Pillars,
not headlines.
Tech moves too fast for daily news to mean anything. We organize what matters into long-running, continuously-updated topic hubs — and let the headline churn happen at the edges.
The AI memory bottleneck.
The constraint that's quietly reshaping the trillion-dollar AI compute build-out: explained, tracked, and updated as it evolves. Six solution paths, four still unresolved.
What I'm following.
In-memory computing: analog and digital compute-in-memory for AI inference
Tracks compute-in-memory architectures that perform matrix operations directly in memory arrays, eliminating data movement overhead.
Carbon nanotubes in energy storage
Tracks CNT deployment as conductive additives enabling higher energy density and faster charging than carbon black.
Carbon nanotubes in computing: transistors, interconnects and the path beyond silicon CMOS
Tracking carbon nanotube transistor development as they demonstrate superior performance metrics but face manufacturing challenges for production beyond 2nm.
HBM4 memory supply crunch
Tracks the yield and qualification bottleneck constraining HBM4 supply despite all three producers achieving mass production.
The advanced packaging race
CoWoS, EMIB, SoIC. Why the assembly line became the new fab.
Export controls & the new silicon geopolitics
The expanding chokepoints between the US, China, and the supply chain in the middle.
Long-form, this week.
Samsung's HBM3E qualification, finally. The full timeline and what it means for NVIDIA's 2026 allocation strategy.
Samsung passed NVIDIA's HBM3E qualification in September 2025, eighteen months late and third in line. The delay cost it the Blackwell cycle, left SK Hynix with 70% of HBM4 allocation, and turned a technical milestone into a symbolic recovery rather than a commercial win.
The hyperscalers built nuclear deals nobody can finance. Here's what happens when reality hits.
Big Tech signed power purchase agreements for 9.8 gigawatts of small modular reactors. The largest investment announced covers 6.9 percent of one plant's construction cost, and the first electrons arrive no earlier than 2030.
Commerce's HBM3E export guidance is narrower than reported. The HBM4 ambiguity is the real story.
The December 2024 HBM rule contains three distinct exemptions and leaves HBM4 in regulatory limbo, but most coverage missed the fine print.
Why every AI training run is now a packaging negotiation.
The bottleneck no longer measured in GPUs but in CoWoS slots. How TSMC quietly became the gatekeeper of the entire frontier-AI roadmap, and what the 2027 packaging map looks like.
Cerebras WSE-4 is generally available. The benchmarks are out, and the numbers are real.
Cerebras began shipping its CS-4 wafer-scale accelerator in August 2026, claiming up to 30 times faster inference than GPU systems. All performance figures come from the vendor; no independent lab has verified the numbers.
The essay shelf.
The hyperscalers built nuclear deals nobody can finance. Here's what happens when reality hits.
Big Tech signed power purchase agreements for 9.8 gigawatts of small modular reactors. The largest investment announced covers 6.9 percent of one plant's construction cost, and the first electrons arrive no earlier than 2030.
↳ AI · MemoryWhy every AI training run is now a packaging negotiation.
The bottleneck no longer measured in GPUs but in CoWoS slots. How TSMC quietly became the gatekeeper of the entire frontier-AI roadmap, and what the 2027 packaging map looks like.