Apr, 2026 : Marvell Acquires Polariton To Advance Optical Scaling
📅 - Marvell Technology is buying its way deeper into optical physics, snapping up Polariton Technologies as pressure builds on data center interconnects under AI workloads. The deal, with no price attached, reflects a growing unease across the industry - current optical scaling paths are starting to look strained just as demand accelerates.
Marvell Technology is not short on optical expertise, but this move suggests incremental gains from silicon photonics alone may not stretch far enough. The company is reaching for something less conventional, and arguably less proven, to keep bandwidth growth aligned with AI infrastructure needs. Swiss Photonics Niche Polariton Technologies comes out of [...][... Check source for end of article ...]
📅 - Dubai Launches AI Data Center in Silicon Oasis with VOLT - Dubai is doubling down on AI infrastructure, with the Dubai Integrated Economic Zones Authority forming a joint venture with VOLT to build an AI-ready data center in Dubai Silicon Oasis. The project adds weight to the emirate's push to anchor global AI workloads locally, not just host them remotely.
The announcement lands alongside a broader shift in how governments are thinking about AI - not as a pilot project, but as core infrastructure. In Dubai's case, the ambition is unusually explicit. Within two years, half of government services are expected to run on agentic AI. That is not a distant roadmap. It is a deadline. And it pulls infrastructure decisions forward, fast.
The joint [...]
📅 - Global IT Spending Hits $6.31T As AI Drives Growth Surge - Global IT spending is set to hit $6.31 trillion in 2026, according to new figures from Gartner, marking a 13.5 percent increase year over year as AI infrastructure demand continues to reshape enterprise budgets worldwide. The numbers suggest momentum is holding, though not evenly across sectors or regions, and not without friction in supply chains and pricing.
The headline figure lands quickly, but the mechanics underneath it take longer to unpack. Growth is not being driven by a broad-based uplift across IT. It is narrower than that, more concentrated. AI workloads are forcing a recalibration of where capital goes, and how fast it moves. Data center investment is accelerating at a pace [...]
📅 - Study: AI Networking Efficiency Could Cut Data Center Emissions - NeuReality is making a case that the real inefficiency in AI infrastructure isn't compute, but the network that feeds it, with new research suggesting that better data movement alone could cut up to 200 tons of CO₂ emissions per rack each year. The findings arrive as operators face rising pressure to scale AI without expanding power footprints.
The numbers are attention-grabbing, but the argument underneath is more familiar. As AI workloads spread across clusters, the system becomes less about raw compute and more about coordination. Data has to move constantly - between GPUs, across nodes, through layers of orchestration. When that movement slows, everything else stalls.
And stalled [...]
📅 - QumulusAI Raises $45M To Accelerate GPU Rollout Pace - QumulusAI has lined up fresh capital to chase a narrowing window in AI infrastructure, securing a $45 million convertible note facility from ATW Partners, with $15 million already deployed. The funding is being funneled straight into GPUs and data center buildouts, as enterprise demand for dedicated AI compute continues to outpace supply across U.S. markets.
QumulusAI is not trying to finesse margins here. It is trying to move faster than procurement cycles typically allow. The company is using the facility to lock in hardware, secure colocation space, and stand up clusters before customers lose patience or shift elsewhere.
Timing matters more than elegance. Enterprises increasingly want [...]
📅 - SUSE And Switch Advance AI Data Center Digital Twins - SUSE and Switch are deepening their partnership around digital twin technology, aiming to turn large-scale data centers into continuously simulated environments as AI workloads grow more complex. The milestone centers on combining open source infrastructure with NVIDIA's simulation stack to operationalize what both companies describe as AI Factories - though the practical implications are still taking shape.
SUSE and Switch are effectively trying to collapse multiple layers of infrastructure into one coordinated system. The pitch is straightforward: instead of running AI models, simulations, and rendering workloads separately, run them together - on the same hardware, in near real time, [...]
📅 - Keppel, Shell Launch Immersion Cooling Data Center Pilot - Keppel and Shell Singapore are preparing to submerge servers in liquid as part of a year-long pilot aimed at rethinking how data centers handle heat. The testbed, scheduled for May 2026, will trial immersion cooling inside a live facility, targeting energy efficiency gains as AI workloads intensify pressure on conventional infrastructure across tropical markets.
Keppel is not abandoning air cooling overnight. Most of its facilities still rely on it. But the direction of travel is becoming harder to ignore. AI workloads are denser, hotter, less predictable. Traditional airflow systems are starting to look like a constraint rather than a default.