• D.C.
  • BXL
  • Lagos
  • Riyadh
  • Beijing
  • SG
  • D.C.
  • BXL
  • Lagos
Semafor Logo
  • Riyadh
  • Beijing
  • SG


Investors in Adept AI will be paid back after Amazon hires startup’s top talent

Aug 2, 2024, 1:01pm EDT
techNorth America
Pascal Rossignol/File Photo/Reuters
PostEmailWhatsapp
Title icon

The Scoop

Investors in Adept, an AI startup whose top employees were hired away by Amazon in June, will get their money back even though ownership of the company hasn’t changed, according to people briefed on the matter.

Adept, which still retains about a third of its employees, will receive around $25 million, while investors like Greylock and General Catalyst, who put $414 million into the company, will roughly recoup their investment. Founded by top researchers from OpenAI and Google, the startup was most recently valued at $1 billion.

The terms of Amazon’s mass hiring will be of interest to regulators weighing whether the poaching amounted to an acquisition of Adept, but circumvented merger notification rules since the startup’s owners remain the same. The US Federal Trade Commission has sent a list of questions to Amazon inquiring about the details of its Adept relationship, Reuters first reported and Semafor has confirmed.

AD

The FTC is already probing a similar deal — Microsoft’s March hiring of most of the employees at Inflection AI, including its co-founder, Mustafa Suleyman. Microsoft agreed to pay Inflection $650 million in licensing fees for its technology.

Adept was working on building advanced, general intelligence AI models and agents capable of carrying out autonomous tasks. But its ambitious plans were burdened by the skyrocketing costs of training AI models. Instead of pursuing its effort to make general intelligence breakthroughs, it chose instead to focus on products it can sell to businesses today, the company said in a blog post when Amazon made the hires.

The company’s co-founders, who include former OpenAI Vice President of Engineering David Luan, decided they could better pursue those advances inside Amazon, which has a growing team focused on building artificial general intelligence. They brought the bulk of the company’s talented researchers with them.

AD

But unlike a typical “acqui-hire,” where a tech company buys a startup in order to get sought-after employees, Amazon didn’t actually buy Adept and didn’t directly pay investors.

Adept’s co-founders didn’t want their backers to lose money, so they arranged for them to be made whole using funds paid by Amazon. The process was overseen by Luan, according to some of the people familiar with the matter. He did not return requests for comment.

Amazon said in a previous statement that it doesn’t consider it an acquisition because it wasn’t interested in owning Adept’s business and technology.

AD

It is difficult to put a value on AI models themselves. Because the technology is changing so fast, current leading models are likely to quickly become obsolete. In order to build AGI, Amazon will need to build models much larger, more expensive and likely more technologically advanced than what Adept has already created.

Title icon

Know More

Silicon Valley has bristled at the aggressive application of antitrust laws by the FTC under Chair Lina Khan.

With acquisitions of startups by larger tech companies under a microscope, deals have tapered off significantly, reducing returns for investors and making some startup business models less attractive to fund.

That scrutiny has become an election issue, with some Silicon Valley supporters of the Democratic Party hoping that a Kamala Harris administration would appoint an FTC chair with a more lenient approach to deals.

Title icon

Reed’s view

Regulators are wary of the Amazon-Adept model of hiring a startup’s talent, because they could be acquisitions pretending to be something else.

But there is a plausible argument that they should be taken at face value.

Adept had three assets: its people, its enterprise sales business, and its AI models. Amazon only wanted the first. And it could have simply hired away the top employees and paid them hefty bonuses.

But it would’ve been difficult for those workers, who may want to start new companies one day, to leave their venture capitalists backers high and dry. So making them whole through payments like the one Amazon made is a way to address that concern.

This kind of mass poaching, or non-acquihires, makes big technology companies more powerful. But the alternative is either discouraging employees from working where they want or creating an environment where venture capitalists are more cautious about deploying capital. Neither of those outcomes serve to increase innovation.

Investors in Adept also aren’t happy about this outcome. They need big returns, not refunds. But it’s better than losing their entire investment, which could have happened under other circumstances. They won’t complain because being a squeaky wheel in Silicon Valley is bad for deal flow.

And Adept could still build a business around enterprise sales and deploying open-source and other AI models. That new business deserves a smaller valuation, but it may be worth trying.

Another factor is that the pursuit of AGI takes such enormous amounts of capital that it can only be attempted by the biggest companies in the world, like Amazon, Microsoft, OpenAI, and Google.

It may one day prove too large even for those companies, and governments will need to get involved.

But if we take a step back and look at the landscape, big tech companies could be seen as actually less powerful than they were before the pandemic. The ChatGPT moment put them off balance, forcing them to invest billions on AI models that have no guarantee of succeeding.

The argument some regulators make is that these AI models might become so powerful and valuable that Big Tech will become even more entrenched. But trying to guess where technology is going and regulating based on those predictions seems impossible given how quickly things are changing. Even the smartest AI researchers don’t know how it will all unfold.

There is also the very real possibility that the models they are trying to build won’t continue to scale and capabilities will begin to taper off. In that case, Big Tech could also face real disruption from startups offering smaller, leaner AI models based on open-source technology.

Title icon

Room for Disagreement

Last year, the FTC laid out its thinking on the AI landscape: “Firms hoping to compete in the generative AI space need expertise, not only on how to develop generative AI but also on how to deploy the fine-tuned AI products. Companies that can acquire both the engineering experience and professional talent necessary to build and package the final generative AI product or service will be better positioned to gain market share.

Since requisite engineering talent is scarce, powerful companies may be incentivized to lock-in workers and thereby stifle competition from actual or would-be rivals. To ensure a competitive and innovative marketplace, it is critical that talented individuals with innovative ideas be permitted to move freely, and, crucially, not be hindered by non-competes.”

AD