Sam Altman’s Delhi Warning: Three Board Decisions for US CEOs
“Democratization of AI is the best way to ensure humanity flourishes… centralization of AI in one company or country could lead to significant disruption.” — Sam Altman, India AI Impact Summit 2026
Sam Altman spoke at this week’s India AI Impact Summit. While he was addressing global leaders, the logic applies directly to your boardroom. Here is what his warning means for three decisions you need to make right now.
Decision 1: Which Workloads Stay With One Vendor?
Altman’s warning about the “centralization of AI” applies to companies just as much as countries. If your critical infrastructure relies on a single provider, you are vulnerable.
The Exercise: List your top 10 AI use cases. Star the 3–5 that matter most to your margins or risk profile (e.g., pricing algorithms, customer data processing, or regulatory compliance).
The Stress Test: Ask yourself: If that vendor raised prices 2x or suffered a month-long outage, how catastrophic is the impact?
Your Move:
Diversify: Add a second provider for backup.
Go Local: Test smaller models you can host yourself.
Protection: Secure contracts with cost caps and clear exit clauses.
The Goal: You don’t need to ditch your main vendor; you just can’t bet the farm on them.
Decision 2: Where Does AI Get Better Every Quarter?
“By 2028, data centers may hold more intellectual capacity than all humans.”
AI ages fast. A model that is “state-of-the-art” today will be average by next year. You cannot treat AI as a three-year capital expenditure; it is a rhythm.
The Strategy: Pick 2–3 workflows and commit to a quarterly upgrade cycle:
Sales Proposals: Move from AI drafting/human polishing to AI agents handling routine tasks.
Fraud Detection: Move from simple flags to autonomous blocks.
Contract Review: Move from basic summaries to active negotiation agents.
The Metric: Target a 20–30% reduction in cycle time every 90 days.
Decision 3: Who Becomes 2x More Productive?
Altman argues that democratization ensures humanity flourishes. In a corporate sense, AI empowers the individuals already using it, making them 25–40% more valuable.
The Plan:
Identify the Power Users: Find your top 20%. Audit their secrets: What tools, habits, and prompts are they using?
Scale the Success: Reallocate budget specifically to the teams that are actively changing how they work.
The Target: Aim for average output per person in key roles to hit 1.5x by 2027. This is your primary lever for revenue-per-employee growth.
A Final Note: Use “Cheap” Models for Routine Work
Altman predicts that AI costs will come down dramatically. Currently, 80% of AI work—classification, summarization, and retrieval—doesn’t need the biggest, most expensive model.
Smaller models cost 10x less, keep data private, and are easier to move. The Rule: Use big models for hard problems, and small models for everything else.
CEO Checklist: Three Questions for Your Next Meeting
Dependency: Which three workloads cannot remain single-vendor?
Iteration: Which two workflows must we upgrade every single quarter?
People: What is our specific roadmap to 2x the average worker’s pr