Introducing Radiant.
By merging the software and compute capabilities of Ori with the long duration capital and powered-land assets of @Brookfield they created an end-to-end, fully integrated AI infrastructure platform.
From shovel to model.
Welcome to the age of abundant AI.
https://t.co/4zqrSWbWbS
The #sovereignAI space is en feu (a tip of the hat to @EmmanuelMacron who pulled in $108B for French #AI this week).
There are four execution archetypes today and we mapped dozens of deployments into each one.
This is the state of play.
https://t.co/vwEw8my9r8
Ever listened to AI-generated speech and thought it sounded just a bit too robotic? Even as open-source voice AI advances, capturing authentic emotion, pacing, and nuance of human conversation remains a hurdle.
Fish Audio S2 Pro changes the equation. This flexible open-weight model features:
✨ Deep emotional control with 15,000+ unique tags
🗣️ Seamless multi-speaker generation in a single stream
🌍 Automatic synthesis across 80+ languages
⚡ Voice cloning from short reference samples
Ready to test it? Our latest guide walks you through exactly how to deploy Fish Audio S2 Pro on Radiant using an NVIDIA H200 GPU. Check out the step-by-step tutorial and listen to the impressive audio samples on our blog: https://t.co/fx9Om1kWh2
Multimodality, multiple models. The difference matters and Deepak Manoor steps us through why with the @NVIDIA Nemotron 3 Nano Omni model and how to run it on the Radiant Cloud.
https://t.co/he6u02jdxR
You can’t power sovereign AI with retrofitted cloud infrastructure.
We are proud to welcome @Radiant_Infra as an Official Partner of RAISE Summit 2026.
With a direct pipeline to a $100B AI infrastructure investment program, Radiant builds and operates purpose-built AI factories specifically designed for sovereign governments, global enterprises, and telecommunications providers.
Built on the @NVIDIA DSX reference design and operating across 20+ locations worldwide, Radiant delivers AI compute with utility-grade economics, operational resilience, and planning horizons aligned with the long-term demands of the intelligence economy.
This July 8-9 at the Carrousel du Louvre, Radiant joins the leaders architecting the physical and capital infrastructure for the next generation of intelligence.
Where visionaries and AI converge.
#RAISE2026 #RAISESummit #AIInfrastructure #SovereignAI
At Radiant, we're always keeping an eye on what's next because the customers we work with are innovating non-stop, and staying ahead of where things are headed is how we'll deliver real value to them. And right now, something genuinely interesting is happening in AI networking.
Copper has been the workhorse of the data center for decades. But as AI models scale into the trillions of parameters and clusters stretch across racks, copper is hitting a wall. The next era belongs to light.
NVIDIA's upcoming Vera Rubin Ultra platform makes that official. On top of the leap in compute density, 800 VDC power delivery, and a fully liquid-cooled Kyber rack design, Vera Rubin Ultra brings Co-Packaged Optics (CPO) into the picture, taking photonics from a theoretical option to something you actually need to build at this scale.
But Rubin Ultra isn't just a silicon story. It's going to take a serious rethink of power, cooling, and facility design at gigawatt scale. That's exactly what we're building Radiant for.
Read more about the shift to light here 👇
https://t.co/dCf2iU1wUM
7.5 GW. 10 new gas plants. One data center. Meta didn't worry about connecting to the grid, they built their own power.
This is the new playbook: capital → land → power → compute. Whoever owns the stack wins.
https://t.co/5RJaOdfhD4
@Radiant_Infra offered one of the most comprehensive glimpses into the full stack of AI infra and the constraints underneath. The answer: Parallel execution and site pre-entitlement.
🧵
The clarity of this conversation between James Pringle and Mahdi Yahya is remarkable. Incredibly accessible but brimming with insight on the state of #AIInfrastructure, #SovereignAI and the importance of an end-to-end offering to accelerate the shovel to token cycle.
Well worth your time.
https://t.co/5E9vBuXkNi
When teams start scaling AI workloads on #Kubernetes one of the first questions is inevitably about what scheduler to use. While the use case matters, Adrian Matei did the benchmarking to facilitate the discussion. Check out the stats on #Kueue, #Volcano and #Slinky: https://t.co/wDEw6CoD40
The definition of the private cloud is evolving in the world of AI-centric data centers. This is less about privacy and more about control. Patrick Wohlschlegel breaks it down in this highly accessible five minute read. https://t.co/7Q9YiiQNc6
#privatecloud#AIinfrastructure #AICloud
Smart money doing smart things:
"In what’s known as a classic powered land playbook, the venture will then look for buyers who want property that’s shovel-ready for data centers."
https://t.co/Fi0DsJLulE
Same #GPUs.
Same workloads.
Different schedulers.
Very different outcomes.
We benchmarked Kueue vs Volcano vs Slinky.
If you’re not thinking about scheduling, you are probably leaving performance on the table.
https://t.co/sQsHDjgJKq
"It's a £500 million commitment from the British Government to invest like a venture investor with the strength of a nation behind it, into AI's successes in the UK".
Sovereign AI (@UKSovereignAI) just launched a £500M UK Venture Fund, to accelerate the success of UK-based AI companies.
James Wise (@Jameswise), who Chair's the £500M British Government commitment, joined us live to share exactly what they have planned:
"We're predominantly going to be working with relatively early stage companies. Viewers may be thinking sort of seed and series A".
"Investing directly between £5 to £10 million. Sometimes a bit more, sometimes a bit less...We'll also be working with some of the real breakout companies in the UK as well".
"This evening, we'll be announcing some of the first investments we've made... but we'll also be announcing some of the first recipients of the compute we're offering as well".
The tide going out in #AIinfrastructure isn't a crisis. It's a filter.
Speculative buildouts are stalling. Capital is repricing. Projects are getting marked down or abandoned.
It is not a problem. Mahdi Yahya explains why.
https://t.co/6HAGUCwhBG
Based on the feedback from the #SovereignAI post - "The Control Plane is the Country" we have written a follow on piece with an execution blueprint. It also draws on the #AI literature -- McKinsey & Company, Tony Blair Institute for Global Change, World Economic Forum and University of Oxford -- and layering in the operational realities that the policy frameworks tend to skip past.
The four-stage nation-state playbook that takes a sovereign program from baseline classification through full operational independence and into ecosystem development. It includes a vendor evaluation framework built around the questions procurement officers should be asking but typically aren't.
The thesis is simple: if your AI infrastructure requires a phone-home to a vendor server in another jurisdiction to function, you have purchased data residency, not sovereignty.
The answer to "who controls the control plane" is the answer to "who has sovereignty." Everything else is preparation.
https://t.co/fRon3JVy3J
More evidence that building behind the meter is the way to build. The math is fairly straightforward here, the grid has already stranded more than 3GW and a variety of pressures are growing (public, permitting, political) that will strand more.
The problem facing #AI is an integrated one - capital, power, land, compute and software. It is only becoming more so.
Congrats to fellow Brookfield portco Bloom Energy on this deal. https://t.co/D9EqN6LedB
As one attendee said, "This was the best talk I saw all week. I can't believe they put it on the last day...10/10." So if you missed it, here it is: Chris Galloway at @NVIDIA#GTC talking about the long-term consequences of foundational design decisions and how they shape autonomy, resilience, and adaptability over the life of an AI system.
https://t.co/PUqCwZlT2c
#SovereignAI is not something you buy.
It’s a systems property.
It emerges from coherent design - and the elimination of dependencies.
Most strategies miss this.