RPT-BREAKINGVIEWS-AI lottery-ticket scramble sparks an R&D spiral

By Robert Cyran

NEW YORK, July 24 (Reuters Breakingviews) - The artificial intelligence race will be won with money. Technology giants frequently tout their vast and rising spending on chip-stuffed data centers, the silicon brains powering chatbots, with Meta Platforms META.O alone set to lay out $60 billion in capital expenditures this year. Yet research and development budgets are nearly as large, due to top $50 billion under boss Mark Zuckerberg, and are growing at the same pace. As the company dangles nine-figure paychecks in front of a limited pool of top-flight researchers, an R&D spiral seems guaranteed.

An all-out battle to land key AI talent is already underway. Meta offered $100 million bonuses for skilled executives, according to OpenAI CEO Sam Altman; start-ups re-hire employees weeks after they leave for rivals; and companies invest billions in firms to hire their employees.

It makes sense to focus as much on flesh and blood as silicon and copper. Throwing more computing power, data and energy at AI has produced steady advances. Yet using entirely new techniques, like allowing a chatbot time to reflect on a response, can result in massive, sudden gains in efficiency and capability. China’s DeepSeek sent Wall Street scrambling by pioneering such novelties, requiring just a tenth of the computing time that other cutting-edge models needed for training.

Given the phenomenal wealth promised by cracking human-level intelligence, spending vast sums on researchers becomes akin to buying a very expensive lottery ticket. Tech firms and their venture capitalist backers have won similar, past gambles. With the ultimate payoff even hazier and the competition now more intense, the incentive to overpay is huge.

After all, while previously unthinkable salaries will lure in bright students, the supply of skilled workers cannot expand quickly. Arduous training takes time, and novel insight is rare.

For proof, look to R&D spending. U.S. companies generally must expense researchers’ salaries, equity awards and the costs of running labs to develop new products in the year they occur.

Granted, that includes other operational costs beyond pay. Companies R&D expenditures are also difficult to compare directly: Microsoft, through its partnership with OpenAI, has effectively outsourced some research. Existing businesses still require investment. However, the timing of the increases and company statements link most of the rise to AI.

Meta, for example, said in its fourth quarter call that R&D accounted for 90% of added headcount, after Zuckerberg excitedly described the company’s AI developments and ambitions. Amazon.com CEO Andy Jassy said in a letter that the company had 1,000 generative AI services and applications in progress or built. OpenAI is expanding furiously. As long as firms are locked in a race for technical supremacy, the stakes will only get higher.

Follow Robert Cyran on Bluesky.