Anjney Midha

swyx × Anjney Midha

AI infrastructure’s binding constraint may be aligned execution rather than money or compute: best-in-class MFU is 60–70%, while small planning errors compound across organizational layers.Amp’s neutral, multi-cloud, multi-silicon grid targets 1.3 gigawatts of demand—roughly $40 billion of cloud spend—against teams that may need 6 gigawatts of spikes over four years, but community support, trust, and culture remain unresolved execution risks.

Peter Diamandis × Anjney Midha × Bonnie Chan × Dave Blundin

AI deployment has already outgrown venture scale, with Peter Diamandis estimating $1 billion a day in U.S. deployment and potentially $3 billion a day by 2030.Strategic investors, public markets, credit and sovereign capital must join the funding effort as reasoning models generate roughly 10 times more tokens and electricity becomes the infrastructure constraint.Vertical applications offer lower-capital early-stage economics, while institutional access and political backlash remain key risks to monitor.

Anjney Midha × Liam Fedus × Ekin Dogus Cubuk

Periodic Labs’ $300 million seed, led by Andreessen Horowitz, backs an AI-scientist thesis where nature supplies the final reward through automated experiments rather than internet-scale training.Scaling may continue while noisy data and missing negative results slow out-of-domain discovery; the about 135 Kelvin superconductivity benchmark provides a measurable north star, while copilots for advanced industries offer a nearer-term land-and-expand business.

Anjney Midha × Liam Fedus × Ekin Dogus Cubuk

Periodic Labs is betting that “Nature is our RL environment,” using experiments to supply the ground truth and reward signals that internet-trained models lack for physics and chemistry.High-temperature superconductivity offers a falsifiable benchmark, while an intelligence layer for industrial R&D could commercialize the stack through scoped deployments, mid-training, simulation, autonomous synthesis, and experimental iteration.

Mark Chen × Jakub Pachocki × Anjney Midha × Sarah Wang

GPT-5 makes adaptive reasoning and agentic behavior the default, shifting competition toward thinking budgets, latency, reliability, and economically relevant discovery rather than saturated benchmarks.OpenAI’s automated-researcher ambition requires longer planning, persistent memory, and honest recovery from failure, while scarce compute, energy, and robotics capacity remain constraints worth monitoring.

Jack Parker-Holder × Shlomi Fruchter × Anjney Midha × Marco Mascorro × Justine Moore × Erik Torenberg

Genie 3 turns text prompts into navigable, real-time worlds with one-minute spatial memory, frame consistency, and emergent physical behavior, combining capabilities previously split across Genie 2, Veo 2, and GameNGen.Its synthetic environments could provide scalable, safer experience for agents and robotics, but Genie 3 remains a research preview with no concrete broader-access timeline and falls short of a faithful world simulator.

Martin Casado × Anjney Midha × Erik Torenberg

US AI policy has shifted from “PauseAI” toward building American leadership, with open weights, an evaluations ecosystem and sovereign AI markets replacing broad restrictions as the central commercial framework.SB 1047’s downstream liability risk and DeepSeek’s frontier proximity exposed the chilling cost of theoretical safety claims, while the action plan’s weak execution detail and omission of academia remain unresolved constraints.

Nathan Labenz × Anjney Midha

Sovereign AI is emerging as control over technical, legal, and cultural dependence, making jurisdiction and indemnification as important to enterprise buyers as model performance.Countries unable to match frontier talent can buy or partner today, then build local productization, continual post-training, and distribution over a 10- to 20-year horizon.Open models fast-following closed systems within roughly six months or less makes localization plausible, while AI factories, export controls, and Gulf infrastructure commitments create long-duration strategic dependencies.

Anjney Midha × Guido Appenzeller

HUMAIN’s announced $100 billion-$250 billion buildout signals that sovereign AI is challenging the cloud-era concentration of workloads in the US and China, with roughly 500 megawatts as its “atomic unit.”High-density AI factories require GPUs, liquid cooling and committed power, while domestic model control is becoming cultural and information infrastructure, raising a strategic choice between exporting US-allied capacity and leaving countries reliant on DeepSeek.

Anjney Midha

AI sovereignty is shifting from adoption to build-or-buy, with compute, energy, data and regulation creating opportunities for alliances between hypercenters and compute deserts.Sovereign NVIDIA orders placed 12–36 months ahead and technically credible founders are key signals, while fragmented US rules, energy constraints and inference liability could drive startups abroad.