Microsoft Confirms 10,000 Layoffs Amid Recession, Plans to Invest Heavily in AI

Microsoft
(Image credit: Shutterstock)

We reported yesterday that Microsoft could announce the elimination of 10,000 positions this week. Today, in a memo sent out (opens in new tab) to all Microsoft employees, CEO Satya Nadella confirmed the disappointing news.

To put the layoffs into perspective, Nadella explained, “[We’re] seeing organizations in every industry and geography exercise caution as some parts of the world are in a recession and other parts are anticipating one… we as a company must strive to deliver results on an ongoing basis, while investing in our long-term opportunity.”

The workforce reduction, which Nadella says represents just under 5% of its global footprint of 220,000 employees, will occur between now and the close of Microsoft’s fiscal third quarter (March 31, 2023). Unfortunately, some employees will receive notices of dismissal as early as today, while others will face their fate with the company in the coming days and weeks. 

The mass layoffs mean that Microsoft will take a $1.2 billion charge during fiscal Q3, in part to cover severance packages for employees. Speaking of which, Microsoft says that it will offer departing employees “above-market severance pay,” six months of continuing healthcare, and stock vesting for an additional six months.

“As a company, our success must be aligned to the world’s success,” Nadella continued. “That means every one of us and every team across the company must raise the bar and perform better than the competition… If we deliver on this, we will emerge stronger and thrive long into the future; it’s as simple as that.”

Despite this massive workforce reduction, Microsoft will “continue to hire in key strategic areas.” Nadella boasted in an interview two weeks ago with CNBC that he is “super long on India” and envisions the country elevating from its current fifth-place position to third place in global economies by 2030. He added, “India is an exception in a challenging world.” 

Artificial intelligence is the next big mountain to climb in computing, and India is number one in the sector. So, it stands to reason that India will play a significant role in Microsoft’s more selective new hires in the coming months and years.

While 10,000 layoffs is a big number, it’s not the largest in the company’s history; Microsoft eliminated 18,000 positions in 2014. However, we should note that workforce reduction occurred during the transition from former CEO Steve Ballmer to Nadella. In addition, two-thirds of that figure came from employees let go following Microsoft’s doomed Nokia Devices and Services acquisition. 

Microsoft isn’t the only big tech company feeling the heat from a slowing economy. Amazon is in the process of laying off 18,000 employees, while Facebook parent Meta is parting ways with 11,000 workers.

 

 

Brandon Hill is a senior editor at Tom's Hardware. He has written about PC and Mac tech since the late 1990s with bylines at AnandTech, DailyTech, and Hot Hardware. When he is not consuming copious amounts of tech news, he can be found enjoying the NC mountains or the beach with his wife and two sons.

  • PlaneInTheSky
    Invest in AI huh.

    There was this video recently on Youtube that showed how terrible Tesla cars are at self parking.

    The Tesla is incredibly hesitant to park, is slow to park, and can't make up its mind at times.

    Odd that a car company like Tesla can't get a fairly easy mathematical question right, while other car companies can easily solve this problem and park effortlessly.

    Someone in the comment sections figured out why this happened. Tesla used "AI", so-called "machine learning", to train its car to park while all the other car companies used mathematical formulas. Tesla relies on data of millions of people who parked their car to train their "AI". The problem is of course that this includes tons of data of people unable to properly park a car.

    What people in the tech world call "AI" has nothing to do with AI of course. Nvidia started using this term for simple mass data gathering of human behavior that is used as a seed to train an algorithm, so-called "machine learning". But this form of "AI", the same AI Microsoft talks about, has the inherent flaw that this data includes unreliable data since humans are far from perfect.

    You see this with ChatGPT that Microsoft invested in too. Another tool that is AI based but makes horrible mistakes. And you can not filter these mistakes out of these AI data sets, that would require going through all your data sets 1 by 1, which is impossible since these algorithms (almost all based on a paper that appeared a few years ago), use millions of data sets.

    Something else that people have noticed with these AI algorithms is that they seem to get "dumber" over time. This is of course a consequence of the fact that early data sets you trained your algorithms on, are of much better quality because you solve the low hanging fruit first. The more and more edge cases you need to solve to improve the algorithm, the less reliable data there is available to solve these problems, and the dumber your algorithm becomes.


    nsb2XBAIWyA
    Reply
  • Mandark
    The fact is intelligence cannot be artificial. When everybody calls, artificial intelligence is just mathematical mimicking. There’s absolutely no intelligence whatsoever behind it it’s brute force calculations.

    looks like Microsoft will be in for a world of hurt, because that crap isn’t going anywhere

    maybe someday when we can make biological brains and copy the human function they can achieve consciousness, and therefore have intelligence, but until then forget it

    this isn’t helping anybody’s lives, now I have to worry about some idiot using auto pilot driving into me. Teslas statistics are a joke and complete lies as well because they’re not complete Data sets. Refer to their highway miles which completely distorts the image that they project.
    Reply
  • brandonjclark
    This comment section tells me that some Tom's readers have no idea what they're talking about when it comes to AI and RL\ML.

    Guys, I'm going to say this one time...

    Our ability to program deterministic and stochastic AI models will IMPROVE over time, just as it always has.

    You're really showing your inability to change and predict the future if you can look at something like OpenAI's ChatGPT and think that it's going to get worse over time or that there is no intelligence behind it.

    Honestly, this is THE future and you'd better get on board very quickly. I won't warn you again.
    Reply
  • korekan
    brandonjclark said:
    This comment section tells me that some Tom's readers have no idea what they're talking about when it comes to AI and RL\ML.

    Guys, I'm going to say this one time...

    Our ability to program deterministic and stochastic AI models will IMPROVE over time, just as it always has.

    You're really showing your inability to change and predict the future if you can look at something like OpenAI's ChatGPT and think that it's going to get worse over time or that there is no intelligence behind it.

    Honestly, this is THE future and you'd better get on board very quickly. I won't warn you again.
    just like an era before computer.
    computer can do nothing some/ most said. turns out it kill a lot manual task in life.
    cant imagine if paying bills still need to write form and queue at the bank for each bill, before that need to queue at the bank to withdraw as well then you forgot 1 bill and need to redo it.
    Reply
  • PlaneInTheSky
    You're really showing your inability to predict the future if you can look at something like OpenAI's ChatGPT and think that there is no intelligence behind it. Honestly, this is THE future and you'd better get on board very quickly. I won't warn you again.



    https://i.postimg.cc/8C2RG8td/dgsgsgsgs.png
    https://i.postimg.cc/zvX6CRKG/sdfsdfsfsfs.png
    Reply
  • Mandark
    brandonjclark said:
    This comment section tells me that some Tom's readers have no idea what they're talking about when it comes to AI and RL\ML.

    Guys, I'm going to say this one time...

    Our ability to program deterministic and stochastic AI models will IMPROVE over time, just as it always has.

    You're really showing your inability to change and predict the future if you can look at something like OpenAI's ChatGPT and think that it's going to get worse over time or that there is no intelligence behind it.

    Honestly, this is THE future and you'd better get on board very quickly. I won't warn you again.
    😂😂😂😂😂

    everything I said is completely correct and provable. Everything you’ve said is not correct and is not provable. Anyone can claim that over time things will get better. The fact is computers will never be able to understand and comprehend, and they can never be conscious, and therefore they are not intelligent, and never will be. I don’t care what mathematical algorithms you use. You can’t compute consciousness.

    it is you who is being naïve
    Reply
  • bit_user
    PlaneInTheSky said:
    What people in the tech world call "AI" has nothing to do with AI of course. Nvidia started using this term for simple mass data gathering of human behavior that is used as a seed to train an algorithm, so-called "machine learning". But this form of "AI", the same AI Microsoft talks about, has the inherent flaw that this data includes unreliable data since humans are far from perfect.
    Did it not occur to you that you can filter out the bad examples? You can even generate synthetic data to train AI, in many cases. Or just let it guess and then score its attempt.

    PlaneInTheSky said:
    you can not filter these mistakes out of these AI data sets, that would require going through all your data sets 1 by 1, which is impossible
    A simple Bayesian classifier could go a long ways towards eliminating bad training samples. Spam filtering was revolutionized by the technique. Spammers had some success trying to work around them, but that's because they made a concerted effort to find things that would slip through. If you're just filtering historical data, then you don't have that kind of "arms race" dynamic, so it should work pretty well.

    PlaneInTheSky said:
    Something else that people have noticed with these AI algorithms is that they seem to get "dumber" over time. This is of course a consequence of the fact that early data sets you trained your algorithms on, are of much better quality because you solve the low hanging fruit first. The more and more edge cases you need to solve to improve the algorithm, the less reliable data there is available to solve these problems, and the dumber your algorithm becomes.
    If your model is too simple, then adding more training samples to handle more cases will indeed make it worse. You can solve that by making it bigger, but then you need more training samples. In many cases, synthetic training samples are adequate. You can also use GAN to generate training samples that tend to be more realistic.
    Reply
  • bit_user
    Mandark said:
    You can’t compute consciousness.
    Time and again, when people have drawn a line and declared that science cannot explain what's beyond it, they've eventually turned out to be wrong.

    100 years ago, we barely knew about micro organisms and didn't even know about DNA. Now, we're not only able to decode and synthesize DNA, but we can even edit the DNA inside the cells of a living animal!

    So far, there's no evidence of anything in the human brain that's not simply a function of chemistry, neurons, synapses, and dendrites. All of these mechanisms can be described in an abstract fashion and emulated by computers. There's no reason to believe a computer big & fast enough couldn't ultimately emulate human-like consciousness. What's exciting is that by experimenting with these abstract models, we're even gaining more insight into human cognition, itself.
    Reply
  • Mandark
    You’re a materialist. But still you don’t know what it is for sure. You can’t know the answer because even neuroscientists don’t know the answer. You cannot prove what you say is true at least not yet.

    investigate the series Closer To Truth by neuroscientist, Dr. Robert Lawrence Kuhn

    he’s a trained neurosciencentist and he can’t answer the question. So how can you know the definitive answer? The answer is you don’t and you can’t. There’s people on both sides of the camp.

    I’m not saying it’s something mystical, but we don’t know what it is at this point and I’m not saying that we won’t figure it out either. .

    also, do some research on Sir Roger Penrose, and watch some of his lectures. He’s brilliant.
    .If Sir Roger Penrose doesn’t know the answer then I guarantee that you don’t either

    Roger Penrose believes that it has something to do with the collapse of the wave, that consciousness is happens on the quantum physics level and consciousness happens around the collapse of the wave function

    I’ve been studying this problem for quite a long time, and for you to come in here and say you know what the answer is, buddy, I’m telling you you don’t because not even the experts and the biggest thinkers know. And we may never know that’s just a fact it may turn out that we can’t figure it out.

    You can believe all you want, but if we will figure it out — my stance is, we’ll see.

    and with that, I’m not going to argue about this with you anymore. I am allowed to make my own comments but I’m not going to dispute this anymore

    https://mindmatters.ai/2018/11/human-consciousness-may-not-be-computable/
    hXgqik6HXc0View: https://youtu.be/hXgqik6HXc0
    https://www.theguardian.com/science/2020/feb/27/why-your-brain-is-not-a-computer-neuroscience-neural-networks-consciousness
    Reply
  • bit_user
    Mandark said:
    You can’t know the answer because even neuroscientists don’t know the answer. You cannot prove what you say is true at least not yet.

    investigate the series Closer To Truth by neuroscientist, Dr. Robert Lawrence Kuhn

    he’s a trained neurosciencentist and he can’t answer the question. So how can you know the definitive answer?
    The brain is a complex system. Biologists tend to get bogged down by complexity, because they focus too much on the details and not the overall system. That's why I believe abstract models are the key to understanding the mysteries underlying cognition.

    Mandark said:
    The answer is you don’t and you can’t.
    Same for you. How much have you read about the insights that deep learning research has provided into cognitive functions and general intelligence? It's a rapidly moving field.

    Mandark said:
    for you to come in here and say you know what the answer is,
    Please go back and reread my post, because I didn't say that, nor do I see how you could even think I did. All I said is that I don't see any reason it's fundamentally impossible.

    Mandark said:
    not even the experts and the biggest thinkers know.
    It's not how smart you are, but whether you have the right tools and skills at your disposal. Science is a process and a community that continually builds and evolves a common knowledge and understanding. If you focus on a single individual, you're missing the point.

    Every day, there are people making discoveries that were missed by their predecessors, even those of superior intellect, because they simply lacked the tools and/or the building blocks.

    Mandark said:
    And we may never know that’s just a fact it may turn out that we can’t figure it out.
    Whenever humans face problems we can't overcome by brute force, we build tools. In the information age, many of these tools are computers and software.

    Mandark said:
    with that, I’m not going to argue about this with you anymore.
    Sure, you can always flip the chessboard and walk away. I can understand it's more comfortable to believe human intelligence is special and it seems as though you've found someone who's saying what you want to hear. I know it can be hard to stay open-minded past that point. But, big thinkers have been wrong before, and human knowledge is always growing and evolving.

    Remember that research is a fundamentally iterative process.

    "If I have seen further it is by standing on the shoulders of Giants.”--Sir Isaac Newton
    Reply