Bill Gates says AI is powerful enough to cause a billion deaths
Gates has spent the past month arguing that AI is a policy problem, not a product category. This week he attached a number to it and asked Washington for the law. Separately, Goldman Sachs put a revenue figure on the AI buildout that its spending implies.
Bill Gates told NBC's "Meet the Press" that AI is "certainly powerful enough to drive events that, you know, cause a billion deaths" — and that the companies building it should not be left to police themselves. The line came with a mechanism, not just a warning: "There's never been a weapon as powerful as the combination of people with ill intent using the latest AI tools." Asked by moderator Kristen Welker whether Washington needs to pass legislation, Gates said, "Absolutely." He then described what he wants on the books: "You need law enforcement and the politicians to get into the discussion about what safeguards and monitoring look like. And that has to be a required thing. And it will be a little bit of overhead for the industry, but not a dramatic slowing of what they're doing."
The interview was conducted on Wednesday, NBC published it Friday morning, and it airs in full on Sunday. That timing matters less than the company it kept. This month Dario Amodei published a case for pacing the frontier and named regulation as the only method that binds companies that would not cooperate voluntarily; Sam Altman told the United Nations to build standards for measuring capability and preserving human oversight. An Anthropic safety researcher, Jacob Coxon, resigned and told Congress the labs are "gambling with our lives" — while also saying he thinks the people running them are sincere about wanting to slow down.
What Gates adds is the number and the demand. We covered the earlier instalments of this campaign in August — Bill Gates says AI has crossed its danger thresholds — now he wants a plan and Bill Gates proposes 'human-reserved' jobs and a token tax — and neither carried a casualty estimate or an explicit call for mandatory federal monitoring. That is the new material.
The politics around him point the other way, which is what makes the interview worth reading rather than filing. Mark Zuckerberg told NBC this week that he does not think the industry needs "some kind of industrywide coordination" and that each lab should simply take the time it needs internally. Speaker Mike Johnson declined to bring the House back to pass AI legislation, saying he would defer to the companies and that they have not offered a workable path. And at the UN this week the White House argued against governments setting ground rules at all — the position we wrote up in Trump to the UN: no global AI rules, and AI is 'super intelligence'. Microsoft's president, Brad Smith, used the same UN stage to make the opposite case with an analogy: two companies racing to ship the next airliner, one of them worried it will crash.
Read Gates's wording closely, though, because the ask is narrower than the warning. He wants monitoring and safeguards to be required — an audit regime — and he is explicit that it should cost the industry "a little bit of overhead" and not "a dramatic slowing." That is the tell: he is asking for verification, not pacing, and verification is the one thing the labs have largely agreed to in principle and resisted in practice. The gap between those two positions is where the next year of AI policy actually lives.
Goldman Sachs strategists expect the five largest US hyperscalers to spend about $800 billion on AI infrastructure this year — and told clients the spending only breaks even if those companies pull in roughly $300 billion a year in AI revenue. The note, led by Ryan Hammond, puts 2027 spending above $1.2 trillion, ahead of Wall Street's $1.1 trillion consensus, and names the five: Amazon, Alphabet, Microsoft, Oracle and Meta. The growth curve is where the interesting part sits. Capex growth runs at nearly 100% this year, slows to 54% in 2027 and 12% in 2028, when the total reaches about $1.4 trillion. Hammond's team argues that on consensus expectations, capital spending as a share of GDP is set to exceed any technology investment cycle since the late-19th-century railway buildout.
The financing side is the pressure point. Hyperscalers are expected to issue roughly $250 billion in global investment-grade debt by the end of 2026, and consensus estimates compiled by Apollo have their operating cash flow tripling from $600 billion in 2025 to about $2 trillion in 2030 — a forecast, not a result. We worked through the denominator of this this morning — The AI buildout is 3.6% of GDP a year — more than any US boom — and Goldman's contribution is the numerator test the Brookings paper deliberately left open: what the spending has to earn to justify itself. A $300 billion annual revenue requirement is a much harder number to argue with than a capex total, because it has to show up in somebody's income statement.
What to watch: whether any member of Congress moves a monitoring bill before the midterms rather than deferring to the labs, and whether the next two quarters of cloud revenue start closing the gap to the break-even figure.
Gates wants mandatory monitoring, not a slowdown — and the labs have agreed to the principle while resisting the paperwork. Which half of that do you think actually happens first? Tell us in the comments.
Sources: NBC News — Bill Gates says AI companies self-regulating isn't enough · Axios — Gates warns AI could cause a billion deaths · Investing.com — Hyperscalers need $300 billion in annual AI revenue to break even · 财联社 (CLS) — 高盛料明年超大规模云服务商AI资本支出增长逾50% · Yahoo Finance — Big Tech's cash flow tripling to $2 trillion
