Please create an account to participate in the Slashdot moderation system

 


Forgot your password?
Close
typodupeerror
×
AI Software Businesses

Software Revenue Lags Despite Tech Giants' $292 Billion AI Spend (indiadispatch.com) 67

Posted by msmash from the tough-luck dept.
Silicon Valley is betting the farm on AI. Data centers are straining power grids. Model training costs are heading toward billions. Yet across the software industry, AI revenue remains theoretical. From a report: Hyperscalers -- combined with Meta and Oracle -- plan to spend $292 billion on AI infrastructure by 2025 -- an 88% increase since 2023. Two-thirds of software companies, however, still report decelerating growth in 2024.

Semiconductor stocks have surged 43% year-to-date on AI expectations, while the software index IGV is up 30%. Microsoft, despite its OpenAI investment, has underperformed the IGV by 19% since ChatGPT's release. Microsoft's AI revenue run rate is 3% of total revenue, according to estimates by investment bank Jefferies. Snowflake expects immaterial AI contribution in fiscal 2025. Salesforce isn't factoring in material contribution from new AI products into FY25 guidance. Adobe's Firefly AI, launched in March 2023, hasn't accelerated revenue.

Software Revenue Lags Despite Tech Giants' $292 Billion AI Spend More | Reply

Software Revenue Lags Despite Tech Giants' $292 Billion AI Spend

Comments Filter:

  • FTFY (Score:5, Insightful)

    by DjangoShagnasty ( 453677 ) on Monday December 23, 2024 @08:24AM (#65034211)
    > Software Revenue Lags Because Of Tech Giants' $292 Billion AI Spend

    • I assume you're saying that the cost of AI has to come from somewhere...

      But AI is being used to replace human workers, so companies no longer need as many licenses.

      People talk about how much savings there is from AI because of payroll, benefits, and sometimes even office space... but just think of the savings from the companies forcing people to pay monthly or annual licensing costs for everything. That adds up pretty quickly the way software companies price things these days

      • I think they meant that the money that would be spent on software is now being spent on AI.

        I don't see AI replacing customer service any time soon, unless you also want to replace customers, possibly with lawsuits.

      • But without workers, no one to make stuff. AI is not at the stage to replace productivity. So money spent on it has to come from somewhere, and the massive amount spent is not coming from useless workers being laid off. Otherwise, just lay off the useless workers without spending on AI and win. The chief problem is with executives who naively believe that the current state of AI is useful.

        Companies make money off of margin, mostly, with cutting back on R&D usually being less important. That's true ev

    • Re: (Score:2)

      by gweihir ( 88907 )

      Exactly. These investments will never pay off.

  • A top level E5 license in the US that includes TONS of MS products and services, a call-in phone number hosting for Teams meetings, etc costs about $20/mo if I remember correctly. One copilot license is $30 PER PERSON. That suggests it (sort of) costs more than developing the entire Office 365 suite and cloud.

    I can say with some certainty that's an introductory price of a hot new product that they want to push so you get used to it and eventually find it hard to live without before they increase the price

    • In my experience CoPilot is a really advanced refactoring and templating/boilerplate engine that usually works pretty well. It's not worth $30. ReSharper is almost as good once you learn all of it's tricks.

      • Re: (Score:2)

        by CEC-P ( 10248912 )
        My company has actually gone down the path of only hiring people for jobs involving reporting, analytics, and business decisions who know how to read, type, and do math. So we haven't really seen a need for Copilot or any AI so far. That is, except for legacy employees that I'd describe as having their entire career propped up by Grammarly and caffeine already. Even those types aren't loving the Teams Meeting copilot "summaries" that are longer than the meeting was and puts "Sorry, I have to let me dog out"

        • Re: (Score:2)

          by gweihir ( 88907 )

          My company has actually gone down the path of only hiring people for jobs involving reporting, analytics, and business decisions who know how to read, type, and do math.

          That should be the standard case. No idea why it seems to qualify as "smart" these days. Hiring people without these qualification probably costs more than they produce in revenue.

      • Re: (Score:2)

        by gweihir ( 88907 )

        Interesting. And ReShaper does not hallucinate, I take it?

        • Re: (Score:2)

          by JBMcB ( 73720 )

          Interesting. And ReShaper does not hallucinate, I take it?

          No, but once in a blue moon there will be a bug that will do something weird to your code, like insert non-printing characters that break indenting. It's rare, but it does happen.

    • If AI lets you fire bad workers and make your phone tree more efficient, cool

      "press 0 to speak to a customer service rep" should be mandatory by law. Without it, people can be blocked from exercising consumer rights.

      It would be fascinating to see how many people choose to follow AI prompts/IVR if there was a direct way to queue for real customer service.

      • When asked to speak my request to the customer service bot, I enunciate clearly, "let me speak to a human". Then it says "I did not understand, please speak your request again". I repeate this three times typically, then the bot says "I did not understand, let me transfer you to a messy and smelly human."

        I don't know if what I say makes a difference there versus speaking gibberish, but if they do record for quality purposes, as it often claimed, then someone may hear this eventually.

        • That is similar to what I do, but already came across companies that will hang up. Tesla UK is an example. Not only they hide the Customer Service phone number from their site, they have aggressive "channel deflection" in place.
            If you have an order number for a car, it is accepted by one of the prompts, but if you ordered a wall connector, the IVR will reject 2x and then hang up.

          In my opinion, AI is a tool for mass enshitification.

    • Wait so they want us to pay for copilot? I don't even want that shit if it paid me

    • Re: (Score:2)

      by gweihir ( 88907 )

      These $30 are not nearly covering the cost. I am reminded of the UMTS frequency auctions in Germany where every mobile phone user would have had to pay $900 per year to ever make them profitable. The investors are simply out of their minds and cannot do basic math anymore. If, say, copilot ends up costing something like $100/month to actually be profitable (not unrealistic), the product is dead already. Even at $30/month it may be effectively dead.

    • Remember, that $30 per person is being sold to the same companies that post up warnings about not using too much toilet paper. Or for software companies, the ones that only allow 5 people to use licensed software at a time (including "free" toolchains because of commercial embedded linux distros).

  • AI bubble warning signs flashing purple (Score:5, Interesting)

    by Tablizer ( 95088 ) on Monday December 23, 2024 @08:52AM (#65034257) Journal

    they ran out of red. The revenue-to-investment ratio is far too small to keep the pace. Investors have limited patience, and AI is hitting a plateau in revenue. Investment money is subsidizing almost every AI tool. It's hardware intensive, so not easy to give away for ad revenue alone, unlike say Facebook.

    The poppage might trigger a general recession, so hold on to your hats.

    • They've gone to plaid. This is a parallel to the dotcom boom/bust cycle from the mid-1990s to 2000, when the bubble burst, and the Clinton Recession began.

      • Re: (Score:2)

        by Tablizer ( 95088 )

        RE: "Clinton Recession" -- Prez has only a very minor short-term impact on the economy. Why does everything think the Prez has a giant economy lever and oil lever? Shhtoopid voters *slap* *slap*

      • Re: (Score:2)

        by Targon ( 17348 )

        Weird how the tech sector only went to "bubble" territory in around 1998 or so before it popped in 2000(after stupid people assumed that Y2K wasn't even a real problem because it was taken care of before 2000 began). So, how did the bubble in your mind turn into the "Clinton" recession? The crash hit due to all of those MBAs with access to venture capital who started an Internet company without even knowing anything about the Internet or technology. If anything, the downturn was caused by the wealthy

        • I'd say that is pretty much it. A lot of stuff was thrown into the market without any real thought. Beenz/Flooz, the "free" Compaq machine that had 1024x768 monitor space, but 800x600 usable, rest devoted to ads. It was also when the old guard changed, and PCs replaced UNIX workstations, which effectively killed SGI and Sun.

          What kept the economy really in the dumps was 2001. 9/11 happened, and no business hired for several years after that, just out of trepidation.

    • Re: (Score:2)

      by gweihir ( 88907 )

      Yep. If companies like Microsoft and Google were not completely stagnant (and starting to devolve) at this time and desperate to avoid that impression, this bubble would not even have happened and OpenAI would never have amounted to much.

    • I don't doubt that a major recession is coming, but I don't see AI as the cause. Other than some jobs, only the rich will lose there, but they have already hedged their bets.
  • ...the spend is mostly on h/w - gpu's, etc...

  • Enterprise 3d TV... (Score:4, Interesting)

    by fuzzyfuzzyfungus ( 1223518 ) on Monday December 23, 2024 @09:10AM (#65034309) Journal
    This feels a lot like the attempt to astroturf an alleged craze into actual sales numbers that accompanied 3d TV a number of years ago; except playing out mostly at the enterprise-focused SaaShole level rather than primarily consumer focused.

    The (predictable) tragedy is that this seems to be directly competing for resources with the unhyped but desperately needed "retain employees who can implement what your customers actually want and fix your shit" strategy. Same old pain points remain unfixed, price is higher, but there's a chatbot and some more aggressive autocomplete!
    • There was enterprise 3D tv - it was VR/AR headsets and glasses. How'd that work out? The useful talent who knows better than AI who gets pushed out or pissed off over AI is still going to be smarter and more useful than AI. If their current employers are too stupid chasing that short term bubble buck, they'll go somewhere else and keep being smart. The stupid big companies are going to have worse products that cost more and more, and there's only one way that'll go for them.

  • The real problem with AI, is the clueless investor (Score:4, Interesting)

    by WarlockD ( 623872 ) on Monday December 23, 2024 @09:51AM (#65034367)
    I mean the pitch is perfect. Anyone who has used ChatGPT knows how janky it is but you see the potential. Image generation isn't perfect but damn is it "good enough". Code generation does have some strange bugs BUT most of that 5000 lines work! We just need the hardware! And the people! Oh, maybe a nuclear powerplant or three! Oh you have lawyer insurance right?

    I mean will it all crash and burn like back in the 00's? Nah. Both Google and Microsoft are invested. The tech IS janky but it can solve so many problems for THEIR particular needs. This isn't where "general tech" is the thing and investors were throwing money at anything that had an ethernet port. I will say, though, many of those small startups that didn't get sold immediately are going to crash and burn.

    • Re: (Score:2)

      by Tablizer ( 95088 )

      To be fair it's incrementally getting better. The real question is whether it gets better fast enough to justify the huge investment sums continually being poured into it. I'd guess "no", but predicting the future is an art.

      We keep predicting a bubble pop, but the incremental improvements keep inventors' knickers moist. It's a tease. Eventually even teases wear out; investors want to hump profits.

      • Re: (Score:2)

        by gweihir ( 88907 )

        Hahaha, no. It is not getting better. It is getting its parameters tuned a bit on the filters and "hardcoded stuff" side, but the actual LLM side probably has stagnated completely a year ago or longer and is getting _worse_ now due to model collapse.

    • Well, my mod points ran out. So someone with points please mode parent up insightful.

    • Re: (Score:2)

      by gweihir ( 88907 )

      Yesterday, somebody posted an LLM search result for a "constant development scam" (I think) which has exactly the characteristics of the AI pitch.

      The real problem with LLMs is that they will not get any better than the pretty pathetic level they have now.

      • Re: (Score:2)

        by Targon ( 17348 )

        No, the problem is that if you don't have a goal in mind, the actual development effort will result in nothing worthwhile. There are some people so obsessed with the idea of AI, they haven't figured out what to use it for, and if it isn't useful for ANYTHING, it will be a waste of time, effort, and money.

      • Seems like once a company gets to the size of Microsoft, etc, profit takes a back seat to maintaining dominance. Most of this AI arms race seems like a corporate FOMO (fear of missing out). Google expediting their AI earlier this year (or was it last year? I forget) when it wasn't ready illustrates that. If things really get hairy, govt has set a precedent to bail out any company deemed "too big to fail." Corporate safety nets and insurmountable obstacles to entry has enabled reckless spending and decision
        • It's about manipulating stock pricing, up or down. They always win.
        • Well that's my point is the companies that use the tech won't feel a thing. You already see it in Google's search results and Copilot from Microsoft. These are products right now working. Now they might not buy as MUCH hardware next year but you can be sure they will keep all this going. None of the big corps are going to need any kind off bailout, but there will sure be a lot less jobs soon from all the splurge.

  • AI is one of those things that is already transformative, and the software companies are well aware that this is a race for dominance. Yes it costs a lot to be in the race but if you aren't in it you are sure to lose.

    My guess is that costs will drop off rapidly as the tech matures. Training will be less expensive, inference will take way less cycles. The stuff they are doing now will be at the level of a commodity in the near future and AGI will be the truly expensive objective. Pretty soon the machines wil

    • Re: (Score:2)

      by gweihir ( 88907 )

      You are hallucinating. AGI is not an objective at this time, because nobody has the slightest clue as to how it can be done or whether it is possible at all. And no, LLMs will not get a lot cheaper than they are now. This is decades old tech and already highly optimized.

    • You gonna let these things do your requirements analysis (my clients would cause the machines to smoke), QA, DevOps, and support? You nuts?

      • I sure will when they are doing it better than the humans can. At this rate it won't be long.

        • GLWT. Personally, after playing with them for a couple of months, I think it's a dangerous fantasy. You can't train them on crap code and expect them to churn good code, let alone process.

          • I've seen varying results as well, but I am noticing the code generation is getting better. Completions tend to be accurate. You can type in a description of what you want to do, at least in some limited circumstances, and get a pretty good template of the solution.

            • Keyword: template. That leaves a lot remaining for human intelligence to solve. And again, requirements need to be specified pretty clearly. That's really always been the hard part.

              • >> That leaves a lot remaining for human intelligence to solve.

                I agree. At least for now, humans get to decide what we want the machines to accomplish for us. I just hope that someday they aren't telling us what they want us to accomplish for them.

                • That's not how I look at it. The people who control the machines control some aspects of their output. I have an interesting example that shows ChatGPT censoring itself - writes content and then deletes it as a violation of its own terms (seems like a bug - text shouldn't have been visible). Looks like it's trying to reduce lawsuit potential. Happy to share if anyone wants to investigate, but it's also kinda personal. That's just one example of how the people who control the machines can control the people

  • There is a chance that it MIGHT produce something of value in the future.
    Scientists are doing great work with tools like Alpha Fold, but today's consumer AI is just crap generators.
    As investors demand profits, expect a tsunami of half-baked, useless, annoying crap to be forced on us, with the most frequent question being, How can I turn this off?

    • Re: (Score:2)

      by gweihir ( 88907 )

      There is some value in existing AI. There is very little of it (basically "better search" at this time) in LLMs. These investments are in no way justified. IMO they are just a result of the desperation of companies that have become very stagnant, like Microsoft, Google, and others.

      AI research had been ongoing for 70 years. Except for LLM research, the investments have probably paid off. My prediction is that the excessive and irrational LLM investments will never even come close to pay off. More like 10% RO

  • More likely that stupid investment is the reason for it.

  • It's a small price to pay, but I can fix their problems.

  • AI is sucking up money at companies trying to figure out how to use. Many before they figure out what it even is... Starting with many IVRs... AI hell.... We gone from press one for English to press one for a human!!!!! Many who were early adopters of AI tech are backing away due to the lack of controls.... Courts are starting to rule that if an AI bot for your company says something. then thats as good as an employee saying the same thing.

Slashdot Top Deals

Profanity is the one language all programmers know best.

Close