Skip to content
OpenAIOpenAI

State of the AI Industry — the OpenAI Podcast Ep. 12

OpenAI CFO Sarah Friar and Khosla Ventures founder Vinod Khosla argue the greatest challenges in AI right now are keeping up with demand and making sure more people get the benefit. They unpack what's driving big investments in compute and why this moment is different from other technology cycles — with meaningful advances in health, agents, and robotics still ahead. Chapters 00:00:00 — What’s the AI story of 2026? 00:07:28 — AI in healthcare 00:12:01 — Scaling compute to match revenue 00:18:05 — Difference between now and dot-com bubble 00:27:41 — Ads in ChatGPT 00:30:05 — Will consumers have more than one AI subscription? 00:36:41 — Winning in enterprise 00:39:44 — How can startups succeed? 00:44:05 — Robotics and beyond

Andrew MaynehostSarah FriarguestVinod Khoslaguest
Jan 18, 202649mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

AI in 2026: agents mature, compute scales, new business models emerge

  1. Vinod Khosla and OpenAI CFO Sarah Friar argue that 2026 is about closing the “capability gap”: moving from chatbot Q&A toward agentic systems that complete real tasks for consumers and enterprises.
  2. They frame demand for AI as fundamentally compute-limited today, with strong price elasticity and expanding use cases—making “bubble” narratives less meaningful than actual usage (e.g., API calls).
  3. Healthcare is highlighted as a major, high-stakes domain already seeing massive usage, but held back by regulation and institutional constraints, even as clinicians increasingly adopt AI tools.
  4. They outline OpenAI’s strategy to scale compute in line with revenue, expand product surfaces and business models (subscriptions, enterprise SaaS, credits, commerce/ads, licensing), and explain where startups can still build durable value—especially around proprietary data, workflows, and permissioning/identity for agents.

IDEAS WORTH REMEMBERING

5 ideas

2026 is positioned as the year agents become visibly useful.

They predict maturation of multi-agent systems that can execute end-to-end tasks (e.g., enterprise reconciliation/ERP workflows, consumer travel planning), shifting AI from “answers” to “outcomes.”

Most users are barely tapping AI’s current capabilities.

Khosla estimates only a single-digit percentage of users utilize even ~30% of what AI can do; the next decade is framed as a learning/adoption journey, not just a model-improvement story.

Healthcare impact is already large, but regulation is the bottleneck.

Friar cites 230M weekly health questions and 66% of US physicians using ChatGPT; Khosla argues prescriptions/diagnosis face FDA/AMA constraints even if AI performance is strong.

OpenAI treats compute as a direct driver of revenue and product velocity.

Friar describes a tight compute-to-ARR relationship (200MW→$2B ARR; 600MW→$6B; 2GW→$20B+) and emphasizes being compute-constrained today while needing to commit years ahead for data center capacity.

“Bubble” should be measured by real usage, not valuations.

Khosla argues stock/valuation swings reflect investor psychology; the meaningful metric is underlying demand (e.g., number of API calls), analogous to internet traffic continuing through the dot-com crash.

WORDS WORTH SAVING

5 quotes

“We’ve handed them the keys to the Ferrari, but they are only learning how to take it out on the road for the first time.”

— Sarah Friar

“Demand is limited not by anything other than availability of compute today.”

— Vinod Khosla

“I always look at bubbles should be measured by the number of API calls.”

— Vinod Khosla

“It’s like we’ve just turned electricity on in the home.”

— Sarah Friar

“Sometime probably towards the end of the next decade, you’ll see a massively deflationary economy, because labor will be near free.”

— Vinod Khosla

2026 outlook: multi-agent systems, robotics, memory, fewer hallucinationsAdoption curves vs capability curvesHealthcare: consumer use, physician adoption, regulationCompute scaling: megawatts-to-ARR correlation, long lead timesBubble debate reframed: API calls vs stock pricesMonetization: tiers, credits, commerce/ads, licensing alignmentEnterprise adoption: vertical specialization, agents, transformational projectsStartup moats: proprietary data, workflows, permissioning/identityRobotics and deflationary economy implications

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome