09/28/2025 What I Read Last Week
#DigitalInfrastructure news: Gigawatt-scale, Open AI Stargate, CoreWeave, TeraWulf, microfluidics cooling system, and Quantum Computing
Weekly Edition of curated news about Digital Infrastructure
These stories absolutely exploded across the tech news cycle last week, so rather than rehash what everyone’s already covered, here are the headlines and links:
[Link] OpenAI signs LOI to deploy ‘at least’ 10GW of AI data centers with Nvidia hardware, Nvidia to invest ‘up to’ $100bn in OpenAI
[Link] OpenAI, Oracle, and SoftBank expand Stargate with five new AI data center sites. New data centers put Stargate ahead of schedule to secure full $500 billion, 10-gigawatt commitment by the end of 2025.
[Link] CoreWeave announced an expanded agreement with OpenAI to power the training of its most advanced next-generation models, reinforcing its position as the essential cloud platform for the most demanding AI workloads. The contract value of this deal is up to $6.5 billion.
[Link] STACK Infrastructure Delivers Digital Infrastructure for the planned Stargate campus in Doña Ana County, New Mexico.
YouTube
[Link] Amazing video from the Bg2 Pod - interviewing Jensen Huang and talking about OpenAI, Future of Compute, and the American Dream. Jensen (expertly) defines what the future holds for GPUs, data centers, and energy. I am currently watching it for a second time (and taking more notes). This is a follow-up from an interview with Jensen from 11 months ago.
[Link] Anthony Pompliano: Why Bitcoin & AI Will EXPLODE Higher Very Soon with Jordi Visser. Fascinating conversation with Jordi Visser diving into AI replacing jobs, the AI-bitcoin mining crossover, and what the NVIDIA/OpenAI deal means. Additionally, he discusses the Bg2 Pod conversation with Jensen Huang.
[Link] Daniel Newman, Futurum Group CEO, and Paul Prager, TeraWulf CEO, join the CNBC show ‘The Exchange’ to discuss how big the power generation figures are for NVIDIA and OpenAI, and if they can reach them. TeraWulf plans to raise $3 billion in debt to support data center buildouts for Google and Fluidstack.
[Link] Cipher Mining signs 168 MW, 10-Year AI hosting agreement with Fluidstack. The agreement represents approximately $3 billion in contracted revenue over the initial 10-year term, and Google will backstop $1.4 billion of Fluidstack’s lease obligations.
[Link] DataBank raises $1.1 billion in hyperscale asset securitization. The proceeds will be used to refinance previous loans put in place to construct DataBank’s ATL4, IAD3, and LGA3 facilities and to fund development of additional data center capacity across the its U.S. platform.
[Link] Nscale announced it has raised the largest Series B in UK and European history, at USD $1.1 billion. The new funding will be used to further Nscale’s deployment of large-scale AI infrastructure across Europe, North America, and the Middle East, enabling the rapid rollout of the company’s “AI factory” data centres for projects like Stargate UK and Stargate Norway, and the expansion of its vertically integrated AI cloud platform.
[Link] Modular has raised $250M in its third financing round to continue its mission to build AI’s unified compute layer – a hypervisor for AI. This brings its total capital raised to $380 million across three rounds since its founding in 2022, and values Modular at $1.6 billion – almost tripling its valuation from its last raise.
[Link] Huawei has unveiled its next-generation data-center scale AI solution that can offer 1 FP4 ZettaFLOPS performance for AI inference and 524 FP8 ExaFLOPS for AI training at its recent Huawei Connect 2025 conference. The Atlas 950 SuperCluster will be built upon 524,288 Ascend 950DT AI accelerators distributed across over 10,240 optically interconnected cabinets.
[Link] Microsoft has developed a microfluidics cooling system for AI chips that etches tiny channels directly into the silicon, allowing liquid coolant to remove heat up to three times more efficiently than traditional cold plates. This breakthrough, which uses AI to optimize coolant flow and mimics nature-inspired designs, enhances chip performance, reduces energy costs, and supports sustainable, high-density data center designs for future AI workloads.
[Link] Lumen is moving forward with a multi-billion-dollar build with plans to add 34 million new intercity fiber miles by the end of 2028, for a total of 47 million intercity fiber miles. A little over a month ago Corning and Lumen reached a supply agreement that reserves 10% of Corning’s global fiber capacity for each of the next two years to interconnect AI-enabled data centers.
[Link] Ciena has agreed to acquire Nubis Communications, a privately-held company headquartered in New Providence, New Jersey. The addition of Nubis, which specializes in high-performance, ultra-compact, low-power optical and electrical interconnects tailored to support AI workloads, will expand Ciena’s portfolio and add critical talent to address a wider range of opportunities inside the data center.
[Link] Equinix announced its Distributed AI infrastructure at its inaugural AI Summit, designed to support next-generation AI, including agentic AI through an AI-ready backbone connecting 270+ data centers across 77 markets, along with a global AI Solutions Lab and Fabric Intelligence capabilities. Equinix and Zayo unveiled an AI Infrastructure Blueprint - a joint architecture framework for powering next-generation AI workloads.
[Link] UAE-based SpaceTech company Space42 is partnering with Microsoft to create the UAE’s first Sovereign Mobility Cloud, which will use Core42‘s Sovereign Public Cloud powered by Microsoft Azure to provide secure, locally-hosted infrastructure for autonomous mobility solutions and related data. The initiative aims to combine Space42’s mobility expertise with Microsoft’s cloud technology to accelerate the deployment of autonomous systems like Space42’s TXAI autonomous taxi service, while ensuring data remains securely within the UAE and meets regional sovereignty requirements. 💾 save for later research / deep dive
Sovereign AI Strategies
[Link] NVIDIA joined Canadian government and industry leaders at the “All In Canada AI Ecosystem” event in Montreal to advance the country’s sovereign AI strategy. The event featured announcements, including TELUS launching Canada’s first fully sovereign AI factory in Quebec, powered by NVIDIA technology, which will serve clients while keeping all data within Canadian borders.
[Link] Cisco has launched a Sovereign Critical Infrastructure portfolio for European customers operating critical systems like banks and healthcare. The key feature is that once deployed, customers have complete control over their infrastructure in air-gapped environments—Cisco has no access to their data or systems. This addresses European digital sovereignty concerns and enables organizations to develop AI models using sensitive data without third-party exposure.
[Link] Google Fellow (and industry icon) Urs Hölzle posted on LinkedIn: “27 years ago Google signed its first data center contract....for 28 sqft (2.5 sqm) of space and 2 Mbps of bandwidth”
Quantum Computing
[Link] IonQ announced it has achieved a record algorithmic qubit score of #AQ 64. This milestone was achieved on an IonQ Tempo system, establishing IonQ as the only company to reach #AQ 64, setting a new standard for quantum systems. With each increment in #AQ value, the useful computational space for running quantum algorithms doubles. A system with #AQ 64 is capable of considering more than 18 quintillion (i.e., 2^64) different possibilities – more than 268,435,456 times more powerful than #AQ 36 (i.e., 2^36), which IonQ achieved in early 2024.
[Link] IonQ, supported by Air Force Research Lab (AFRL) research, announced a key advance in quantum networking today. The company successfully converted photons from visible to telecom wavelengths on a prototype system, enabling trapped barium ions to interface with standard fiber optic infrastructure. This breakthrough brings long-distance quantum computer connectivity closer to reality.
[Link] Quantum computing company Alice & Bob revealed research findings showing their cat qubits maintain resistance to bit-flip errors for more than 60 minutes. The hour-long coherence times mark a substantial leap forward from the prior benchmark of 430 seconds established on Alice & Bob’s Boson 4 chip in 2024.
The weekly 🤯
[Link] Routing photonic entanglement toward a quantum internet. Researchers at Tohoku University developed a highly efficient photonic router for the quantum internet, capable of directing single and entangled photons with minimal loss (0.06 dB, ~1.3%) and high fidelity (>99%). The router uses a novel electro-optic design with a parallelogram-shaped interferometer, maintaining photon polarization at telecom wavelengths. It successfully routed two-photon entangled states with 97% interference visibility, making it compatible with existing telecom networks and advancing practical quantum technologies like secure communication and quantum computing.
📑 bookmark for later
[Link] Quantum Computing in 2025: 80 Key Players Shaping the industry
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.