Featured

September 2026

How Long Does a GPU Last?

GPU depreciation at hyperscalers isn't about hardware lifespan—it's about accounting timing: chips earn half their lifetime cash in two years but are depreciated evenly over six.

Perspectives

September 2026

The 10 Companies AI Can't Scale Without

As AI infrastructure scales past GPU constraints, supply chain bottlenecks are shifting to memory, power equipment, and data center land—creating 2027-2028 pricing opportunities the market hasn't yet valued.

Perspectives

September 2026

How the GPU Became Collateral

Three years ago nobody would lend billions against GPUs. Now the paper is rated A3 and sits in insurance portfolios.

Perspectives

August 2026

How Matic Got Robots into 10,000 Homes

Seven lessons from Matic's co-founder on their journey building robots that are actually deployed and loved in homes

All Articles

Thank you! Your submission has been received!

Oops! Something went wrong while submitting the form.

Perspectives

August 2026

Who Makes Money When Inference Gets 10x Cheaper?

A layer-by-layer map of where the dollars land when the price of intelligence collapses.

Funding Announcements

August 2026

Investing in Synthefy: Building Foundation Models for the World’s Numbers

We are proud to share that Wing led Synthefy's $6.5 million seed round to build Structured Data Foundation Models (SD FMs)

Perspectives

August 2026

The Simulation Stack in Robotics

Physics-based simulation engines and neural network world models offer complementary approaches to generating synthetic training data at scale for robot learning, though both face sim-to-real gaps that startups are addressing through domain randomization, asset generation, and real-to-sim reconstruction technologies.

Perspectives

August 2026

The AI Memory Stack

AI's memory problem isn't buying more HBM. It's managing a full hierarchy — GPU cache to cold storage — with a different winner at every tier.

Perspectives

July 2026

How AI Tokens Are Made

Inside the five scheduling, memory, and routing problems that turn GPU compute into output tokens — and why neoclouds keep buying the companies that solved them.

Perspectives

July 2026

The Data Pyramid in Robotics

The robot data pyramid framework combines seven types of training data—from internet video and human demonstrations to simulation and real-world deployments—to address the critical data bottleneck limiting the development of physical AI and autonomous robotics systems.

Perspectives

July 2026

Kimi: When Frontier Weights Become Open

How Kimi K3's open-weight release is reshaping AI industry margins and benefiting compute infrastructure over model APIs.

Funding Announcements

July 2026

Investing in Bespoke Labs: The Training and Verification Layer for Agentic AI

Why we led Bespoke Labs' Series A — and why the training and verification layer is the most important real estate in the agentic stack.

PORTFOLIO

The companies these perspectives come from.

View Portfolio

Team

The investors writing these pieces.

Meet the team