The 2026 AI Forecast: Foundation Models, IPOs, and Robotics with Sarah Guo and Elad Gil

Discussion of unprecedented AI adoption rates among historically slow-to-adopt professions like physicians, lawyers, and accountants. Despite bubble concerns, professionals who traditionally resist technology are embracing AI tools for documentation, clinical decision support, and reasoning with unstructured data at remarkable speed.

• •Physicians showing massive enthusiasm for AI clinical decision support tools like Abridge and Open Evidence
• •Lawyers, accountants, and compliance professionals adopting AI faster than any previous technology wave
• •Professions requiring reasoning with unstructured data are finding AI particularly valuable
• •Adoption speed contradicts market sentiment about AI being overhyped or not delivering value

Prediction that 2026 will see breakthrough foundation models for scientific domains including physics, materials science, and mathematics. Elad forecasts specific wins like new material inventions or mathematical conjectures being proved, triggering overstated hype cycles that will later prove understated in long-term impact.

• •Next generation of foundation models will target physics, materials science, and mathematical reasoning
• •Expect 1-2 high-profile wins like new material discovery or conjecture proofs in 2026
• •Initial breakthroughs will trigger overhyped claims that 'science is solved'
• •Long-term trend will be understated - these models will eventually transform scientific research

Sarah predicts sentiment collapse around robotics companies as deployment timelines meet reality, while Elad draws parallels to self-driving's 15+ year journey. They debate whether Tesla's Optimus and Waymo will dominate like they did in autonomous vehicles, or if startups can overcome capital and supply chain advantages of incumbents.

• •Small-scale deployment of humanoid robots expected in 2026, but not everything will work perfectly
• •Hype cycle will trigger negative sentiment when early deployments show limitations
• •Self-driving took 15+ years with incumbents (Waymo, Tesla) ultimately winning - robotics may follow similar pattern
• •Startups face structural disadvantages: capital requirements, manufacturing expertise, and supply chain access
• •Chinese robotics companies likely to be major global winners alongside Tesla

Analysis of the coming AI IPO wave, with Sarah sharing insights from a major tech hedge fund manager on why they must buy AI IPOs regardless of fundamentals due to retail demand and annual performance benchmarking. Discussion of how foundation model companies could use IPOs as massive fundraising vehicles.

• •Massive retail appetite to participate in AI beyond NVIDIA will drive IPO pops
• •Hedge funds face game theory pressure to buy AI IPOs due to annual benchmarking and retail demand
• •One successful AI IPO will trigger wave of followers going public
• •IPOs provide foundation model labs with alternative massive fundraising mechanism
• •Fund manager admits buying strategy independent of fundamental company analysis

Deep dive into why consumer AI products have lagged despite technical capabilities. Sarah predicts breakthrough consumer agent software in 2026, while they explore structural barriers including incumbent fear, lack of creative ambition, and scarcity of exceptional consumer product talent (estimated at only a few hundred globally).

• •New wave of consumer agent software coming in 2026 with truly differentiated experiences
• •Incumbents (Google, OpenAI) can copy and distribute fast, making defensibility critical
• •First instinct of founders has been building 'better versions' of old experiences vs. truly novel products
• •Requires either deep research proximity or creative ambition to build something different
• •Estimated only a few hundred exceptional consumer product people globally with right context
• •Previous generation saw ideas get absorbed into platforms, creating founder hesitation

Discussion of the surprising funding of 3-8 new AI research labs in 2024 and competing approaches to AI progress. Covers Ilya's 'age of research' thesis about testing ideas at scale, alternative architectures like diffusion and SSMs, and the resource allocation challenge between inference and research for existing labs.

• •Multiple neo labs funded to explore alternative approaches: diffusion, SSMs, continual learning, self-improvement
• •Ilya's thesis: there's a compute floor (not infinite) where secret ideas can be tested at scale
• •Question whether multiple architectures will prove valuable at different domains vs. single approach winning
• •Existing labs face resource tradeoff: inference generates revenue but consumes compute needed for research
• •Capital aggregation will eventually consolidate to handful of players as scale requirements increase

Prediction that defense tech and drone-based autonomous systems will accelerate dramatically in 2026, driven by Trump administration approach and startup density. Elad argues this represents a fundamental reworking of warfare that's currently in a hype cycle but ultimately underthought given its massive scale.

• •Defense tech and autonomous drone systems will see major acceleration in 2026
• •Trump administration and new Secretary of Defense driving policy support
• •Sufficient density of defense startups now exists to create momentum
• •Represents fundamental shift in how warfare is conducted
• •Challenge: budget must shift from prime contractors to new companies at scale

Sarah's non-AI prediction that GLP-1 drugs remain underrated despite enthusiasm, with second-order effects including opening pathways for other peptide/hormone therapies. The biohacking community's peptide experimentation serves as early indicator of broader societal adoption, similar to how bodybuilders pioneered supplements.

• •GLP-1 adoption will continue to be inexorable despite already high enthusiasm
• •Success creates pathway for other engineered peptides and hormone therapies
• •Demonstrates that delivery mechanisms matter for effective medicines
• •Biohacking community's peptide use (Dubai injections, etc.) signals future mainstream adoption
• •Pattern mirrors how bodybuilding community pioneered creatine and other supplements
• •Will fuel significantly more investment in similar therapeutic opportunities

Compilation of predictions from leading AI founders and researchers including breakthroughs in reasoning systems across all industries, proactive AI that completes tasks before being asked, context-aware products, faster inference enabling new experiences, mass-scale consumer agentic AI, and enterprise AI agents in vertical domains.

• •Reasoning systems will revolutionize biology, self-driving, robotics beyond just language models
• •AI will shift from reactive to proactive - completing tasks before users ask
• •Context extraction will become critical product feature, reducing user burden
• •Faster inference will unlock entirely new product experiences
• •Mass-scale consumer agentic AI will emerge - interface breakthrough comparable to ChatGPT
• •Enterprise AI agents will focus on deep vertical/domain-specific areas with proper workflow integration
• •American open models will regain leadership from China at frontier scale
• •Energy-efficient AI becomes critical as compute demand outstrips grid capacity