World haven't seen Dario Amodei that energized, loud and even a little aggressive. What made him abandon the usual zen state of his genius mind on the scene of NY Times DealBook Summit yesterday?

First, he is not happy with the conversation about "AI bubble"

When Andrew Ross Sorkin started their dialog on this note, Dario was quick to say that Anthropic AI business is strong and growing at a crazy rate while his competitor—no names named, although there is none other than OpenAI—is led by a person who is happy to over-hype things. Exact quote: "I think there are some players who are not managing that risk well, who are taking unwise risks." Pressed to name names, he refused, but the target was obvious to everyone in the room.

This wasn't the measured, diplomatic Dario of past interviews. For the first time, Amodei laid out in granular detail why Anthropic approach to the AI investment boom isn't just different from its rivals, it's the only one that pencils out without catastrophic risk.

Anthropic revenue grew from zero to 100 million in 2023, then to one billion in 2024, and is projected to land between eight and ten billion by the end of this year. That's 10x growth, year after year, for three consecutive years.

But Amodei immediately undercut any triumphalism. When Sorkin suggested extrapolating to 100 billion next year, Amodei shut it down: "I don't believe that. Just to be clear, I don't believe that at all."

This is where his argument became genuinely novel. Every AI company faces what Amodei calls "the cone of uncertainty". You have to decide now how much compute to buy for early 2027, when that infrastructure comes online. But you don't know if your revenue then will be 20 billion or 50 billion. The lag time between ordering data centers and serving customers creates an unavoidable timing mismatch.

"If I don't buy enough compute, I won't be able to serve all the customers I want. I'll have to turn them away and send them to my competitors," Amodei explained. "If I buy too much compute, of course, I might not get enough revenue to pay for that compute. And in the extreme case, there's kind of the risk of going bankrupt."

What separates Anthropic from the companies that push the risk dial too far, betting everything on the most optimistic scenarios -- is how they manage this cone. "We basically say we want to buy enough compute that we're confident, even in the 10th percentile scenario, might be in a bad position, but we think we can pay for it." They're planning for the pessimistic case, not the optimistic one. While others might be banking on that 100 billion revenue fantasy, Anthropic is buying compute they can afford even if growth slows dramatically.

The second technical insight concerns chip depreciation. When Sorkin asked about the depreciation schedule, Amodei reframed the entire question. "When new chips come out, the issue isn't the lifetime of the chips. Chips keep working for a long time. The issue is new chips come out that are faster and cheaper."

This matters enormously. Capital doesn't evaporate. It just yields less competitive advantage over time. And crucially, "we assume very aggressive kind of continuation of the chip efficiency curve." They're already pricing in the competition. The risk isn't sudden obsolescence but gradual depreciation, which can be modeled and planned for.

Together, these points form an argument that massive AI investment can be managed responsibly: the uncertainty is quantifiable, the planning can be conservative, and the depreciation is gradual rather than catastrophic.

Dario also explained why he thinks his corporate customers are not going to switch his Claude model for other models: "Companies have great difficulty switching from one model to another because they have downstream customers who use the model. And these customers like the current model, they prompt and interact with the models in certain ways, as models have different personalities. It's actually quite hard to switch." This makes Anthropic's business model robust and stable.

Dario does not believe in "AGI Moment"

Another thing that visibly energizes Dario is talk about AGI. He wasn't ambiguous: "I've never liked these terms: AGI, artificial superintelligence. I don't know what it means. There's an exponential growth in intelligence, just like we had an exponential with Moore's law: chips getting faster and faster until they could do any simple calculation faster than any human. I think the models are just going to get more and more capable at everything."

This challenges the entire narrative structure that dominates AI discourse. Most people imagine AGI as a discrete moment, a threshold we cross when machines suddenly "wake up" or achieve human-level intelligence across all domains. Amodei rejects this entirely. "Every few months we release a new model, gets better at coding, it gets better at science. Now models are routinely winning high school math olympiads. They're moving on to college math olympiads. They're starting to do new mathematics."

What he's describing isn't a race to a finish line but a continuous acceleration. This has profound implications for both investment timing and competitive dynamics. There's no winner-takes-all moment where one company achieves AGI and everyone else becomes irrelevant. Instead, it's steady capability gains across all tasks, which means steady revenue opportunities and predictable scaling economics.

He even offered concrete evidence of this shift happening right now inside Anthropic: "For the first time, I've had internal people at Anthropic say, 'I don't write any code anymore. I don't open up an editor and write code. I just let Claude Code write the first draft and all I do is edit it.'" The transition from "AI assists my coding" to "I edit AI's code" marks a qualitative change in the human-AI workflow, one that arrived gradually, not suddenly.

What is the right approach to AI regulation

Dario wasn't happy when Sorkin tried to push him to comment on Trump's personality, but he had strong words about the dominant Silicon Valley and current administration view on AI regulation.

"Saying that for 10 years we won't regulate that technology, it's like ripping out the steering wheel of a car because you don't need to steer for 10 years."

His detailed explanation revealed the depth of the philosophical split in Silicon Valley. There are those, including influential figures like David Sacks (now the AI czar at the White House) and major venture capital firms like Andreessen Horowitz, who argue AI should be treated like previous technological revolutions—the internet, telecommunications—where markets figured things out and regulation came later, if at all. They've pushed for a ten-year moratorium on AI regulation.

Amodei thinks this view fundamentally misunderstands what's being built. "I think those who are closest to AI don't feel this way. If you poll the actual researchers who work on AI, not investors who invest in some AI application companies, not general tech commentators who think they know something about AI, but the actual people who are building the technology, they're excited about the potential, but they're also worried."

What are they worried about? National security risks. Model alignment. Economic displacement. The concentration of power. These aren't abstract concerns but concrete challenges that need active management as the technology scales.

The irony, Amodei noted, is that he may be "the most optimistic about the upsides" of anyone in the debate. He wrote an essay called "Machines of Loving Grace" predicting AI could extend human lifespan to 150 years within a decade of achieving what he calls "a country of geniuses in a data center". He thinks economic growth could accelerate to five or even ten percent annually, compared to the historical average of around 1.4 percent.

"But nothing that powerful doesn't have a significant number of downsides. And we as a society, as a polity, need to think ahead about those downsides."

He made a point of emphasizing that his concerns about regulation aren't about strangling innovation or protecting Anthropic's market position. In fact, "almost all of the AI regulation that we've supported has exemptions for small players. The main AI bill we supported, SB1047, doesn't even apply at all to startups with under 500 million in revenue."

The accusation from Sacks that Anthropic is engaged in "regulatory capture" particularly seemed to irritate him. "This isn't about any particular administration; this is about policy questions. I mean, going all the way back to 2016, I've written papers about AI, before I even had a company, before there even could have been a plan around anything like regulatory capture."

The warning as strategy

I’d suggest that Dario was energized at DealBook as he sees now a beautiful clarity of purpose. He spent years watching the scaling laws continue to hold true. He's seen his company's revenue grow 10x annually for three years straight. He's watched his engineers stop writing code themselves and start editing AI output instead. And he's concluded that the technology he's building will fundamentally reshape civilization.

That conviction brings responsibility. "I warn about these things not to be a prophet of doom, but because warning about them is the first step towards solving them," he explained. "And if we don't warn about them, then we'll just kind of blindly walk into the landmine and blow us up. If we warn about them, if we see the landmine, we can walk around it and we can avoid it."

This is the throughline connecting all of Dario's uncharacteristic behavior. He's not just competing for market share or defending Anthropic's approach to capital allocation. He's trying to establish that there's a responsible path through the AI transition, one that manages uncertainty conservatively, maintains democratic values, and plans for both the economic disruption and the enormous gains ahead.

The question for everyone watching is whether Silicon Valley and Washington will listen to the person closest to the technology, or whether they'll keep driving with the steering wheel ripped out, hoping the road stays straight.