Writing · Build / Thinking / Field / DYOR

Articles include research.

Writing is the unified content surface; build notes, thinking notes, field records, and DYOR research all live in the same Articles board.

New content publishes to Cloudflare without a site release.

Writing

Articles

Build notes, thinking notes, field records, and DYOR research share one filterable Articles board; DYOR still keeps dedicated visual report pages.

High-End MLCC vs HBM: AI Hardware Demand, Bottlenecks, and Expansion Difficulty

High-end MLCC is not an HBM substitute and not an EMIB/CoWoS substitute. HBM is a memory bandwidth bottleneck close to the accelerator. MLCC is a power-integrity and reliability bottleneck. Both benefit from AI server complexity, but their value density, expansion economics and earnings transmission are very different.

Bitcoin Miners' Power-to-AI Pivot: BTDR vs HUT vs IREN vs CIFR

The core variable is no longer hash-rate ranking. It is whether power can be upgraded into financeable AI load. HUT and CIFR have already converted part of their power estates into long-duration AI colocation leases; IREN has pushed meaningful capacity into AI Cloud contracts and GPU cloud economics; BTDR owns a 3.0GW raw power portfolio and a fast-growing AI Cloud ARR base, but its large-scale AI colocation contract proof still lags. The highest certainty today is HUT, the highest beta is IREN / CIFR, and the largest relative re-rating option is BTDR if Tydal and its U.S. sites sign credible tenants.

ARM / QCOM / INTC / AMD: Four Chip Capacity Models

The four companies do not answer the same capacity question. Arm primarily sells IP, so capacity sits with licensees and the ecosystem. Qualcomm is mostly a fabless SoC vendor with some owned RF/filter manufacturing. Intel is the only one with large-scale owned wafers and advanced packaging. AMD is a fabless CPU/GPU/AI accelerator company whose upside depends heavily on TSMC, advanced packaging, and HBM supply.

SpaceX IPO and Space Stocks: RKLB Upside, Proxy Risk, and Dilution

The first-order reaction to SpaceX's S-1 is obvious: the space sector finally gets a publicly traded leader with scale, brand power, launch dominance, Starlink, NASA relationships, and a broader AI/orbital-compute narrative. Public space stocks have mostly been partial exposures: Rocket Lab for launch plus space systems, AST SpaceMobile for direct-to-device, Intuitive Machines for lunar services, and Planet Labs or BlackSky for Earth-observation data.

T1 Energy / TE: A High-Leverage U.S. Solar Manufacturing Turnaround

The TE bull case is not simply 'AI needs power, so every energy stock works.' It is a narrower and more fragile thesis: if T1 links G1_Dallas module revenue, G2_Austin cell production, 45X/domestic-content eligibility, and non-related customer demand, it can become a scarce U.S. solar manufacturing asset. If G2 financing or FEOC/45X eligibility breaks, common equity absorbs heavy downside through leverage and dilution.

Sivers / SIVEF: A Laser-Source Bottleneck Candidate in AI Photonics

Sivers Semiconductors is easy to label as AI photonics, CPO, 1.6T, SATCOM, or defense. The cleaner first-principles view is narrower: it is not a GPU vendor, switch ASIC vendor, full optical-module vendor, or hyperscaler network architect. It combines two enabling component layers: high-power DFB/CW laser arrays in Photonics and mmWave RFIC/beamforming technology in Wireless.

POET / MXL / HIMX: three layers of AI interconnect optionality

POET, MXL and HIMX can all be mapped to AI interconnect, optics, CPO or edge AI, but the first-principles layers are different. POET is a manufacturing-platform option: can Optical Interposer / EOI move optical-engine production closer to wafer-scale integration? MXL is the signal-chain layer: optical modules and AI data-center links need PAM4 DSPs, TIAs and retimers. HIMX is a display and ultra-low-power sensing company with AR, WiseEye and CPO optionality through FOCI.