SAN FRANCISCO — Inside Anthropic headquarters, President and co-founder Daniela Amodei keeps coming back to a phrase that’s become a sort of governing principle for the artificial intelligence startup’s entire strategy: Do more with less.

It’s a direct challenge to the prevailing mood across Silicon Valley, where the biggest labs and their backers are treating scale as destiny.

Firms are raising record sums, locking up chips years in advance, and pouring concrete across the American heartland for data centers in the belief that the company that builds the largest intelligence factory will win.

OpenAI has become the clearest example of that approach.

The company has made roughly $1.4 trillion in headline compute and infrastructure commitments as it works with partners to stand up massive data center campuses and secure next-generation chips at a pace the industry has never seen.

Anthropic’s pitch is that there’s another way through the race, one where disciplined spending, algorithmic efficiency, and smarter deployment can keep you at the frontier without trying to outbuild everyone else.

“I think what we have always aimed to do at Anthropic is be as judicious with the resources that we have while still operating in this space where it’s just a lot of compute,” Amodei told CNBC. “Anthropic has always had a fraction of what our competitors have had in terms of compute and capital, and yet, pretty consistently, we’ve had the most powerful, most performant models for the majority of the past several years.”

Daniela Amodei and her brother, Dario Amodei, who is Anthropic’s CEO and a Baidu and Google alumni, helped build the very worldview they’re now betting against.

Dario Amodei was among the researchers who helped popularize the scaling paradigm that has guided the modern model race. It is the strategy that increasing compute, data, model size, and capabilities tends to improve the model in a predictable way.

That pattern has effectively become the financial bedrock of the AI arms race.

It underwrites hyperscaler capital spending, justifies towering chip valuations, and keeps private markets willing to assign enormous prices to companies that are still spending heavily to reach profitability.

But even as Anthropic has benefited from that logic, the company is trying to prove that the next phase of competition won’t be decided only by who can afford the largest pre-training runs.

Its strategy leans into higher-quality training data, post-training techniques that improve reasoning, and product choices designed to make models cheaper to run and easier to adopt at scale — the part of the AI business where the compute bill never stops.

To be clear, Anthropic isn’t operating on a shoestring. The company has roughly $100 billion in compute commitments, and expects those requirements to keep rising if it wants to stay at the frontier.

“The compute requirements for the future are very large,” Daniela Amodei said. “So our expectation is, yes, we will need more…

Shares of grocery delivery service Instacart dropped about 7% in extended trading on Wednesday, following a report that said the U.S. Federal Trade Commission has begun an investigation into the company’s pricing practices.

The FTC sent a civil investigative demand to Instacart, Reuters reported, citing unnamed people.

“The Federal Trade Commission has a longstanding policy of not commenting on any potential or ongoing investigations,” the FTC told CNBC in a statement. “But, like so many Americans, we are disturbed by what we have read in the press about Instacart’s alleged pricing practice.”

A study released last week showed that prices for the same products in the same supermarkets that work with Instacart can vary by around 7%, which can result in over $1,000 in extra annual costs for customers. Instacart responded by saying that retailers determine prices listed in the app.

In 2022, Instacart spent $59 million to acquire Eversight, a company specializing in artificial intelligence-driven pricing and promotions for retailers and consumer packaged goods. Instacart sought to “create compelling savings opportunities for customers in real-time” with Eversight, according to a regulatory filing.

Instacart declined to comment on the investigation but said reporting has mischaracterized its pricing system.

“Much of what’s been reported has mischaracterized how pricing works on Instacart,” a company spokesperson said in an email. “First, our retail partners control their pricing strategies, and we work with them to align their online and in-store pricing wherever possible. Second, these tests are not dynamic pricing nor surveillance pricing – prices on Instacart do not change in real time nor are they based on supply or demand, and we never use personal, demographic, or user-level behavioral data to set item prices.

“These tests are a form of randomized A/B testing, similar to the way retailers have long run pricing tests between different stores.”

Earlier on Wednesday, Robert Garcia, a U.S. House Democrat from California, sent a letter to Instacart CEO Chris Rogers asking for a report on how the company sets prices.

“It is unconscionable that corporations are adding to Americans’ financial strain with algorithmic — and potentially surveillance — pricing,” wrote Garcia, who is ranking member of the House Committee on Oversight and Government Reform.

— CNBC’s Annie Palmer contributed to this report

Read Reuters’ full report here.

Nvidia has established itself as the undisputed leader in artificial intelligence chips, selling large quantities of silicon to most of the world’s biggest tech companies en route to a $4.5 trillion market cap.

One of Nvidia’s key clients is Google, which has been loading up on the chipmaker’s graphics processing units, or GPUs, to try and keep pace with soaring demand for AI compute power in the cloud.

While there’s no sign that Google will be slowing its purchases of Nvidia GPUs, the internet giant is increasingly showing that it’s not just a buyer of high-powered silicon. It’s also a developer.

On Thursday, Google announced that its most powerful chip yet, called Ironwood, is being made widely available in the coming weeks. It’s the seventh generation of Google’s Tensor Processing Unit, or TPU, the company’s custom silicon that’s been in the works for more than a decade.

TPUs are application-specific integrated circuits, or ASICs, which play a crucial role in AI by providing highly specialized and efficient hardware for particular tasks. Google says Ironwood is designed to handle the heaviest AI workloads, from training large models to powering real-time chatbots and AI agents, and is more than four times faster than its predecessor. AI startup Anthropic plans to use up to 1 million of them to run its Claude model.

For Google, TPUs offer a competitive edge at a time when all the hyperscalers are rushing to build mammoth data centers, and AI processors can’t get manufactured fast enough to meet demand. Other cloud companies are taking a similar approach, but are well behind in their efforts.

Amazon Web Services made its first cloud AI chip, Inferentia, available to customers in 2019, followed by Trainium three years later. Microsoft didn’t announce its first custom AI chip, Maia, until the end of 2023.

“Of the ASIC players, Google’s the only one that’s really deployed this stuff in huge volumes,” said Stacy Rasgon, an analyst covering semiconductors at Bernstein. “For other big players, it takes a long time and a lot of effort and a lot of money. They’re the furthest along among the other hyperscalers.”

Originally trained for internal workloads, Google’s TPUs have been available to cloud customers since 2018. Of late, Nvidia has shown some level of concern. When OpenAI signed its first cloud contract with Google earlier this year, the announcement spurred Nvidia CEO Jensen Huang to initiate further talks with the AI startup and its CEO, Sam Altman, according to reporting by The Wall Street Journal.

Unlike Nvidia, Google isn’t selling its chips as hardware, but rather providing access to TPUs as a service through its cloud, which has emerged as one of the company’s big growth drivers. In its third-quarter earnings report last week, Google parent Alphabet said cloud revenue increased 34% from a year earlier to $15.15 billion, beating analyst estimates. The company ended the quarter with a business…