Anthropic's 'do more with less' bet has kept it at the AI frontier, Daniela Amodei tells CNBC

In the heart of San Francisco, Anthropic's President and co-founder, Daniela Amodei, emphasizes a guiding principle that sets the artificial intelligence startup apart: "Do more with less." This mantra starkly contrasts with the prevailing narrative in Silicon Valley, where larger labs and their investors prioritize scale as the key to success. With companies racing to secure substantial investments, lock in chip supplies, and construct massive data centers, the belief persists that whoever builds the biggest AI infrastructure will dominate the market. OpenAI exemplifies this mindset, having committed approximately $1.4 trillion towards compute and infrastructure as it partners with various entities to establish extensive data center campuses and acquire next-generation chips at unprecedented rates. However, Anthropic proposes a different path, one that advocates for strategic spending, algorithmic efficiency, and intelligent resource allocation to remain at the forefront without the need for extravagant expansion. "At Anthropic, our goal has always been to judiciously utilize our resources while operating in a compute-intensive environment," Amodei shared in a CNBC interview. Despite having significantly less compute power and capital than competitors, Anthropic has consistently developed some of the most advanced models in recent years. Dario Amodei, Daniela's brother and Anthropic's CEO, played a pivotal role in popularizing the scaling concept that most AI companies now follow. This framework suggests that increasing compute, data, and model size typically enhances performance in a predictable manner, forming the financial foundation of the current AI race. This model underpins massive capital expenditures and justifies high valuations for companies still striving for profitability. While Anthropic benefits from this logic, it seeks to demonstrate that future competition will not solely hinge on who can afford the largest pre-training processes. Their strategy prioritizes high-quality training data, advanced post-training methods for improved reasoning, and product designs that reduce operational costs and enhance scalability — crucial elements in a sector where expenses are relentless. It's important to note that Anthropic is not operating on a minimal budget; they have around $100 billion in compute commitments, with expectations for further increases as they strive to maintain their competitive edge. "The compute requirements for the future are substantial," Amodei remarked. "To stay at the forefront, we will need to expand our compute capabilities." Nonetheless, Anthropic contends that many of the figures circulated in the industry lack direct comparability, suggesting that the overwhelming confidence regarding optimal spending is often overstated. Amodei pointed out that the structure of various deals can lead to discrepancies in reported numbers, creating pressure for companies to make early commitments for hardware. Reflecting on the scaling narrative, she noted that even those who pioneered this approach have been taken aback by the consistent growth in performance and business success. "We have been surprised, even as the creators of this scaling belief," she stated, adding, "Every year, we think the exponential growth can't continue — and yet it has." This perspective encapsulates the duality of today's growth environment. If exponential growth persists, companies that secure resources early may appear visionary. Conversely, if the trend falters or adoption fails to keep pace with advancements, those who overcommitted may face significant financial burdens from infrastructure built for unmet demand. Amodei highlighted the distinction between technological advancements and economic realities, an essential nuance often blurred in public discourse. Technologically, Anthropic does not foresee any slowdown in progress, but the challenge lies in how swiftly businesses and consumers can adapt these innovations into practical applications. "Regardless of technology quality, it takes time for businesses and individuals to effectively utilize it," she explained. Anthropic's focus on enterprise solutions underpins its significance as a critical player in the generative AI landscape, with substantial revenue derived from companies integrating Claude into their operations, a model that tends to foster long-term customer retention compared to consumer applications. Revenue has reportedly surged tenfold year-over-year for three consecutive years, and Anthropic has established a unique distribution footprint in a competitive market. The Claude model is accessible across major cloud platforms, including through partnerships with competitors, reflecting a market demand for flexibility. As both Anthropic and OpenAI approach potential public market readiness while navigating a rapidly evolving landscape, their differing strategies will be put to the test. While OpenAI continues to pursue a massive infrastructure buildout, Anthropic remains agile, adjusting operations according to costs, availability, and customer needs, all while concentrating on enhancing model efficiency. As 2026 approaches, the implications of these contrasting approaches will be pivotal. The outcome of the AI arms race may depend on whether the market continues to favor expansive growth or shifts toward demanding efficiency, positioning Anthropic's approach as a potential advantage. Amodei's assertion that "the exponential continues until it doesn't" raises a crucial question: What will happen if the industry's favored growth model falters?