Shield AI Air Force Deal, Google's TurboQuant, and Meta's Brain-Predictive AI
Chamath Palihapitiya
· Chamath Palihapitiya
· March 29, 2026 at 17:18
· ⏱ 5 min read
| Read on Substack ↗
Summary
The newsletter summarizes three recent tech breakthroughs: Shield AI's $12.7B valuation from a $2B funding round (defense autonomy), Google's TurboQuant quantization technique (up to 8x faster AI inference on Nvidia H100s), and Meta's TRIBE v2 brain-predictive AI model. The market implication is that AI inference efficiency gains (TurboQuant) directly benefit Nvidia's GPU infrastructure, while defense AI and neuroscience AI remain early-stage but signal continued investment.
•Shield AI raised $2B ($1.5B Series G) at $12.7B post-money, a 140% increase from $5.3B in March 2025.
•Shield AI's Hivemind autonomy software selected by U.S. Air Force for Collaborative Combat Aircraft program.
•Google's TurboQuant compresses KV cache to as low as 3 bits per value with no accuracy loss, cutting memory use by 6x and enabling up to 8x faster attention on Nvidia H100 GPUs.
•TurboQuant works online without retraining or calibration, tested on Gemma, Mistral, and Llama-3.1 models.
•Meta's TRIBE v2 trained on 1,000+ hours of fMRI data from 720 subjects, predicts brain activity from video, audio, and language without retraining for new individuals.
•TRIBE v2 achieves 2-3x improvement over previous methods and is released under CC BY-NC license for scientific use.
The article benchmarks Google's TurboQuant on Nvidia H100 GPUs, showing 6x memory reduction and up to 8x faster attention. Such inference efficiency gains would increase H100 throughput and demand, di
The article benchmarks Google's TurboQuant on Nvidia H100 GPUs, showing 6x memory reduction and up to 8x faster attention. Such inference efficiency gains would increase H100 throughput and demand, directly benefiting Nvidia's data center revenue.
Risk: TurboQuant is a research paper; adoption timelines are uncertain. Competitors (AMD, Intel) could develop similar quantization methods.
This newsletter, published March 29, 2026,
features Chamath Palihapitiya
discussing NVDA.
1 trade idea extracted by AI with direction and confidence scoring.