Google owns DeepMind, a subsidiary of Alphabet, so it is naturally dependent on the search engine giant to fund its research and operations. But because DeepMind is dependent on Alphabet, it may choose to focus on areas of research that are more profitable than those that are less profitable. DeepMind’s stated mission is to advance science and benefit mankind, which means it may be able to ignore short-term profits in favor of long-term gains.
While revenue at DeepMind has increased, it’s not clear what exactly has caused the revenue to skyrocket. Alphabet hasn’t announced any deals with DeepMind outside the group, so it’s hard to determine the specific reason for the increase in revenue. But DeepMind doesn’t sell its products or services directly to consumers, so it may have been able to achieve the revenue increase by using “creative accounting”.
While large tech companies have taken interest in deep learning, recruitment of top AI researchers has become a competitive arms race. Some leading researchers can earn seven-figure salaries at tech giants. Meanwhile, small-tech labs and academic institutions struggle to retain their best researchers. Last year, DeepMind reported that it would spend PS467 million on salaries for its 1,000 employees. Despite its modest size, it pays its staff very well. The company plans to hire more than one thousand new researchers by 2020.
Though the financial reports are vague, the numbers seem promising. In 2018, DeepMind reported revenue of PS560 million. In 2020, the company expects to generate a PS826 million profit. This is up from the PS265 million it made last year. But while revenue growth is promising, the company isn’t making money fast enough to keep investors happy. The company has to make money to stay afloat, and it can’t do so unless it can sell services to Alphabet companies.
Google’s decision to forgive PS1.1 billion of debt to DeepMind in 2019 may prove beneficial in the long run. It might also help DeepMind report positive earnings by 2020. Alphabet is willing to forgive DeepMind’s debt in the future, as long as they stay loyal to the company. Otherwise, it would be left with no funding and no customers, and will face stiff competition from tech giants.
The research at DeepMind doesn’t seem to translate into a viable business model. One of its best-known projects, AlphaStar, is a reinforcement learning system that has mastered the real-time strategy game StarCraft 2. Even though it was probably subsidized by Google because it costs millions of dollars, its capabilities have limited practical applications outside of tuning billions of extras. Its use in a business model for applied AI is limited.
As Google’s AI research lab, DeepMind has struggled for years. It has been a loss-making operation for years, but in 2020, it posted a profit, with revenues of PS826 million. Previously, DeepMind posted losses of hundreds of millions of pounds. The company’s revenue has tripled in a year, but the company has not provided a rationale for the rise.
As a subsidiary of Alphabet, DeepMind doesn’t sell its own products to consumers, though it has partnered with several companies. But it hasn’t announced any of these partnerships publicly. A source in the industry said the revenue increase could be due to a creative accounting approach. DeepMind’s financial report doesn’t give much detail about its sources of revenue, but analysts say the company is making progress.
Meanwhile, DeepMind’s chief executive is defending the controversial deal with the NHS. The company’s CEO has publicly stated that the data sharing with Google is “not in breach of UK data laws.” He said that this agreement is a result of a data transfer that he says will lead to “significant improvements” in medical care. Nevertheless, he also said that the deal is profitable and will lead to a more competitive DeepMind.