RAMmageddon – DigitalWasteland

Jens Klasen
  • 0.0 (0)
  • Technology

Jens Klasen on the upheaval in enterprise IT. Memory is scarce, servers keep getting pricier, and hypervisor, Kubernetes and cloud decisions are all up for grabs. Season one, "Rewind", starts with Jens's latest briefing and travels back to RAMmageddon Part 1. Hosted by Matilda. Voices are AI-generated: Jens's voice is a clone of his own, the ideas and analysis are his. Jens works at Everpure (Portworx); any connection to his employer is disclosed. Original briefings at klasen.ai.

  1. 1d ago

    RAMmageddon Part 11 – The Bill of Materials Tips Over

    Season two starts in the present. In the third quarter of 2026, the PC market drops by more than 20 percent while Samsung and Micron report record profits. Jens explains why that is not a market disruption but a redistribution, and why the next two years are a window for modernization. In this episode Memory and SSD now make up 36 to 42 percent of a mainstream PC's component costs, up from about 16 percent at the start of 2025 (Omdia)IDC and Omdia: PC shipments down 20.1 and 21.2 percent in Q3 2026Samsung's preliminary guidance of 107.4 trillion won operating profit (divisional results due October 29), and Micron's record quarter (14 weeks instead of 13)Where the wafers go: Nvidia, TSMC, HBM and NAND allocated through 2027What it means for mid-sized companies in Germany, Austria and Switzerland: client refresh, data center, capacity costsModernize instead of buying more, and why small, task-specific AI models fit the shortageFull briefing with all sources: https://klasen.ai/en/briefing/rammageddon-teil-11-stueckliste-kippt/ Voices are AI-generated (Jens's voice is a clone of his own). Content and analysis by Jens Klasen. Transparency: Since September 16, 2026, Jens has been a Cloud Native Architect at Everpure (formerly Pure Storage), on the Portworx team. This podcast is a personal project; the opinions are his own.

    RAMmageddon Part 11 – The Bill of Materials Tips Over

  2. Oct 5 ·  Bonus

    RAMmageddon Bonus – I Cloned My Voice

    A bonus episode about this podcast itself. On a Wednesday evening, Jens asked how good voice cloning really is these days. By Friday afternoon, RAMmageddon – DigitalWasteland was live on Spotify and Apple Podcasts, in a voice he never recorded. Matilda asks how it went, what works surprisingly well, and where you have to listen closely. In this episode Three steps to the clone: a muffled local open-source model, an instant clone from a two-minute voice memo, and a professional clone trained on just under 50 minutes of EnglishWhy the microphone and the room matter more than the modelAbout 40 hours from idea to first episode, and a full season in two languages for the price of a monthly subscriptionWhere to listen closely: pronunciation, consistency, recording qualityJudgment stays human: the AI finds inconsistencies, deciding what is right is still Jens's jobIs this okay? Why transparency is what mattersAll of Jens's briefings: https://klasen.ai/briefing/ Voices are AI-generated (Jens's voice is a clone of his own). Content and analysis by Jens Klasen. Transparency: Jens works as a Cloud Native Architect at Everpure (formerly Pure Storage). This podcast is a personal project.

    RAMmageddon Bonus – I Cloned My Voice

  3. Oct 2

    Rewind 17 – RAMmageddon Part 1: Hardware Renewal 2026

    The season finale rewinds to where it all began: February 2026 and the very first RAMmageddon briefing, "Time for preparation, not panic orders". With memory prices soaring and server prices rising fifteen to twenty percent, Jens argues for extending support and using the time gained to rethink architecture – not for panic orders or a blind lift-and-shift into the cloud. The episode closes the circle back to episode 01, The Mediator Between Worlds. In this episode The facts as of February 2026: DRAM, server prices, CPU lead times, sold-out capacityWhy this is a structural reallocation driven by AI demand, not a temporary dipSupport extensions as a sensible bridge – if the time is used activelyProofs of concept now: virtualization, KubeVirt, SDN and container platformsWhy an unprepared lift-and-shift into the cloud is a cost trapThe formula behind the whole series: extend support, rethink architecture, run POCs, order with precisionOriginally published in German on February 10, 2026: https://klasen.ai/briefing/rammageddon-teil-1-hardware-renewal/ English version: https://klasen.ai/en/briefing/rammageddon-teil-1-hardware-renewal/ Voices are AI-generated (Jens's voice is a clone of his own). Content and analysis by Jens Klasen.

  4. Oct 2

    Rewind 16 – RAMmageddon Part 2: The Geopolitical Escalation

    Three weeks after Part 1, the hardware crisis gained a new dimension: geopolitics. US semiconductor tariffs, Europe's structural dependency, the Taiwan risk and, from February 28, 2026, the war in Iran blocking the Strait of Hormuz turned a chip shortage into a perfect storm. Jens explains why this makes data sovereignty and vendor diversification mandatory – and why the crisis is also a chance to finally modernize software instead of swapping hardware. In this episode Section 232 and Section 301: the tariff escalation and what it means for EuropeTaiwan, TSMC and Europe's place in the queueHow a war in the Middle East hits IT supply chains and data center energy costsThe silver lining: forced to think sustainably, with time to fix causes instead of symptomsSoftware resilience instead of hardware redundancy – and why "AI equals GPU" is often an expensive mistakeThe extended formula, with data sovereignty and vendor diversification as guardrailsOriginally published in German on March 5, 2026: https://klasen.ai/briefing/rammageddon-teil-2-geopolitische-eskalation/ English version: https://klasen.ai/en/briefing/rammageddon-teil-2-geopolitische-eskalation/ Voices are AI-generated (Jens's voice is a clone of his own). Content and analysis by Jens Klasen.

  5. Oct 2

    Rewind 15 – RAMmageddon Part 3: Why Software Is Now the Only Answer

    In March 2026, Google's TurboQuant cut the memory needs of AI models by a factor of six – proof, for Jens, that smarter software beats more hardware. Meanwhile memory prices kept breaking records, and Europe's Chips Act could not deliver fast enough. Jens explains why AI-assisted coding turns months-long modernization proofs into two-week sprints, and why it only works in experienced hands. In this episode TurboQuant: what it changes, and what it doesn't (inference, not training)The state of the market in March 2026: DRAM contract prices, HP's memory cost share, SK Hynix's outlook to 2027Europe's Chips Act: impressive investments, a stagnant ten percent share of global productionVibe coding: "Code has become cheap. Architecture has stayed expensive."Mistral Vibe and Devstral as a sovereign European optionFour concrete recommendations: start now, use senior expertise, prove it with POCs, think about data sovereigntyOriginally published in German on March 31, 2026: https://klasen.ai/briefing/rammageddon-teil-3-software-antwort/ English version: https://klasen.ai/en/briefing/rammageddon-teil-3-software-antwort/ Voices are AI-generated (Jens's voice is a clone of his own). Content and analysis by Jens Klasen.

  6. Oct 2

    Rewind 14 – An AI Prompt Is Not Text. It Is Software.

    A detour from the hardware crisis: Jens built a personal AI coaching app, JK-Fitt, mostly with Claude Code – and learned his biggest lesson from a single component, the coach's system prompt. His first version was friendly, competent and completely interchangeable, until he realized he had described personalities but no decision logic. The takeaway for every boardroom: AI tools accelerate building, not knowing what should be built. In this episode Why AI-assisted coding is about discipline, not speedFour rounds of prompt rewrites – and "a bug in my thinking"The three layers of a working system prompt: attitude, expert sources, and a rulebook of eight rulesAI tools as a microscope: they magnify whatever clarity – or lack of it – already existsThree rollout rules: start with the product owner, measure rework instead of speed, treat system prompts like source codeOriginally published in German on April 10, 2026: https://klasen.ai/briefing/ki-prompts-sind-software/ English version: https://klasen.ai/en/briefing/ki-prompts-sind-software/ Voices are AI-generated (Jens's voice is a clone of his own). Content and analysis by Jens Klasen.

  7. Oct 2

    Rewind 13 – RAMmageddon Part 4: Hardware Lens Off, Service Lens On

    "We understand that we should modernize. But where do we start?" In Part 4, Jens answers the question he kept hearing from CTOs, IT leaders and even a CFO. Classic sizing tools only ask which box fits the refresh. The service lens reads the very same discovery data and asks which software drives the load, and what is necessary versus simply grown over the years. In this episode The market as of late April 2026: server DRAM up roughly another 60 percent, NAND allocations closed, six-month CPU lead timesTwo lenses, one dataset: how the same telemetry leads to "buy the same again" or to "modernize first, then size leaner, then buy"Six reading lenses a refresh report overlooks, from naming conventions and provisioned vs. active memory to workload concentration and datastore mappingA case from the field: a nine-host Hyper-V environment with 27 percent RAM utilization, and why the answer was a third less hardwareThree consequences for refresh planning: run a discovery first, get an architectural reading, and accept that the result will be uncomfortableOriginally published in German on April 30, 2026: https://klasen.ai/briefing/rammageddon-teil-4-service-brille/ English version: https://klasen.ai/en/briefing/rammageddon-teil-4-service-brille/ Voices are AI-generated (Jens's voice is a clone of his own). Content and analysis by Jens Klasen.

  8. Oct 2

    Rewind 12 – RAMmageddon Part 5: The Map – Seven Patterns in Which Money Evaporates in 2026

    "Got it, but where do we start?" After Part 4, that was the question Jens heard most. Part 5 is his answer: a map of seven patterns that show up in almost every mid-sized data center once you know what to look for, each read through the same three questions. What do we see, what does it mean, and which lever fits? He closes with a bridge: sovereign European cloud on a portable open-source platform layer, as both transition and target architecture. In this episode The map in three layers: consumption, modernization and architectureThe seven use cases: database IOPS hotspots, zombie VMs, legacy operating systems, SQL licenses, container candidates, cloud bursting, and data modernization as the AI door-openerThe hotspot matrix: quick wins at the top, data modernization as the strategic lever with the longest lead timeWhy hyperscalers are a more expensive bridge in 2026, and what six sovereign German and European providers offer insteadJens's definition of sovereignty: not "no cloud", but the freedom to switch providers at any timeOriginally published in German on May 5, 2026: https://klasen.ai/briefing/rammageddon-teil-5-landkarte/ English version: https://klasen.ai/en/briefing/rammageddon-teil-5-landkarte/ Voices are AI-generated (Jens's voice is a clone of his own). Content and analysis by Jens Klasen.

See All (20)

Trailer

About

Jens Klasen on the upheaval in enterprise IT. Memory is scarce, servers keep getting pricier, and hypervisor, Kubernetes and cloud decisions are all up for grabs. Season one, "Rewind", starts with Jens's latest briefing and travels back to RAMmageddon Part 1. Hosted by Matilda. Voices are AI-generated: Jens's voice is a clone of his own, the ideas and analysis are his. Jens works at Everpure (Portworx); any connection to his employer is disclosed. Original briefings at klasen.ai.

Information