Daniela Amodei on Anthropic’s ‘Do More With Less’ Strategy

SAN FRANCISCO – At the heart of Anthropic’s headquarters, President and co-founder Daniela Amodei frequently returns to a guiding principle for the artificial intelligence startup’s overarching strategy: achieve more with less. This philosophy stands in stark contrast to the prevailing sentiment across Silicon Valley, where leading AI labs and their investors are fixated on the idea that scale is the ultimate determinant of success.

Companies are securing unprecedented funding, pre-ordering essential computing hardware years in advance, and constructing massive data centers with the conviction that the entity capable of building the largest AI infrastructure will emerge victorious. OpenAI has become the most prominent example of this approach, having committed to approximately $1.4 trillion in compute and infrastructure resources. This includes establishing vast data center campuses and acquiring next-generation chips at a rate the industry has never witnessed.

Anthropic, however, presents an alternative path through this AI arms race. Their strategy centers on judicious resource allocation, algorithmic efficiency, and intelligent deployment to maintain a leading position without engaging in a direct competition of sheer scale.

“I believe our consistent aim at Anthropic has been to be as resourceful as possible with our available assets, all while operating in a domain that inherently demands substantial compute power,” Amodei told CNBC. “Anthropic has historically operated with a fraction of the compute and capital available to our competitors. Yet, for the better part of the last several years, we have consistently delivered the most powerful and performant models.”

Daniela Amodei, along with her brother and Anthropic’s CEO Dario Amodei – a former researcher at Baidu and Google – were instrumental in shaping the very paradigm they are now challenging. Dario Amodei was among the pioneers who popularized the scaling hypothesis, the concept that progressively increasing compute, data, model size, and complexity leads to predictable improvements in model capabilities. This pattern has, in essence, become the financial underpinning of the current AI investment landscape, justifying massive capital expenditures by hyperscalers, bolstering the valuations of chip manufacturers, and encouraging private markets to assign enormous price tags to companies still heavily investing to reach profitability.

While Anthropic has benefited from this scaling logic, the company is now demonstrating that the next phase of AI competition may not solely be decided by the ability to fund the largest pre-training runs. Their strategy emphasizes the use of higher-quality training data, advanced post-training techniques to enhance reasoning abilities, and product design choices aimed at reducing operational costs and facilitating wider adoption at scale – a critical consideration in the AI sector where compute expenses are a perpetual concern.

It is important to note that Anthropic is not operating with a minimal budget. The company has secured approximately $100 billion in compute commitments and anticipates these requirements will continue to grow as it strives to remain at the forefront of AI development. “The compute demands for the future are substantial,” Daniela Amodei acknowledged. “Our projection is that, yes, we will require more compute to maintain our position at the frontier as we expand.”

Nevertheless, Anthropic contends that the headline figures bandied about in the sector are often not directly comparable, and that the industry’s collective certainty regarding the “optimal” expenditure is less robust than it appears. “Many of the numbers presented are not precisely apples to apples due to the structural intricacies of some of these agreements,” she explained, referencing a market environment where companies feel compelled to make early commitments to secure hardware well in advance.

She further elaborated that even insiders who helped formulate the scaling thesis have been surprised by the sustained compounding of performance and business growth. “We have continued to be surprised, even as those who championed the belief in scaling laws,” Daniela Amodei remarked. “A sentiment I often hear from my colleagues is that the exponential trend continues until it ceases. And every year, we’ve found ourselves thinking, ‘This cannot possibly continue on an exponential trajectory’ – and yet, it has.”

This observation encapsulates both the optimism and the underlying anxiety of the current AI buildout. If the exponential trend persists, companies that secured early access to power, chips, and infrastructure may appear prescient. Conversely, if this trend falters, or if adoption rates lag behind capability advancements, those that overcommitted could find themselves burdened by years of fixed costs and long-lead-time infrastructure built for demand that never materializes.

Daniela Amodei drew a critical distinction between the technological advancement curve and the economic adoption curve, a nuanced point that often becomes blurred in public discourse. From a technological standpoint, she stated that Anthropic has not observed any deceleration in progress, based on their current observations. The more complex challenge lies in how rapidly businesses and consumers can integrate these advanced capabilities into practical workflows, where procurement processes, change management, and human factors can impede even the most sophisticated tools.

“Regardless of the technology’s sophistication, its integration into business or personal contexts takes time,” she emphasized. “The pivotal question for me is: How swiftly can businesses, in particular, but also individuals, effectively leverage this technology?”

This focus on enterprise adoption is a key reason why Anthropic has become a closely watched indicator for the broader generative AI market. The company has strategically positioned itself as an enterprise-first model provider, deriving a significant portion of its revenue from businesses integrating Claude into their workflows, products, and internal systems. This type of usage tends to be more enduring than consumer applications, where churn can increase once the initial novelty wears off.

Anthropic reports a tenfold year-over-year revenue growth for three consecutive years. Furthermore, it has cultivated a distribution network that is distinctive in a market characterized by intense competition. The Claude model is accessible across major cloud platforms, including through partners who are also developing and marketing competing AI models. Daniela Amodei views this multi-cloud presence not as a concession, but as a reflection of customer demand. Large enterprises seek flexibility across different cloud environments, and cloud providers aim to meet the evolving needs of their key clients.

In practice, this multicloud strategy also serves as a method to compete without making a singular, high-stakes infrastructure bet. While OpenAI appears to be anchoring its extensive buildout around bespoke campuses and dedicated capacity, Anthropic is prioritizing flexibility. The company can dynamically shift its operational base based on cost, availability, and customer demand, while concentrating its internal efforts on enhancing model efficiency and performance per unit of compute.

As 2026 commences, this strategic divergence carries significant implications. Both Anthropic and OpenAI are navigating the transition towards public market readiness, albeit while operating within a private market landscape where compute needs are escalating faster than market certainty. Neither company has announced definitive IPO timelines, but both are undertaking actions indicative of preparation, including strengthening finance, governance, forecasting capabilities, and establishing an operational cadence resilient to public scrutiny. Simultaneously, both are actively raising capital and negotiating increasingly large compute agreements to fuel the next stage of model development.

This sets the stage for a genuine test of strategy over rhetoric. If the market continues to favor large-scale investment, OpenAI’s approach may remain the industry benchmark. However, if investors begin to prioritize efficiency, Anthropic’s “do more with less” philosophy could provide a distinct competitive advantage.

In this context, Anthropic’s contrarian bet is not that scaling is ineffective. Rather, it posits that scaling is not the sole decisive factor. The company believes the winner of the next phase may be the AI lab that can achieve continuous improvement while operating within an economic framework that is sustainable in the real economy.

“The exponential trend continues until it doesn’t,” Daniela Amodei concluded. The critical question for 2026 is what will transpire in the AI arms race – and for the companies constructing it – if the industry’s favored curve finally ceases its predictable trajectory.