RCRTech

A week ago, President Trump signed Executive Order 14420, which details security risks associated with data centers, AI, and advanced manufacturing. In particular, it appliesto high-voltage infrastructure operating at or above 69 kV, spanning not only physical equipment — transformers, inverters, generators, and BESS — but also associated software, firmware, and remote-access capabilities.

0 In sum – what to know: AI orders surge – Dell booked a record $60.9 billion in AI orders during the quarter and exited with a $95 billion AI backlog. Deployments intensify – AI infrastructure projects are becoming increasingly complex, with some requiring upwards of 50 unique designs to optimize performance, power, cooling, and the data center environment.

0 In sum – what we know: Supply-chain revival — Ming-Chi Kuo reports Rubin CPX is back on a production schedule beginning in Q1 2027, after Nvidia dropped it from the GTC 2026 roadmap in favor of Groq 3 LPX racks. (unchanged) A major redesign — According to Kuo, the chip abandons the shelved design’s 128 GB of GDDR7 for 168 GB of HBM4, a 2,300 W power envelope matching Rubin, and its own standalone MGX ETL racks.

0 In sum – what to know: Infrastructure expands – AWS plans to launch Saudi Arabia’s first Cloud Infrastructure Region in December, while up to 50MW of capacity is planned for HUMAIN’s AI Zone by 2028. Sovereign AI – Mistral and HUMAIN are targeting local infrastructure, customizable models and customer-controlled data, compute and operations, particularly for regulated industries.

RCRTech is a daily newsletter focused on AI Infrastructure — the data centers, fiber networks, edge interconnections, chips, data platforms, hardware and software that make AI work for enterprise verticals including telecom, finance, pharmaceutical and automotive.

In yet another staggering infrastructure investment, Anthropic is hoping to power Claude’s exponential growth by investing $35 billion for long-term compute access through Nvidia-backed Lamda (with Nvidia itself holding the lease on the data center, as well as providing chips to Lamda). Hut 8 will develop the facility, which will deliver about 350 MWs of capacity, according to the Wall Street Journal.

Progress towards greater autonomy in network operations depends on the readiness of operational data to feed AI models and agents. But that is about much more than data lakes. To move isolated pilots to trusted, closed-loop operations, CSPs must first answer fundamental questions about the nature and scope of their data.

Back from holiday, first day of Fall, and what do we have: a couple of stories, either side of the weekend, say something about the contradictory state of the mobile market – that telecoms has become too important to fail, but is not easy enough or valuable enough to pay for itself. Or something like that; that existing regulation and economics are tough, anyway, and that future R&D is being neglected, leaving it ripe for disruption by ‘big tech’.

0 In sum – what we know: Three wafers per rack – The CS-4 Nexus architecture houses three WSE-3 Turbo engines, lifting sparse FP16 compute to 750 PFLOPS per rack from 125 in the CS-3. Same silicon, more power – The WSE-3 Turbo keeps the CS-3’s 900,000 cores and 44GB of SRAM, with gains coming from packaging, power delivery, and higher clocks.

In another example of what some criticize as “circular financing,” Nvidia is investing $3.5 billion in convertible bonds issued by MediaTek, with MediaTek adopting the Nvidia NVLink Fusion platform and collaborating on RTX Spark™ and DGX Spark™ PC chips. Some analysts say this gives Nvidia a competitive edge, while others argue “circular financing” adds further risk.