45k user reviews & 1 million votes, over 29k web hosting brands & 85k plans since 2004!
Menu
WHTop → 📋 News → XCENA Raises $135M to Scale Memory-Centric AI Infrastructure
XCENA Raises $135M to Scale Memory-Centric AI Infrastructure
📅 - XCENA has raised $135 million in Series B funding, giving the memory-centric computing startup a $570 million valuation as AI infrastructure buyers confront mounting memory bottlenecks. The company plans to use the capital to expand globally, scale customer deployments and advance MX1, its computational memory product now being tested with selected partners worldwide across demanding AI workloads.
The round, co-led by Atinum Investment and IMM Investment, brings XCENA's total fundraising to $185 million. That is a serious amount of money for a company operating in one of AI infrastructure's less glamorous but increasingly critical layers: memory movement.
Not model training. Not GPUs on [...][... Check source for end of article ...]
📅 - SIA-Deloitte Report Puts Chips at Center of AI Buildout - A new SIA-Deloitte report says semiconductors make up more than 95% of the content value inside a leading AI server rack, recasting the data center boom as a chip supply contest rather than a software story. The study projects AI data center chip revenue could reach $1.2tn by 2028, intensifying policy and procurement pressure globally this year.
The report is useful because it drags the AI conversation down from model announcements and benchmark theater into the rack. That is where the money is going. Accelerators, memory, networking silicon, power chips, controllers, sensors, transceivers. Thousands of packaged chips doing unglamorous work so the glamorous systems can run.
The [...]
📅 - Gartner Warns AI Agent Governance Failures Ahead by 2027 - Gartner is warning that enterprises are setting themselves up for a second wave of AI agent disappointment, not because the technology cannot act, but because companies are granting and restricting that action badly. The firm expects 40% of enterprises to demote or decommission autonomous agents by 2027 after governance failures surface in production environments.
The message is pointed because it lands in the middle of a spending cycle where "agentic AI" has become the default answer to every board-level productivity question. Automate the workflow. Let the agent take action. Remove friction. Cut process time. Sounds clean in a vendor demo.
Then the agent touches a live [...]
📅 - SoftBank Plans €75bn AI Data Center Buildout in France - SoftBank said it will invest up to €75bn to build 5GW of AI data center capacity in France, starting with €45bn and 3.1GW in Hauts-de-France by 2031. The plan, unveiled around Choose France, pairs massive compute demand with French power, industrial land and a Schneider Electric manufacturing partnership in Dunkirk to localize core infrastructure for Europe buyers.
This is a huge number, even in the inflated arithmetic of AI infrastructure. Five gigawatts is not a campus. It is a national-scale energy and industrial planning question dressed as a data center program.
France wants the project read as proof that it can host AI infrastructure without becoming merely another customer market [...]
📅 - Bull, Foxconn to Build European AI Infrastructure Supply Chain - Bull and Foxconn will manufacture AI infrastructure across facilities in the Czech Republic and France, committing more than €120 million to a European supply-chain project aimed at AI factories and neo-cloud providers. The arrangement gives Europe a more local route for servers, racks and validation, though critical chips still come from global supply chains outside Europe.
The deal is industrial before it is technological. Bull brings high-performance computing and AI systems design. Foxconn brings manufacturing scale, component discipline and the ability to produce complex electronics without treating every system as a custom project.
That combination is the point.
Europe has spent [...]
📅 - NVIDIA Moves Beyond GPUs With Vera CPU and Rubin Systems - NVIDIA says its Vera CPU is now in production and its Vera Rubin rack-scale platform is ramping for agentic AI factories, drawing planned adoption from AI labs, hyperscalers, clouds and server makers. The announcements put CPUs, networking, confidential computing and full-rack integration at the center of the next AI infrastructure purchasing cycle starting this fall globally.
The company is not just selling faster chips here. It is trying to define what the AI factory stack should look like when agents move from answering prompts to running code, calling tools, retrieving data, evaluating results and generating follow-on actions.
That changes the infrastructure conversation. GPUs still [...]
📅 - Intel’s AI Server Plan Turns On Inference Economics - Intel used Computex 2026 to argue that AI infrastructure is entering a less theatrical phase: not bigger training clusters, but cheaper inference, denser CPUs, alternative accelerators, and servers that can run inside real enterprise constraints. The company introduced Xeon 6+, rackscale systems with SambaNova and Foxconn, and a Crescent Island GPU built around LPDDR5X rather than HBM.
The more interesting part is what Intel did not quite say. It is not trying to beat NVIDIA at NVIDIA's strongest point. Not here. Not directly. Intel is instead trying to define the next procurement fight around power, memory capacity, deployment cost, and operational control.
📅 -