Tech giants pay talent millions of dollars

The artificial intelligence arms race is heating up, and as tech giants scramble to come out on top, they’re dangling millions of dollars in front of a small talent pool of specialists in what’s become known as the AI talent war.

It’s seeing Big Tech firms like Meta, Microsoft, and Google compete for top AI researchers in an effort to bolster their artificial intelligence divisions and dominate the multibillion-dollar market.

Meta CEO Mark Zuckerberg recently embarked on an expensive hiring spree to beef up the company’s new AI Superintelligence Labs. This included poaching Scale AI co-founder Alexander Wang as part of a $14 billion investment into the startup.

OpenAI’s Chief Executive Sam Altman, meanwhile, recently said the Meta CEO had tried to tempt top OpenAI talent with $100 million signing bonuses and even higher compensation packages.

Google is also a player in the talent war, tempting Varun Mohan, co-founder and CEO of artificial intelligence coding startup Windsurf, to join Google DeepMind in a $2.4 billion deal. Microsoft AI, meanwhile, has quietly hired two dozen Google DeepMind employees.

“In the software engineering space, there was an intense competition for talent even 15 years ago, but as artificial intelligence became more and more capable, the researchers and engineers that are specialized in this area has stayed relatively stable,” Alexandru Voica, head of corporate affairs and policy at AI video platform Synthesia, told CNBC Make It.

“You have this supply and demand situation where the demand now has skyrocketed, but the supply has been relatively constant, and as a result, there’s the [wage] inflation,” Voica, a former Meta employee and currently a consultant at the Mohamed bin Zayed University of Artificial Intelligence, added.

Voica said the multi-million dollar compensation packages are a phenomenon the industry has “never seen before.”

Here’s what’s behind the AI talent war:

The inflated salaries for specialists come hand-in-hand with the billion-dollar price tags of building AI models — the technology behind your favorite AI products like ChatGPT.

There are different types of AI companies. Some, like Synthesia, Cohere, Replika, and Lovable, build products; others, including OpenAI, Anthropic, Google, and Meta, build and train large language models.

“There’s only a handful of companies that can afford to build those types of models,” Voica said. “It’s very capital-intensive. You need to spend billions of dollars, and not a lot of companies have billions of dollars to spend on building a model. And as a result, those companies, the way they approach this is: ‘If I’m going to spend a billion dollars to build a model, $10 million for an engineer is a relatively low investment.'”

Anthropic’s CEO Dario Amodei told Time Magazine in 2024 that he expected the cost of training frontier AI models to be $1 billion that year.

Stanford University’s AI Institute recently produced a report that showed the estimated cost of building select AI models between 2019 and 2024. OpenAI’s GPT-4 cost $79 million to build in 2023, for example, while Google’s Gemini 1.0 Ultra was $192 million. Meta’s Llama 3.1-405B cost $170 million to build in 2024.

“Companies that build products pay to use these existing models and build on top of them, so the capital expenditure is lower and there isn’t as much pressure to burn money,” Voica said. “The space where things are very hot in terms of salaries are the companies that are building models.”

Machine learning engineers are the AI professionals who can build and train these large language models — and demand for them is high on both sides of the Atlantic, Ben Litvinoff, associate director at technology recruitment company Robert Walters, said.

“There’s definitely a heavy increase in demand with regards to both AI-focused analytics and machine learning in particular, so people working with large language models and people deploying more advanced either GPT-backed or more advanced AI-driven technologies or solutions,” Litvinoff explained.

This includes a “slim talent pool” of experienced specialists who have worked in the industry for years, he said, as well as AI research scientists who have completed PhDs at the top five or six universities in the world and are being snapped up by tech giants upon graduating.

It’s leading to mega pay packets, with Zuckerberg reportedly offering $250 million to a 24-year-old AI genius Matt Deitke, who dropped out of a computer science doctoral program at the University of Washington.

Meta directed CNBC to Zuckerberg’s comments to The Information, where the Facebook founder said there’s an “absolute premium” for top talent.

“A lot of the specifics that have been reported aren’t accurate by themselves. But it is a very hot market. I mean, as you know, and there’s a small number of researchers, which are the best, who are in demand by all of the different labs,” Zuckerberg told the tech publication.

“The amount that is being spent to recruit the people is actually still quite small compared to the overall investment and all when you talk about super intelligence.”

Litvinoff estimated that, in the London market, machine learning engineers and principal engineers are currently earning six-figure salaries ranging from £140,000 to £300,000 for more senior roles, on average.

In the U.S., the average salary for a machine learning engineer is $175,000, reaching nearly $300,000 at the higher end, according to Indeed.

As tech giants continue to guzzle up the best minds in AI with the lure of mammoth salaries, there’s a risk that startups get left behind.

“Some of these startups that are trying to compete in this space of building models, it’s hard to see a way forward for them, because they’re stuck in the space of: the models are very expensive to build, but the companies that are buying those models, I don’t know if they can afford to pay the prices that cover the cost of building the model,” Voica noted.

Mark Miller, founder and CEO of Insurevision.ai, recently told Startups Magazine that this talent war was also creating a “massive opportunity gap” in traditional industries.

“Entire industries like insurance, healthcare, and logistics can’t compete on salary. They need innovation but can’t access the talent,” Miller said. “The current situation is absolutely unsustainable. You can’t have one industry hoarding all the talent while others wither.”

Voica said AI professionals will have to make a choice. While some will take Big Tech’s higher salaries and bureaucracy, others will lean towards startups, where salaries are lower, but staff have more ownership and impact.

“In a large company, you’re essentially a cog in a machine, whereas in a startup, you can have a lot of influence. You can have a lot of impact through your work, and you feel that impact,” Voica said.

Until the price of building AI models comes down, however, the high salaries for AI talent are likely to remain.

“As long as companies will have to spend billions of dollars to build the model, they will spend tens of millions, or hundreds of millions, to hire engineers to build those models,” Voica added.

“If all of a sudden tomorrow, the cost to build those models decreases by 10 times, the salaries I would expect would come down as well.”