In 2016, Seattle-based startup Turi was helping almost 100 customers create and manage software that uses machine learning, a powerful type of artificial intelligence. Its technology was so promising that Apple snapped it up for $200 million.
The deal was a triumph for investors and founders, but one backer thought Turi — and the broader tech industry — might be better off if the startup had spurned Apple’s advances. Matt McIlwain, managing director at Madrona Venture Group, said it’s important that at least some emerging tech businesses remain independent, rather than falling into the arms of Apple, Amazon, Facebook, Google or Microsoft.
“It is economically beneficial to society to have more stand-alone, independent companies. We generally think that’s better than just having these companies consolidated into larger ones,” McIlwain said. “There have to be some in each era that are willing to take the risk and stand the test of time as independent companies so we have the next generation of leading platforms.”
Regulators and lawmakers are investigating whether the largest U.S. technology companies have become too powerful. Acquisitions are a major part of the probes, with the Federal Trade Commission re-examining hundreds of small deals from the previous decade made by Apple, Amazon, Facebook, Google and Microsoft.
Artificial intelligence is a prime example of an important technology field where the majority of promising startups have been gobbled up, frequently ending public availability of any products they created. Last year saw a record 231 acquisitions of AI startups, up from 42 in 2014, according to data from CB Insights. Apple has been the top acquirer since 2010, followed by Google, Microsoft, Facebook, Intel and Amazon.
“If big tech companies buy them all up, they eliminate these future competitors, and have a chance of actually owning the winners,” said Sean Gourley, chief executive officer of machine-learning startup Primer AI. “It’s a real shame. We lost something. There may have been different approaches in this field, but now we only have what these larger companies decide.” Bloomberg Beta, the venture capital arm of Bloomberg LP, is an investor in Primer AI.
It’s particularly important to have a broad, diverse community developing AI because the technology is informing more decisions and has been susceptible to bias, according to researchers. Representatives for Turi and Apple declined to comment, as did Amazon, Facebook, Google and Microsoft.
Apart from consolidating promising technology, many of the acquisitions were done to amass talent. There’s a shortage of workers with experience in deep learning and machine learning, and many companies, not just in tech, are competing for these specialists. Google’s 2013 purchase of DNNResearch brought deep-learning godfather Geoffrey Hinton to the internet giant. In 2014, Google also snapped up DeepMind, a London research outfit led by Demis Hassabis that used software to beat the best players at the strategy game Go and is working on health applications for AI.
The industry’s deal spree has raised concern that the biggest technology companies have a lock on brain power in a field that’s considered critical to the future of computing, global competitiveness and even military supremacy.
“It’s the most important technology we will see in our lifetime,” said Diego Oppenheimer, CEO of Seattle-based Algorthimia, which provides a marketplace for machine-learning algorithms. “When you look at it that way, if it is concentrated on the few, it’s going to be really hard to compete with those few.”
At least some of the motivation behind the acquisition of smaller AI firms has to be more than talent hoarding, according to Frederic Laurin, partnership director at Mila, a prominent Montreal-based deep learning research lab. “The other potential explanation is they see those firms as competitors,” he said.
For startups, selling can be the best option when promising technology is failing to become a real business. Joining a larger company can provide a bigger audience for a startup’s ideas, along with more resources to develop products quicker. Giants such as Google and Facebook have massive data sets that are crucial for training AI models, for instance.
“It’s easy to get funding for an AI startup, or at least it was. Easier than translating an idea into revenue, so the exit often becomes an acquisition,” said Babak Hodjat, who invented some of the technology that became Apple’s Siri digital assistant and sold parts of his AI company Sentient Technologies to outsourcing giant Cognizant Technology Solutions Corp. last year.
Before Apple bought Turi, venture capital firm Opus Capital backed the startup because it had an “awesome product and a very technically sharp team,” said Preeti Rathi, who was at Opus at the time. But the market wasn’t ready for the technology, she said.
“When a good team meets a not-ready market, it takes a lot of capital,” said Rathi, who is now a general partner at Icon Ventures. “Startups don’t typically have that much cash,” so the Apple sale was a good outcome for Turi, she added.
One new reason AI startups look to be acquired by a big patron is the cost of computing power. As AI models get more complex, startups are paying a lot for cloud computing services to train and run those models. And they often pay their rivals, Amazon, Microsoft and Google, which are the dominant cloud providers. This is pricing smaller firms out of the space, said Laurin of Mila, the Montreal-based deep-learning lab.
Even Mila, with 450 researchers and support from several universities, can’t keep up sometimes, Laurin said. Yoshua Bengio, a deep-learning pioneer and scientific director at Mila, has told Laurin that there are research papers from Google and Facebook that Mila can’t replicate because it doesn’t have access to the same computing power.
In 2015, Elon Musk, Sam Altman and other technologists helped start non-profit research group OpenAI with a $1 billion commitment because they were concerned about big tech companies dominating such an important technology. Now OpenAI has started a for-profit arm and taken a $1 billion investment from Microsoft partially to fund the intensive computing needs its work requires.
Still, some promising AI startups have stayed independent, and new ones are being formed all the time. There were 2,235 venture capital AI funding deals last year worth a total of $26.6 billion, according to CB Insights, which also counted 24 AI unicorns — companies valued at $1 billion or more.
“New startups are emerging every day, some of which are building large businesses and will emerge as leaders,” said Rathi. “This is despite the current generation of incumbent competitors — Google, Amazon, etc. — that are quite formidable.”
Madrona’s McIlwain casts an envious eye at data and machine-learning software firm Databricks Inc., which is now valued at more than $6 billion and has a customer list that includes Cisco Systems Inc., HP Inc. and ViacomCBS Inc. Turi may have become more valuable if it too had stayed independent, he said, but it also could have gone worse.
It’s been a hard road for other AI independents. Employees at Clarifai, founded in 2013, were initially confident they had found a defensible niche. Instead, the company struggled to maintain momentum. Last year, the startup cut about 20% of its staff, according to two people familiar with the situation. Clarifai Founder and CEO, Matt Zeiler, didn’t respond to requests for comment.