=========================================== SAGE INTELLIGENCE BRIEF Monday, June 01, 2026 =========================================== LEAD STORY Intel killed Hyperthreading. The next-gen Xeon "Diamond Rapids" pushes to 192 cores and drops SMT entirely, while the Clearwater Forest line on 18A hits up to 288 cores with 576MB of L3. Intel's claiming the 192-core 6990E+ beats AMD's 192-core Epyc 9965 by 30% per thread. For anyone capacity-planning data center refreshes or VDI density, this changes the math. Dropping SMT means per-core licensing for Windows Server, SQL and VMware gets cleaner to calculate, but it also means you size differently than you have for the last decade. --- IT INFRASTRUCTURE ARCHITECTURE Intel ditches Hyperthreading in Diamond Rapids 192 cores, no SMT. Intel's betting that fat physical cores beat logical threads for AI and dense workloads. The licensing implication is real: per-core licensed software gets predictable again when you don't have to reason about thread contention, but you're paying for more physical cores to hit the same throughput. Watch how this lands against your VMware and SQL renewals. Source: https://www.theregister.com/systems/2026/06/01/intels-next-gen-xeons-to-pack-192-cores-abandon-smt/5248940 Netflix open sources Project Headroom to cut AI bills A Netflix engineer built a tool to slash AI inference costs and then open sourced it. If you're running any in-house inference, this is worth a look before your next GPU spend. Free tooling that reduces compute waste is the kind of thing that pays for itself in a quarter. Source: https://www.theregister.com/ai-ml/2026/05/31/netflix-wiz-creates-app-to-slash-ai-bills-then-open-sources-it/5248702 DuckDB Quack brings multi-user analytics over HTTP DuckDB announced Quack, a client/server protocol over HTTP that lets multiple instances connect to a shared backend. This pushes DuckDB from a single-node embedded engine toward something you can run as a small analytics service. For lean shops that don't want to stand up a full warehouse, this is a credible middle path. Source: https://www.infoq.com/news/2026/05/duckdb-quack-protocol/ Hardware prices are the new normal, not a blip The Register is treating rising hardware prices as structural, not transient. Tariffs, supply shocks and AI demand are keeping prices elevated. Budget your refresh cycles accordingly and stop waiting for prices to "come back down." They aren't. Source: https://www.theregister.com/personal-tech/2026/06/01/exploding-rockets-and-exploding-hardware-prices-make-for-a-lousy-new-normal/5248601 --- CYBERSECURITY & COMPLIANCE No notable developments tonight. --- CLOUD PLATFORMS & STRATEGY No notable developments tonight. --- NETDEVOPS & NETWORK AUTOMATION Intel pivots to power efficiency and cost in new chip designs Intel's framing its latest silicon around power efficiency and cost as AI workloads strain data center budgets. For network and infrastructure teams, the relevant part is thermal and power density planning. If your DC fabric and PDU capacity were sized for older Xeon power curves, the new parts shift your per-rack math. Plan your power budget around the chips you're actually buying, not last gen's TDP. Source: https://www.networkworld.com/article/4178385/intel-focuses-on-power-efficiency-and-cost-with-new-chip-designs.html --- AI IN INFRASTRUCTURE & AIOPS Erin Brockovich targets data center secrecy Brockovich is going after the lack of transparency around data center water and power consumption. This is a regulatory and reputational risk signal. If your AI roadmap depends on cheap hyperscaler capacity, expect more public scrutiny and potentially more cost passed down as utilities and municipalities tighten rules. Source: https://techcrunch.com/2026/05/31/erin-brockovich-takes-aim-at-data-center-secrecy/ The AI psychosis debate Equity's running a segment on whether tech leadership is "uniquely prone to AI psychosis." Strip the clickbait and there's a real point: decision-makers are over-indexing on AI capability and under-indexing on operational reality. When you're being pitched AI tooling, separate the demo from what survives production. Skepticism is a feature. Source: https://techcrunch.com/2026/05/31/making-sense-of-the-debate-over-ai-psychosis/ --- HARDWARE, GPU & COMPUTE Nvidia unveils RTX Spark Superchip at Computex Jensen Huang showed off an Arm CPU plus Blackwell GPU platform with 128GB unified memory, pitched as turning Windows into an "agentic AI OS." For the channel, the watch item is whether agentic local AI moves real workloads off the cloud onto the endpoint. If it does, your endpoint refresh spec and your data governance both change. Don't buy the OS marketing, but track the unified memory architecture. Source: https://www.tomshardware.com/laptops/nvidia-unveils-rtx-spark-superchip-at-computex-2026-new-platform-promises-to-turn-windows-into-an-agentic-ai-os-with-arm-cpu-blackwell-gpu-and-128gb-unified-memory Intel Clearwater Forest puts 18A in the data center Up to 288 cores and 576MB of L3 on Intel's 18A node. This is Intel proving 18A can ship at volume in the data center, which matters for anyone who wrote Intel off on fab competitiveness. More cores per socket means fewer nodes for the same workload, which changes rack count, cooling and switch port counts. Factor the density into your next DC design, not just the per-core spec sheet. Source: https://www.tomshardware.com/pc-components/cpus/intel-xeon-6-clearwater-forest-puts-18a-in-the-data-center-with-up-to-288-cores-576-mb-of-l3-cache-new-xeon-699 Nvidia Computex keynote sets the AI infrastructure tone ServeTheHome covered Huang's two-hour keynote spanning AI, PCs and robotics. The signal for infrastructure teams is that Nvidia keeps expanding from the GPU into full platform plays. The more they own the stack, the more your buying decisions get locked to their ecosystem. Watch the lock-in, not just the performance. Source: https://www.servethehome.com/nvidia-computex-2026-keynote-live-coverage/ Dell XPS 13 lands at $599 as a MacBook Neo rival Dell dropped a $599 XPS 13 with touchscreen and backlit keyboard, with student pricing, aimed squarely at the MacBook Neo. For fleet buyers, a premium-feature 13-inch at that price point reshapes the standard-issue laptop conversation. If you're refreshing road-warrior fleets, this is a credible Windows option that doesn't feel like a budget compromise. Source: https://www.zdnet.com/article/dell-xps-13-2026-macbook-neo-competitor/ --- NETWORK MANAGEMENT & MONITORING No notable developments tonight. --- MANAGED SERVICE PROVIDERS No notable developments tonight. --- IT VENDOR ECOSYSTEM & M&A No notable developments tonight. --- EDGE COMPUTING & IOT TV RS-232 ports as an automation hook ZDNet's reminding people the RS-232 serial port on commercial displays is still a real automation interface. For anyone running digital signage, conference room or retail edge deployments, serial control is the unglamorous backbone that keeps display fleets manageable. If you're standardizing AV in client sites, spec displays with RS-232 and build the control into your automation. Source: https://www.zdnet.com/article/what-is-the-tvs-rss232-serial-port-and-what-it-can-do/ Old Android phone as a Wi-Fi extender ZDNet walks through turning a retired Android phone into a Wi-Fi extender at no cost. It's a consumer trick, but the underlying point matters for edge and remote deployments: cheap repurposed hardware can patch coverage gaps before you commit to full mesh. Useful for temporary sites or pop-up deployments where a permanent solution isn't justified yet. Source: https://www.zdnet.com/article/how-to-turn-android-phone-into-wifi-extender-for-free/ --- WHAT TO WATCH The core-count arms race between Intel and AMD just got real again. Intel shipping 288 cores on 18A and dropping SMT, while AMD holds at 192 on Epyc, resets how you size everything: licensing, power, cooling and switch density. This is the week to revisit your DC refresh assumptions, because the per-socket density jump changes the cost model for the next three years. --- CONVERSATION STARTER Intel says its 192-core Xeon 6990E+ runs 30% faster per thread than AMD's 192-core Epyc 9965, and it does it without Hyperthreading. If that benchmark holds in production, you can hit the same throughput with fewer physical cores, which directly cuts your per-core software licensing on Windows Server, SQL and VMware. That's a budget line worth modeling before your next renewal cycle. ===========================================