back to article DeepSeek means companies need to consider AI investment more carefully

The shockwave following the release of competitive AI models from Chinese startup DeepSeek has led many to question the assumption that throwing ever more money at costly large-scale GPU-based infrastructure delivers the best results. As The Register reported earlier, shares of some of the largest American tech brands in the …

  1. Doctor Syntax Silver badge

    "Historically, the AI industry has relied on scaling models, increasing data volume, and enhancing hardware performance for growth"

    Don't forget the disdain for other people's copyright.

    1. ComputerSays_noAbsolutelyNo Silver badge

      I wonder what the reading on the Irony Meter is:

      https://arstechnica.com/ai/2025/01/i-agree-with-openai-you-shouldnt-use-other-peoples-work-without-permission/

    2. JRStern Bronze badge

      there is no "historically" for AI, other than flame out and failure.

  2. cookiecutter

    Why is no one taking about the education gap

    Deepseek might be stolen tech.

    But no one seems to be taking about the way China has got millions of engineers, highly trained, highly motivated & paid.

    In the US, UK because of morons like Mckinsey, Welch, Accenture PWC, we've been shipping out jobs, training other countries. Denegrating experts. Cutting wages, making degrees more expensive. Companies have created a situation where only a bootlicking idiot would do anything over 50% effort becsuse you know you're going to be fired when the CEO wants a new car.

    All the western government's seem to care about its the banking scum who mushroom money upwards, force companies to send yet more expertise offshore & make work more and more pointless.

    Compare wages in tech in the 90s & 00s to today. Abd is NOT because there's more engineers, companies are still screaming for staff apparently. It's because all that money has mushroomed upwards.

    Personally I would LOVE a few big hacks of the banks & consultancy firms..I'm talking genuine weeks outage scale ransomware so we can laugh at them for creating the situation

    1. Anonymous Coward
      Anonymous Coward

      Re: Why is no one taking about the education gap

      China does have millions of engineers, but not many of them are naturally talented...we have a good number of naturally talented engineers in the UK, those that grew up with tech that naturally ended up in the industry...but they're often thrown on the slag heap with all the factory produced "battery farm" techies that come off the production line.

      "In the US, UK because of morons like Mckinsey, Welch, Accenture PWC, we've been shipping out jobs, training other countries. Denegrating experts. Cutting wages"

      Absolutely, back in the day...late 90s / early 00s...I was doing odd bits and pieces for very large companies. British Airways, Research in Motion, Lufthansa, DHL, Hitachi, Cisco, Sony...as a teenager! Because back then, it was always about finding the talent and using it...the big contractors hadn't jumped on the bandwagon yet...I designed an internal intranet site for British Airways in the late 90s which was largely static HTML (but with a tiny bit of PHP thrown in for content management) that was used by some cargo workers to track stuff...I built it in an afternoon, I was paid £5,000 for it, cash...I was 15! I knew a guy that worked in the department there and he asked if I would be interested in building something, so with it being BA, I figured I'd do it, because at the very least it's something I can put on my CV at some point...when I finished it and demo'd it, the guy shook my hand and gave me £5k...I was gobsmacked...being 15 at the time, the first thing I did was get on my bike and ride to the chippy...bought myself a nice fish supper...I was sat outside the chippy tucking in with a 5 grand wad in my pocket...I didn't count it until I got home...up until I counted it, I assumed it was just a couple of hundred quid...as for the other firms, I wasn't always paid as handsomely as that, but the barrier to entry for these places was much lower, like RIM for example, they had an office in Egham (near Heathrow) and someone I knew worked there...they twigged that I was proficient at coding, and he'd take me in after hours to help him with some coding...I got paid little token amounts for this, like £50 here and there out of his own pocket...point is, this sort of thing just wouldn't happen now...Hitachi flew me out to CeBIT to assist with setting up prototypes...Cisco would occasionally send me out to businesses in my area to fix problems with switches and routers (even though I didn't have any kind of Cisco cert)...again, I knew people there through acquaintances...people I met just doing techie stuff...I was quite into early wifi and I used to go around sniffing it out looking for stuff with no passwords then knocking on doors and offering to fix it...you can't do that anymore, you'll get banged up for it.

      There is just so much about the tech industry that has been lost / hidden away behind massive fucking walls and moats...I don't recognise the tech industry anymore...it's not what it once was.

      "Compare wages in tech in the 90s & 00s to today"

      I often do, my earning capacity peaked when i was about 15 during this era...what is crazy about the earnings during that era is that the work was a lot less sophisticated. You could get £1,000 for a single page static HTML website back then for a random local business that you could throw together in an afternoon, you wouldn't even have to ask for the money, you would just do the work and wait to be paid because you knew people would just pay well. There was no negotiating, it was just understood that tech work was valuable...I never once had to go cap in hand during this era.

      It wouldn't be a crap website either, my designs were always at least at the cutting edge of what was possible, because I always (and still do) experiment with tech and sharpen my skills in my own time...I regularly look back on the Wayback machine at sites I built back then and track the progress of my skills on there, between around 1997 and 2003 I shifted my approach at least twice a year and you can see the jumps in improvement as the tech improved and my abilities improved...frames, to tables, to sliced up PSD files, to CSS...the jumps were sometimes enormous...then we reach about 2007 and websites have all been pretty "samey" ever since, web design has been fucking boring for decades now, all that has changed is that we now have frameworks and they get bloatier with each new release / development all in the name of "making it easier"...I don't really do web design anymore (at least not professionally, I do for my own kicks though) because it's just not worth it. It takes longer now, because the direction that tech went in, and the money is orders of magnitude less...mostly because of "market rates" rather than what the work is actually worth to the business you're doing it for.

      These days you're lucky to get £1000 for a fucking full blown app that takes a month.

      I look at the nearest modern equivalent to the web stuff I was doing back then, which is essentially deploying a Wordpress site with an off the shelf template, and you can get that done for £20 now through a freelance website populated mostly by Indians.

      1. The man with a spanner

        Re: Why is no one taking about the education gap

        "China does have millions of engineers, but not many of them are naturally talented."

        What stunning arrogance!

        This is the attitude that ensured that British Leyland maintained its technological domination over the world car market place. . . .no need to worry about the Japanese with their obsesion with incremental improvement in quality - our natural talent is there for all to see.

        (Note: may contain irony)

  3. Anonymous Coward
    Anonymous Coward

    "Chinese startup shakeup doesn't herald 'drastic drop' in need for infrastructure buildout, say analysts"

    Emphasis on "anal".

  4. Gene Cash Silver badge
    Flame

    "companies need to consider AI investment more carefully"

    Why? They never have before. Why should they start now? They just chant "AI AI AI" worse than Steve Ballmer and consider it the solution to ALL their problems including the decision of what's for lunch.

    I'm hoping they all lose their shirts. Then perhaps they'll examine a technology for suitability next time... nah, wait, who am I kidding.

    Icon for hoping all the AI companies and AI-using companies burn to the ground. Edit: especially the ones using chatbots for "support"

  5. DS999 Silver badge

    People may be looking at this the wrong way

    Moore's Law gave us more computing power for the same $$$ for decades, and it resulted in MORE investment in computing not less. I'm not sure we should assume that being able to train LLMs with 1/10th of the computing power (that's basically what DeepSeek's numbers/claims amount to) will result in everyone spending less on AI. Maybe it means Microsoft doesn't feel like they need to spend $85 billion on AI and could get away with less but I doubt they'd knock it all the way down to $8.5 billion. They'd figure they could do MORE AI or bigger models or whatever, and invest somewhere between those two figures.

    What's more, at current pricing very few companies can play in the AI game. If it costs 1/10th as much that opens the door to a lot of other companies, and they will be wanting to buy AI hardware to take their shot at becoming the next trillion dollar company.

    So once everything shakes out, Nvidia might see their order book get more full, rather than as feared there are cutbacks since they can do their training with much less hardware.

    1. HuBo Silver badge
      Thumb Up

      Re: People may be looking at this the wrong way

      Quite right imho! DeepSeek mostly shows that Chinese folks are on the job, doing what they do, which is (often) to re-make what we (in the West) do, less expensively (eg. EVs, solar panels, ...), but also at a lower quality point in my experience.

      Japan (also Korea) followed that route some time ago and initially produced cars (for example) that were inexpensive, but also cheap and unreliable (a bit crappy), and that today are of great quality while maintaining favorable price points.

      There's a learning curve to those things and we need to stay ahead of it with top-notch education, innovation, and entrepreneusrhip, to maintain our lead and global relevance, without relinquishing our world-famous freedoms, inclusivity leadership, and open-mindedness!

      1. Bebu sa Ware
        FAIL

        Re: People may be looking at this the wrong way

        "we need to stay ahead of it with top-notch education, innovation, and entrepreneusrhip, to maintain our lead and global relevance, without relinquishing our world-famous freedoms, inclusivity leadership, and open-mindedness!".

        Game over then?

        Dylan Thomas' words from his "Do Not Go Gentle into That Good Night" come to mind:

        "Rage, rage against the dying of the light."

      2. Emir Al Weeq

        Re: People may be looking at this the wrong way

        >inexpensive, but also cheap and unreliable

        They may have been inexpensive and cheap (what's the difference?) but compared to the UK domestic* produce they were considered very reliable and better equipped. So they sold.

        * RoW: YMMV

    2. flayman

      Re: People may be looking at this the wrong way

      I've read this MIT Technology Review [https://www.technologyreview.com/2025/01/24/1110526/china-deepseek-top-ai-despite-sanctions/] on DeepSeek and it seems to say that sanctions and limitations directly led to innovations that increased computational efficiency in training models. This suggests that AI in the west is suffering from serious bloat. Moore's law is one thing, but as computer hardware has become exponentially more powerful, this has resulted in massively bloated operating systems and application software because the need to program efficiently has fallen away. There was a time when the application designer had to fight for and justify every byte in memory. Imagine what amazing things could be achieved if those sort of engineering limitations came back!

  6. CowHorseFrog Silver badge

    Maybe the real q is why is something valued so highly when it has achieved nothing of value ?

    1. CowHorseFrog Silver badge

      Then again we must also remember that America also values celebrities...

  7. JRStern Bronze badge

    But Chinese startup shakeup doesn't herald 'drastic drop' in need for infrastructure buildout, say analysts

    But Chinese startup shakeup does herald 'drastic drop' in need for infrastructure buildout, says me.

    They showed that by going beneath CUDA they could so some purely mechanical tuning and run 10x faster and/or cheaper.

    Almost like NVidia was purposely providing support only for very inefficient operation.

    Nah, now why would they do that?

  8. harrys Bronze badge

    stating the bloody obvious.....

    500W plus gpu's, even in home PC's

    look out the window!

  9. JimmyPage
    Stop

    Mandy Rice Davies applies

    (RIP)

    1. Anonymous Coward
      Anonymous Coward

      Re: Mandy Rice Davies applies

      for those who are too young to remember

      https://en.wikipedia.org/wiki/Well_he_would,_wouldn%27t_he%3F

  10. IGotOut Silver badge

    And yet...

    ....there is no killer use for this insane amount of money that could of been spent doing something useful.

  11. Mike Friedman

    Just the latest IT fad.

    Repeat after me. AI is a fad. It will never been anything more than a niche thing.

    Remember how Blockchain was EVERYWHERE 10-12 years ago? It was going to REVOLUTIONIZE EVERYTHING!

    Note. It did not do so and is now .....a niche application.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like