Why Silicon Photonics Is the Quietest Revolution in AI Hardware
The AI hardware debate obsesses over GPU compute and memory bandwidth. Meanwhile, the interconnect layer — the part that moves data between chips — is where the real constraint lives.
Read More
Quantum Error Correction: Where the Math Meets the Fab
Surface codes look clean on a whiteboard. Getting them to work on actual hardware — with real decoherence times and crosstalk — is a different problem entirely. Here's what the benchmarks don't tell you.
Read More
Custom Silicon for the Edge: RISC-V's Moment Has Arrived
The question is no longer whether RISC-V can compete — it's which companies will own the EDA stack and the toolchain that makes RISC-V custom silicon practical for volume production.
Read More
Beyond FinFETs: What 2D Materials Mean for the Transistor Roadmap
Intel and TSMC are betting on 3D stacking to extend Moore's Law. The 2D materials community thinks the answer is a different channel material altogether. The evidence is more interesting than the headlines suggest.
Read More
The Case for Autonomous Fab Operations
Semiconductor yields are still partly determined by human judgment calls made at 3am. Autonomous process control isn't just an efficiency play — it's how the industry gets to consistent sub-2nm volume production.
Read More
Quantum Networking Is Not a Science Project Anymore
QKD deployments have been running on dark fiber for two years in Asia and Europe. The US government just noticed. Here's what the infrastructure buildout looks like and who the buyers actually are.
Read More
Gate Oxide at Sub-2nm: The Material Problem Nobody Talks About
The transistor scaling roadmap has a materials bottleneck that gets almost no coverage outside of device physics conferences. Equivalent oxide thickness below 4 angstroms is not a lithography problem — it's a chemistry problem.
Read More
Spintronics and the Memory Architecture AI Actually Needs
SRAM is fast but power-hungry. DRAM is denser but too slow for in-memory compute. The neuromorphic community has been arguing for a decade that spin-transfer torque RAM is the answer. Here's where that bet stands.
Read More
Why Deep Tech Investing Requires a Different Clock
Standard venture fund economics were designed for software. The ten-year clock, the DPI expectations, the portfolio construction math — none of it maps cleanly onto companies building physical hardware at the device level.
Read More
The Geopolitics of Semiconductors Is Now a Fund Thesis
The CHIPS Act, export controls, and allied nations' domestic fab investments have changed the investment calculus for semiconductor companies in ways that go beyond individual company risk profiles.
Read More
Light Over Copper: The Physics Case for Photonic Interconnects
Copper interconnects lose energy as heat. Photonic interconnects lose energy as light. The second number is smaller — and the gap grows wider as data rates climb past 400Gbps.
Read More