The Morning Download: AI Turns Innovation Flywheel

What's up: Data center demands drives AWS copper deal; OpenAI forges deal with chipmaker Cerebras; California is investigating xAI.

Good morning. The fast pace of innovation in artificial intelligence is putting pressure on other parts of the tech ecosystem to keep up. Sooner or later, all companies will have to accelerate the generation of new ideas.

The advance of AI continues to drive more innovation in the chip sector. OpenAI has struck a multibillion-dollar agreement to buy computing capacity from startup Cerebras Systems, which is backed by Chief Executive Sam Altman, the WSJ reports. Cerebras designs artificial-intelligence chips that it says can run AI models and generate responses faster than industry leader Nvidia, according to the WSJ. (More on that deal, below.)

AI’s impact is rippling far beyond chips, though. As the core models and systems at the heart of AI evolve, everything connected to it must keep pace. For example, the demand for copper used in data center networks has led to breakthroughs in mining, the Wall Street Journal reports. Amazon Web Services secured a two-year supply pact with Rio Tinto’s Nuton venture for copper from an Arizona mine, supporting AI data centers. The project uses bacteria and acid to extract copper from low-grade ore, offering a domestic supply with reduced energy and water use, according to an exclusive report from the WSJ.

Growing copper consumption for data-center construction has helped to offset slower demand in cyclical manufacturing sectors and construction, pushing copper prices to hit record highs this month.

As such innovations make faster and higher-functioning AI more accessible to the market, companies will find new ways to put it to use. The early adopters will innovate, but sooner or later everyone else will be under pressure to keep up with change. The leadership challenge of the day is to rethink organizations so they can keep up with the pace and intensity of technological change.

OpenAI and Cerebras began partnership talks in the fall and signed a term sheet by Thanksgiving, Cerebras CEO Andrew Feldman told the Journal in an interview. Demand for faster AI compute drove the deal, he said.

Sachin Katti, an OpenAI infrastructure executive, said that the company began considering a partnership with Cerebras after its engineers provided feedback that they wanted chips that could run AI applications, specifically for coding, more quickly.

“The biggest predictor of OpenAI revenue is how much compute is there,” Katti said in an interview. “The last two years, consistently, we have tripled compute every year and the revenue has tripled every year.”

Global spending on AI is expected to reach $2.52 trillion this year, according to a Thursday forecast from market research and IT consulting firm Gartner. That’s an increase of 44% year-over-year.

What’s driving the growth in AI spending? Not surprisingly, it’s mostly cash spent by tech vendors and hyperscalers on AI infrastructure like chips and data center hardware, John-David Lovelock, a distinguished VP analyst at Gartner, told The WSJ Leadership Institute.

Enterprises are still seeing high failure rates from their AI projects, Lovelock said, leading to greater skepticism toward the technology.

For businesses, that means most of their IT dollars will go toward AI projects “with more reliable ROIs.” So rather than developing new AI capabilities, enterprises are more likely to ask their existing software providers to bring AI features to them, Lovelock said.

What’s most notable in the numbers is that AI spending will become nearly the totality of IT spending by 2030, Lovelock added. “AI is going to be part of every IT purchase,” he said. “It’s either going to be in the product, it is the product, or it is supporting an AI product.”

— Belle Lin

Meta’s former generative AI chief, Ahmad Al-Dahle, has joined Airbnb as its new chief technology officer, WSJ reports. Al-Dahle left Meta after almost six years in the wake of the company’s recent AI overhaul. Airbnb said the addition of Al-Dahle would help it “shape how people travel, live, and interact with AI in a way that strengthens human connection.”

Ericsson has a proposal to cut around 12% of its workforce, some 1,600 jobs, as it seeks to lower costs and boost competitiveness, WSJ reports  “The mobile network market still shows no growth and inflationary pressures are expected to continue,” a spokesperson said Thursday.]

Business school graduates preparing for McKinsey’s famously grueling tests may need to sharpen their prompt-engineering skills. For the first time, the firm is evaluating how junior recruits work with its internal AI chatbot, Lilli. In a pilot, candidates used Lilli to analyze a case study and refine their conclusions, one person tells the FT.Interviewers assessed how applicants prompted the tool.

Countries worldwide are looking to reduce foreign reliance and hone their own capabilities in a technology that could profoundly affect their economic competitiveness and national security.

But a June contest to develop homegrown AI models hosted by the South Korean government shows, it is easier said than done. Most finalists used some foreign open-source code, including from China. WSJ reports.

Gu-Yeon Wei, an electrical-engineering professor at Harvard University, tells the WSJ that it isn’t realistic to require every single piece of code be written entirely in-house when pursuing AI-model development. “To forgo open-source software,” Wei said, “you’re leaving on the table this huge amount of benefit.”

The race seeks to identify two homegrown winners by 2027 able to achieve 95% or higher parity in performance with leading AI models from the likes of OpenAI or Google.

Outrage over the creation and dissemination of nonconsensual, sexually explicit AI deepfakes by Elon Musk’s companies is leading more entities to take action. 

California announced Wednesday that its attorney general is investigating xAI, maker of the Grok chatbot behind the images. California Gov. Gavin Newsom in a post regarding the action, called the AI company’s decision to allow the functionality, “vile.”

Grok began allowing users to edit images with text prompts in late December. Last week an analysis found that Grok generated about 7,750 sexually suggestive or nudifying images an hour, WSJ reports.

Earlier this week the U.K.’s communications watchdog said it launched an investigation into Musk’s X social-media platform, adding that the sexualized images of children created by Grok “may amount to child sexual abuse material,” WSJ reports.

As government entities plan a response, businesses may hold the real power to effect change. In early January, xAI launched an enterprise subscription tier for its Grok chatbot, the Information reports.

A brutal crackdown tha