A new high: DeepMind burned $650 million a year, but Google waved to forgive $1.5 billion in debt

A new high: DeepMind burned $650 million a year, but Google waved to forgive $1.5 billion in debt

649 million, DeepMind’s loss last year reached a new high, even more than Google spent to buy it. But unlike the fate of Boston Dynamics, which has been repeatedly resold, Google’s attitude towards it is: give as much as you want.

Founded in 2010 by Demis Hassabis, Mustafa Suleyman and Shane Legg, DeepMind was acquired by Google for $600 million in 2014. The company is known to the community for its extensive research results, such as AlphaGo, which beat the world champion Go, AlphaFold 2, which predicted protein structures, and more.

  A new high: DeepMind burned $650 million a year, but Google waved to forgive $1.5 billion in debt

DeepMind began to be widely known as AlphaGo, which defeated the strongest human beings in the field of Go. In May 2017, Ke Jie, the world’s No. 1 player at the time, lost 0 to 3 in the man-machine battle.

However, DeepMind is also remembered for “burning money”.

A recent filing with the UK Companies Registry showed that DeepMind lost $649 million in 2019, a 1.5% increase from its 2018 loss. The bulk of this was for “personnel and other related expenses”, which totaled £468 million ($634 million), up from £398 million ($540 million) in 2018.

Currently, DeepMind has about 1,000 employees worldwide, including the world’s top AI research scientists, whose annual salary may exceed $1 million. DeepMind’s top talent typically holds doctorates from schools such as Oxford, Cambridge, Stanford or MIT, and is sought after by tech giants such as Facebook, Apple, Amazon and Microsoft.

A DeepMind spokesperson said: “During the time covered by these fees, DeepMind’s groundbreaking results have laid the groundwork for solving the grand challenge of 50 years in biology, protein structure prediction. DeepMind is also working with teams at Google to make a real-world impact. We His team is involved in many projects, including improving the predictability of wind power, promoting ecological research in the Serengeti, and more.”

Despite a slight uptick in losses, DeepMind’s 2019 revenue reached £266 million ($360 million), a 158% increase from £103 million ($140 million) in 2018. However, the main source of DeepMind’s revenue is the research and development fees paid to DeepMind by other Alphabet subsidiaries, including Google.

DeepMind was acquired by Google for $600 million in 2014, and currently DeepMind relies on the steady flow of capital from its parent company Alphabet.

In 2019, Google Ireland forgave DeepMind on intercompany loan repayments and all accrued interest totalling £1.1 billion ($1.5 billion), the documents show.

DeepMind is headquartered in the Google building near London’s King’s Cross, but it plans to move to a new location early next year. In 2019, DeepMind spent a total of £1.3 million ($1.76 million) on buildings, furniture, networking and production equipment, the documents show.

Additionally, DeepMind donated £6.3 million ($8.53 million) to academia in 2019, down from £13.5 million ($18.28 million) in 2018.

In terms of risks and uncertainties, the document states: “Machine learning research and applications is an emerging market that is constantly changing and highly competitive. As such DeepMind will continue to face risks and uncertainties, which may increase its gains in the market. The ability to sustain success has a significant impact.”

But despite DeepMind’s year-on-year losses, parent company Alphabet has pledged to continue raising funds for DeepMind.

Alphabet CEO Sundar Pichai said on the company’s second-quarter earnings call: “I’m very pleased with the progress we’ve made in AI. For me, the most important thing is that we have achieved SOTA at the company level, And leading the way. I’m very happy with the pace of work of the engineering and R&D teams at Google and DeepMind.”


In December 2020, DeepMind’s AlphaFold 2 solved a major challenge in biology for 50 years-the problem of protein molecular folding, which shocked the world for a while.

Azeem Azhar, founder of the Exponential Perspective podcast, said: “The question, in my opinion, is why aren’t more companies in other industries supporting core research in exponential technologies like AI, synthetic biology, battery technology, quantum computing. Support can be Direct investments like Google’s can also be corporate ventures, or partnerships with universities, but very few companies (like Volkswagen or Bosch) do that.”

Azeem Azhar also added: “You won’t see a huge social return in five years, maybe 20, 30 years or even longer. For example, the Intel 4004 was released in 1970, but only half of the UK was released in 2001. of households own a computer.”


The Links:   1D500A-030 2M120B150J-120 LCD-STOCK