
Editor's Note
On November 19, Anthropic officially listed on Nasdaq (stock code: ANT). The IPO issue price was $83, raising about $81.3 billion and setting a new historical record in the global capital market; the final closing price was $97.17, approximately 17% higher than the issue price, with a fully diluted valuation of $1.77 trillion.
This is an exclusive interview with the founder siblings Amodei on the day of the listing. CNBC renowned host Andrew Ross Sorkin posed a series of sharp market questions: How to fulfill the $900 billion computing power commitment? Will actively “controlling speed” stifle business growth? Will open-source assaults and low-priced tokens level out profits? And how do they maneuver between Nvidia, Musk, and the Pentagon?
Below are selected dialogue excerpts, totaling about 4,700 words, with a reading time of about 8 minutes.
This Issue's Key Insights
Scaling Laws are just a smooth mathematical curve on paper, but when they are realized in physical reality as overwhelming economic productivity, the shock is something we initially didn’t dare to believe ourselves.
Even if we hold self-evolving technology that can push the vehicle speed to 200 mph, we prefer to keep the speed at 90 mph. The extra computing power is used to strengthen safety barriers— in the business marathon, only absolutely reliable models that gain the trust of large global enterprises can dominate the market.
Competition for large models is primarily about quality, not price. The exponential leap in cognition means: once the critical point is crossed, models can leap from writing code to tackling molecular biology and sophisticated embodied intelligence. Commercial value will only concentrate on frontier models.
We signed a long-term computing power agreement worth $900 billion, with confidence based on our cash reserves and existing annual revenue; history has repeatedly proven that heavily investing in models with stronger cognitive abilities yields very high returns, which are ultimately worthwhile.
The frenzy of data centers cannot be shouldered by the general public. We established a strict rule: any extra price differences caused by data center connections to the local power grid will be fully covered by Anthropic out of its own pocket.
I wrote a core principle in my letter to investors, which is also central to how we operate internally: always underpromise and overdeliver.
We must never be constrained by others in our computing foundation. Collaborating with so many partners, from Nvidia, Amazon, and Google to Broadcom, to design the best chips in the world represents a tangible strategic advantage for us.
Scaling Laws Realized: Curves on Paper, Cognitive Explosion in Reality
Q: When you first founded Anthropic, almost no one in the industry understood Scaling Laws better than you. But today, facing a market value starting at “hundreds of billions,” was all of this in your expectation?
Dario Amodei (Co-founder and CEO of Anthropic): Frankly speaking, viewing theoretical extrapolation is one thing, and seeing the reality unfold is another. In the lab, Scaling Laws are merely a coordinate line on draft paper, predicting mathematically that as long as parameters and computing power are stacked, models can unlock corresponding cognitive levels and economic processing capabilities at specific nodes. However, it's easy to draw lines on paper, but witnessing this line evolve into a production impact that sweeps through the real world is a completely different matter. The theory has long predicted the scale this technology should have, yet when it truly materialized, we were still profoundly shocked.
Q: Anthropic has become the largest "Public Benefit Corporation" (PBC) in the capital market. For today's secondary market investors, what does this structure of “writing social welfare into the charter” mean?
Daniella Amodei (Co-founder and President of Anthropic): Our mission is to create something beneficial for everyone. As a PBC, we are legally obligated to maintain an absolute balance between shareholder returns and public benefits. When we, the co-founders, started the company and learned about the concept of a public benefit corporation, we immediately felt this is what we wanted to do. We must permanently embed safety mechanisms and public benefits into the corporate framework. For mature investors, this is not a business burden but a strategic hedge that allows the company to avoid self-destruction in the wave of disruptive technology.
The Philosophy of Controlled Speed: Why Actively Keep It at 90 MPH When You Can Go 200 MPH?
Q: Investors are most concerned about growth, yet you repeatedly emphasize “pacing.” If safety means actively slowing technological evolution, how should the market accept this seemingly anti-commercial setting?
Dario Amodei: This contradiction is objectively present, but as foundational technology becomes increasingly powerful, the entire industry will eventually hit this wall; there’s no escaping it. Our distinction is simply choosing to face it more honestly and earlier.
For example: Imagine you are driving a car at 60 mph, and suddenly you invent an engine that can instantly boost your speed to 200 mph— in the AI field, this is when models begin to acquire the ability for Recursive Self-Improvement. Controlling speed means we know we can rev up to 200 mph, but we choose to drive at 90 mph. We pull out the excess speed and computing power redundancies and focus on solving safety alignment.
In the long run, this is actually a visionary business strategy. The ultimate outcome in the large model race isn’t about who runs the craziest, but who is most worthy of global enterprises entrusting their core business. Who can maintain absolute safety at extreme speeds will eat up the largest share of the enterprise-level market.
Q: After today’s listing, what kind of stock price increase do you consider a "success"? A 10% or 20% increase? If the increase is too large, would it rather deviate from your expectations?
Dario Amodei: Regarding fluctuations in secondary market stock prices, compliance requires us to remain silent.
Daniella Amodei: Stock prices are for today, but Anthropic is a typical ultra-long-cycle company. Today is an exciting milestone, but it’s merely a fueling station in a long marathon. Tomorrow morning the sun will rise as usual, and our R&D efforts and bottom-line principles remain unchanged. For true long-term tech investors, the volume fluctuations brought by the listing are far less significant than the evolution of the foundational technology.
Frontier Returns: It’s About Quality, Not Price; Most Value Is Concentrated on the Frontier
Q: The current competitive intensity is unprecedented: open-source models are attacking, OpenAI is desperately lowering Token pricing, and Microsoft is selling your models while pushing its own lower-priced models. Will this round of price wars completely flatten the profit margins of large models?
Dario Amodei: We must identify two underlying business logics here.
First, traditional products compete on price, but the core battlefield for large models is always on the "frontier.” This is a market that consumes quality because the economic value unleashed by cognitive ability leaps is nonlinear. Climbing the cognitive ladder, what models can solve transitions from trivial Q&A to writing underlying architecture and then to automated molecular biology research and advanced robotic control. The business volume created by frontier models at the peak of technology vastly outpaces that of mediocre models. Almost all excess profits will ultimately be siphoned to frontier models.
Second, only by mastering the most hardcore frontier technologies can you possess the ability to downscale to produce lower-priced models. For instance, our Haiku 5.5 has dropped directly to a quarter of the price of the previous generation Haiku 4.5, far cheaper than top-tier Opus and Fable 5 families. The “computing power multiplier” we have developed while sprinting to the frontier large model can be directly reused for lightweight models. Even in the lower-tech frontier market, we can still establish a protective moat with competitively low costs.
Computing Power Bet: With a $900 Billion Long-Term Procurement Commitment, How Can You Ensure There Are No Disasters?
Q: The energy throughput of data centers is nearing the physical limits, even causing severe backlash from local power grids. Facing increasingly stringent energy audits, how can you ensure the infrastructure can continue to expand?
Daniella Amodei: Our underlying logic is that the costs of AI expansion must not fall on the general public. Many local residents are concerned that connecting data centers to the power grid will drive up local electricity prices; in response, Anthropic has made a clear commitment: any price differences arising at the grid level will be covered by us. High-tech companies must ensure that the local population's interests are protected when entering any community; this is the only legitimate source of long-term viability for computing power infrastructure.
Q: But you are burdened with a staggering $900 billion long-term computing power agreement. Will such astronomical capital expenditure drag down cash flow before achieving overall profitability? Is this a phased investment or an irreversible permanent state?
Dario Amodei: First, we need to clarify the cycle: This $900 billion is spread over a very long future contract. Compared to our vast cash reserves and currently explosive annual revenue, this long-term commitment is well within a healthy boundary.
In terms of computing power forecasting, we prefer to adopt a conservative lower-limit strategy, purchasing on the spot market when explosive demand arises. More importantly, historically, we have repeatedly seen that investing heavily to train models with stronger cognitive abilities is ultimately worthwhile. Hence, we are willing to invest this significant computing expenditure.
Q: Including the funds raised from the IPO, you may have about $250 billion in cash. How beneficial is this reserve for obtaining investment-grade credit ratings and expanding low-cost bonds in the future?
Dario Amodei: We will have to see the results. We hope to achieve an investment-grade rating at some future point, and debt is also a tool available to us that we intend to utilize at some point in the future.
Giants’ Competition and Geopolitical Reflection: Nvidia Is Not the Only Wild Card
Q: Regarding the deep safety responsibilities of the models, even Microsoft CEO Satya Nadella recently admitted, “Relying solely on stacking computing power does not necessarily lead to safety alignment,” and there is even a need to establish specific mechanisms to re-train AI. As a company that has always marketed safety, can your enterprise-level defense withstand scrutiny?
Daniella Amodei: The experiences gained from engaging deeply at the enterprise level are invaluable. Anthropic has always borne the ultimate responsibility: ensuring that the foundational pre-training processes are foolproof while equipping models with the most stringent safety safeguards. Through co-creation with enterprise clients, we are more aware than any ivory tower laboratory detached from practical business of the real risks involved in deploying Claude in the real commercial world.
From the very first day of our founding, we have been a “safety-first” company, and this is the entire reason for our independence. The destructive potential of this technology is as real as its commercial miracles; those who can completely seal systemic loopholes will control the lifeline of the enterprise-level market.
Q: The secondary market environment has been extremely harsh in recent weeks, and Jensen Huang and Nvidia are also at the center of public opinion storms. Facing this turmoil, how did you respond to investors' concerns during the roadshow?
Dario Amodei: We are very confident. Technological breakthroughs are measured in years, while market noise is measured in weeks or even days. Those short-term fluctuations cannot shake the competitive barriers we are building.
I emphasized a phrase in my letter to investors, which is a core principle of how we operate internally and something we tell potential hires: always underpromise, and let the results speak for themselves (Underpromise and overdeliver).
Q: But Nvidia is your investor and participated in this IPO, and there are serious differences between you and Jensen Huang’s public stance on semiconductor export controls to China. How do you handle this complex competitive tension? Should Jensen Huang return to the front line?
Dario Amodei: We have worked with Nvidia for many years to build world-class data centers and are extremely close business partners. However, in the mature business world, healthy cooperation does not mean alignment on every public policy issue. Our policy positions are entirely based on our independent extrapolation of global security dynamics and the technological paths for general AI; we stand by our position while fully respecting our partners’ demands. As for Jensen Huang’s personal position, we will not comment, as it is up to Nvidia to decide.
Daniella Amodei: To add, regarding the underlying chip ecosystem, Anthropic’s core moat is “comprehensive multi-sourcing hedging.” We are deeply tied to Nvidia but also customized chips with Amazon, formed deep alliances with Google, and collaborated closely with Broadcom. Not tying our livelihoods to a single giant gives us strong strategic maneuverability in computing architecture.
Q: Speaking of computing power diversification, Musk is also one of your computing power suppliers. There are rumors that after the contract expires, you may reduce the use of some SpaceX computing power and shift elsewhere. Is this true? Additionally, how is your progress with the Pentagon?
Dario Amodei: It's uncertain how things will unfold in the future. The quality of computing power delivered by SpaceX is outstanding, and current engineering collaboration is extremely smooth.
Daniella Amodei: Regarding defense cooperation, there are currently no details we can publicly disclose. However, it is clear that since our inception, Anthropic has maintained an extremely tight and efficient cooperation mechanism with national security agencies. Ensuring that Claude defends technological sovereignty and national security to the highest standards is our core red line, and we expect this cooperation to continue.
Q: Sibling collaboration has created a trillion-dollar technology giant, a rarity in commercial history. Beyond this moment, stripping away all the glory, what does this mean for both of you?
Daniella Amodei: Before entering the market, we both remarked that this is a miracle belonging to the entire team, and we feel extremely proud of the team's contributions. But after the bell rings, we are still the same company and are eager to continue advancing our mission. If there were a slight personal wish after the IPO— it would be that once the storm of fame and fortune subsides, my brother and I can have a bit more time that belongs only to the two of us.
Source: CNBC, Host Andrew Ross Sorkin dialogues with Anthropic Co-founders Dario Amodei and Daniella Amodei on the day of their IPO on November 19, 2026, https://www.cnbc.com/video/2026/11/19/watch-cnbcs-full-interview-with-dario-and-daniela-amodei.html
免责声明:本文章仅代表作者个人观点,不代表本平台的立场和观点。本文章仅供信息分享,不构成对任何人的任何投资建议。用户与作者之间的任何争议,与本平台无关。如网页中刊载的文章或图片涉及侵权,请提供相关的权利证明和身份证明发送邮件到support@aicoin.com,本平台相关工作人员将会进行核查。