Want to understand the current state of AI? Check out these charts.

Read the full articleWant to understand the current state of AI? Check out these charts. on MIT Tech Review

What Happened

If you’re following AI news, you’re probably getting whiplash. AI is a gold rush. AI is a bubble. AI is taking your job. AI can’t even read a clock. The 2026 AI Index from Stanford University’s Institute for Human-Centered Artificial Intelligence, AI’s annual report card, comes out today and cuts th

Fordel's Take

The Stanford 2026 AI Index dropped today. It's a 400-page annual data compilation tracking benchmark progress, compute costs, adoption rates, and research output across the US, China, and EU. No opinion — just longitudinal data.

The cost charts are the ones that should change decisions. Frontier model inference costs dropped over 99% between 2022 and 2025. If your RAG pipeline routes every query through Claude Opus or GPT-4o, your architecture is priced against 2023 assumptions. Most teams are. That's a real ops budget problem, not a philosophy debate.

Teams running multi-step agents should audit which steps actually need frontier models — classification, chunking, and reranking do not. Greenfield builders can skip this; they default to tiered routing already.

What To Do

Do tiered model routing with Haiku or Flash for classification steps instead of passing every agent subtask to Opus because inference costs have dropped 99% and the delta in output quality for simple routing tasks is negligible.

Builder's Brief

Who

founders and PMs using market data to justify AI product investment to boards or investors

What changes

Stanford AI Index 2026 becomes the citable baseline for capability and adoption claims in fundraising decks and policy submissions

When

weeks

Watch for

Which specific Stanford metrics get cited in congressional testimony or EU AI Act enforcement guidance — signals which numbers will drive regulatory pressure

What Skeptics Say

Annual AI indices from academic institutions systematically undercount energy consumption, failed deployments, and labor displacement costs while over-indexing on capability benchmarks — producing an optimism bias that shapes policy on incomplete data. Stanford's HAI funding sources create structural conflicts with negative findings.

Cited By

MIT Tech Review Want to understand the current state of AI? Check out these charts.