was just too much to keep up with. Here are some of my highlights:
[Link] SemiAnalysis technical review: NVIDIA GTC 2025 – Built For Reasoning, Vera Rubin, Kyber, CPO, Dynamo Inference, Jensen Math, Feynman.
NVIDIA AI reference architectures released - from Hewlett Packard Enterprise, Cisco, Dell, and Supermicro.
[Link] ECOBLOX announced its selection as a strategic partner for NVIDIA's Mobile AI Factories and has launched new products to support them. NVIDIA AI Factories are data centers using NVIDIA tech for AI model creation, training, and inference, aiming to boost productivity, efficiency, and agility in manufacturing.
[Link] EPRI, NVIDIA and collaborators launch Open Power AI Consortium to transform the future of energy. During the GTC event EPRI launched the new Open Power AI Consortium, to drive the development and deployment of an open AI model tailored for the power sector, accelerating AI adoption to reduce operating costs while improving the energy customer experience.
[Link] Nvidia is working with several telecommunications and research companies to develop an artificial-intelligence-native wireless network for 6G. Jensen Huang said Tuesday the chipmaker plans to work with T-Mobile, Mitre, Cisco, ODC and Booz Allen Hamilton to make a wireless network that is integrated with AI.
[Link] Giga Computing and Start Campus are collaborating on a technical study to assess the integration of Giga Computing’s GIGAPOD platform and GIGABYTE POD Manager into AI-ready data center infrastructures like SINES DC. The study will focus on optimizing energy efficiency, AI-driven operations, and system resilience.
Check out the Jensen Huang GTC keynote - around 45:56 where he talks about Blackwell and data center infrastructure (scale-up, then scale-out).
[Link] Crusoe announced construction has begun on the next phase of its AI data center at the Lancium Clean Campus in Abilene, Texas. The second phase of construction, expected to be completed in mid-2026, includes six additional buildings, bringing the total facility to eight buildings, approximately 4 million square feet, and a total power capacity of 1.2 gigawatts.
[Link] STACK Infrastructure announced it has secured $4 billion in green financing for its 1+GW campus in Stafford, Virginia, as well as two campuses in Portland, Oregon, and Toronto, Canada.
[Link] A data center campus of up to 5GW that utilizes its own behind-the-meter power supply is being pitched for 50,000 acres of land in Texas. Energy Abundance Development Corporation has unveiled the project, Data City, which could be built near the city of Laredo, with the first 300MW and 1 million square feet of space set to launch in 2026.
[Link] Crypto and AI data center firm Core Scientific plans new AI data centers in Dalton, Georgia. Core Scientific’s Dalton footprint totals around 195MW.
[Link] Bain Capital-backed hyperscaler Bridge Data Centres, has secured US$2.8 billion in senior secured bank financing to fuel its data center expansion, marking one of the largest-ever bank facilities for pan-Asian data centre operators.
Data centers: big boom, big scale, big money:
[Link] Picture this: the data center boom in charts. The number of hyperscale data centers has nearly doubled since the end of 2019, reaching 1,136 at the end of 2024, and hundreds more are in the pipeline.
[Link] From Billions to Trillions: Data Centers' New Scale of Investment. Fueled by AI advancements, data centers are experiencing a new era of investment, with companies like Apple, the Stargate project, Amazon, and others committing trillions of dollars to expand infrastructure in the U.S. and globally.
[Link] Superintelligence Strategy paper. Former Google CEO Eric Schmidt and co-authors suggest building AI training data centers in remote locations to protect them from nation-state attacks, proposing measures like increased transparency, AI-assisted inspections, and strict control over AI chip sales to prevent AI dominance and potential misuse.
Quantum Computing
[Link] NVIDIA Accelerated Quantum Research Center to bring quantum computing closer. Quantum computing innovators like Quantinuum, QuEra and Quantum Machines, along with academic partners from the Harvard Quantum Initiative and the Engineering Quantum Systems group at the MIT Center for Quantum Engineering, will work on projects with NVIDIA at the center to explore how AI supercomputing can accelerate the path toward quantum computing.
[Link] The Quantum Economic Development Consortium released the State of the Global Quantum Industry report, which seeks to capture key metrics that characterize the size and impact of the global quantum industry. The report provides a data-driven perspective, through an analytical framework, on the industry’s composition, investment, market size, workforce & pipeline, and intellectual property.
[Link] Blockchain is a side interest of mine, so this was particularly interesting. D-Wave announced that it has published a new research paper introducing a novel blockchain architecture that uses techniques from its quantum supremacy demonstration. By adding quantum to traditional blockchain computation, the new architecture could enhance blockchain security and efficiency. As part of this research, D-Wave scientists deployed the blockchain architecture across four of its cloud-based annealing quantum computers in Canada and the United States, performing distributed quantum computing for the first time.
[Link] Colt announced the successful completion of a quantum-secured encryption trial across its optical wave network. Colt collaborated with a number of technology partners for the trial including Adtran, Ciena, ID Quantique, Nokia, and Toshiba, with a view to Colt offering a suite of services for global businesses to prepare for a quantum-secured future.
[Link] Ciena survey: Data Center experts predict at least 6X increase in DCI(Data Center Interconnect) bandwidth demand over next 5 years, with 43% of new data center facilities expected to be dedicated to AI workloads.
[Link] Lanner Electronics and Arrcus are collaborating to provide AI-optimized networking solutions at the telco edge by integrating Lanner’s MGX Edge AI platform with Arrcus’ ACE networking platform. This collaboration aims to enable telcos and enterprises to deploy AI inferencing and AI-driven RAN applications with low latency, high scalability, and optimized power efficiency.
[Link] T-Mobile is selling its 800 MHz spectrum, inherited from Sprint, to Grain Management, a private equity firm planning to use it for the utilities industry. Grain invests in global broadband technology and telecommunications infrastructure - and has a pretty interesting portfolio of companies.
Weekly link of 🤯
[Link] Scientists from the RIKEN Center for Emergent Matter Science (CEMS) and collaborators have discovered a groundbreaking way to control superconductivity—an essential phenomenon for developing more energy-efficient technologies and quantum computing—by simply twisting atomically thin layers within a layered device. By adjusting the twist angle, they were able to finely tune the “superconducting gap,” which plays a key role in the behavior of these materials.
Here is a summary of GTC 2025 news - courtesy of Perplexity.ai
Here are the key takeaways from Nvidia's GTC 2025:AI Acceleration and Reasoning Token Explosion: Nvidia is focusing on massive 35x improvements in inference cost to enable the training and deployment of more complex AI models. This is driven by stacked scaling laws in pre-training, post-training, and inference time.
New GPU Architecture - Vera Rubin: Nvidia introduced the next-generation Rubin Ultra GPU, which will offer significant performance gains. The Vera Rubin NVL 144 system, based on Rubin Ultra, is set to arrive in the second half of 2026.
Kyber Rack Architecture: A new rack design that rotates racks 90 degrees for increased density, comprising 4 canisters with two layers of 18 compute cartridges each. This new architecture increases the NVLink world size to 576 GPU dies.
Nvidia Dynamo: An open AI engine stack focused on making it easier to deploy and scale inference. It offers features like Smart Router, GPU Planner, and improved NCCL Collective for Inference.
Isaac GR00T N1: Introduced as the world's first open and fully customizable AI foundation model for building humanoid robots.
Photonics Integration: Nvidia is integrating photonics into its accelerated computing infrastructure, enabling AI factories to connect millions of GPUs across sites while reducing energy consumption and operational costs.
DGX AI Supercomputers: New systems leveraging the Blackwell AI platform, including DGX Station and DGX Spark, allowing a wide range of users to prototype, optimize, and execute AI models from their personal desktops.
Increased Computing Power Demand: Jensen Huang stated that the computing power needed for new AI is "easily 100 times more than we thought we needed this time last year".
Follow me on X for more links and news.
Coinbase is the easiest place to buy and sell cryptocurrency. Use my referral link to signup.