The latest AI Infrastructure posts we have indexed from venture capital blogs and newsletters, plus the highest-scoring pieces of the past year and the people who write about it most.
Enterprise AI deployment spending surged 5.4x to $177B in 2026, driven by autonomous agents that now consume 80% of model tokens and enable new use cases in coding and customer support; frontier models command 69% of spend despite open-source gaining usage share.
Decision models—a new AI category that makes bounded choices rather than generating text—launched at $100M revenue in a week, then got replicated for $95 in open-source form within five days, signaling that software commoditization has reached a new inflection where defensibility now rests on integration and operations rather than capability alone.
Venture capital is shifting from software to physical technologies as AI collapses software development costs and the SaaS model loses defensibility; the strongest opportunities now lie at the intersection of AI-accelerated atoms—chips, energy, robotics, biotech—where capital intensity and technical moats remain defensible.
Software AGI has arrived as of November 2025, but only in coding; the real frontier is verification, not generation. Meanwhile, capability gaps in medicine and security reveal the pattern—models can do the work, but humans haven't scaled to review it.
As AI capex bubbles and shifts to local deployment, robotics emerges as the natural "application layer"—but mass adoption depends on extensibility, cheap hardware, and hobbyist-driven innovation, mirroring how personal computers displaced mainframes.
Nvidia's leadership crisis and market turmoil are testing AI's venture funding thesis; Anthropic's IPO pricing will signal whether a valuation reset extends from public markets into private AI rounds, with Nvidia's balance-sheet support for infrastructure as the transmission mechanism.
AI is reshaping drug discovery economics by improving success rates and reducing development time and cost, potentially enabling pharma companies to earn "valuation credit" like software businesses and spawning a wave of AI-native biotech startups building discovery platforms or their own therapeutics.
Inference providers compete in a thin-margin compute commodity today, but the few that execute AWS's playbook—cutting prices on raw GPUs while climbing the stack to sell higher-value products like fine-tuning and custom model serving—could build durable, high-margin businesses with real customer lock-in.
Tomasz Tunguz (33) · Lightspeed Venture Partners (31) · Saanya Ojha (23) · Newcomer (14) · Investing in AI (13) · Ed Sim (13) · A16Z (13) · Radical VC (10) · USV (8) · Clouded Judgement (8)
Suggest a blog or newsletter for the SandHill index. We track the RSS feed and feature the best pieces in the weekly digest.
The latest theses & analyses from 800+ venture investors, curated into one email. Join 5,000+ investors, founders and LPs.