be.x2m.in

Tag: AI

  • The Spenders: Who Writes the Checks

    This is post 2 of 2 in the series “Who Actually Gets Paid in AI”

    A series on the AI buildout, layer by layer: who is spending, who is growing, who has the cash engine today, and who only has a story about tomorrow. Starts with the bubble question.

    1. Is This an AI Bubble?
    2. The Spenders: Who Writes the Checks

    Before the names, my labels, because I will use them in every post in this series.

    Real: the growth and the cash are here today, and the price roughly pays for what already exists.
    Real but rich: a solid business with real growth, but the price already assumes years of flawless execution.
    Potential: the opportunity is real, but the numbers are early, small, or not yet durable.
    Hype: the story and the price run far ahead of any revenue or cash.

    Part 1 asked if this is a bubble. This part looks at the six companies writing the biggest checks. First finding: none of the six is Hype. Every check is written by a real cash engine. The difference is how much engine is left after the check, and who pays if the bet fails.

    Company One line Label
    Microsoft Azure grew 48% and free cash flow stayed positive at $18.6 billion in the quarter, the only big spender whose cash never went under water. Real
    Alphabet Cloud more than doubled and operating cash flow is about $180 billion over twelve months, but the quarter’s free cash flow was negative $9.1 billion on $51.4 billion of building. Real
    Meta Revenue up 26% on an ads engine that prints cash, operating cash flow about $133 billion and growing, but it also raised new shares, so owners split the future with more people. Real
    Amazon AWS grew 39% and the cash engine is real, but $77 billion of capex in one quarter took twelve month free cash flow to negative $42.4 billion. The return on this spend is the unproven part. Potential
    Apple The control case. It lets others write the checks, made about $110 billion of free cash flow in nine months, and the stock has run well ahead of its own history while its AI revenue is still mostly other people’s models on its devices. Real but rich
    Oracle The backlog is real and huge, but the build is funded with debt and new stock, with a credit rating one notch above junk. If the bet fails, lenders and diluted owners pay first. Potential

    So the spenders are not the dot com story. Nobody here is spending money they do not have from a business that does not exist. The honest risk is narrower: the returns have to arrive before the patience runs out, and for Amazon and Oracle, before the bill gets uncomfortable.

    I am writing what I think today so I can come back and check if I was right or wrong. These posts will keep me honest.

    My call today: when these six report their next full round in late January, at least four will still show operating cash flow growing year over year. If two or more stall, my Real labels here are wrong, and I will change them in public.

    Reminder on the labels: Real means growth and cash are here and the price roughly matches. Real but rich means a solid business priced for perfection. Potential means a real opportunity not yet proven in durable numbers. Hype means story and price far ahead of revenue and cash.

  • Is This an AI Bubble?

    This is post 1 of 2 in the series “Who Actually Gets Paid in AI”

    A series on the AI buildout, layer by layer: who is spending, who is growing, who has the cash engine today, and who only has a story about tomorrow. Starts with the bubble question.

    1. Is This an AI Bubble?
    2. The Spenders: Who Writes the Checks

    October was rough for AI stocks.

    Alphabet fell about 19% in the month. Meta fell about 15%. It was not because the businesses suddenly stopped growing. Alphabet revenue was up 25%. Meta revenue was up 26%.

    The market hated the spending.

    Both companies are spending huge amounts to build AI capacity. That spending has pushed free cash flow down hard, even into negative numbers for the quarter. Their bet is simple. Spend through a weak free cash flow period now, and the future free cash flow will be amazing.

    That bet may be right. It may also be wrong. But I keep coming back to an old line from Jeff Bezos in 1997. When forced to choose between making the accounting look pretty and the present value of future cash flows, he said, “we’ll take the cash flows.”

    I agree. Free cash flow is the destination.

    But while the building spending is front loaded, I am watching a different number. Operating cash flow. That is the cash the core business collects before the building spending. It tells me if the engine still works.

    Alphabet operating cash flow is about $180 billion over the last twelve months, and growing. Meta is about $133 billion, and growing. Their free cash flow has collapsed because they are building, not because the engine has stopped.

    So is this a bubble?

    Parts of it, yes. Some AI names are priced for perfection, with huge multiples and no cash flow to show yet. If anything goes wrong, there is nothing under the price.

    But the biggest spenders do not look like the dot com bubble to me. That bubble was built by companies with little revenue and borrowed hope. These are profitable giants funding the buildout from enormous operating cash flow.

    Both things can be true. AI can be a real shift, and some AI prices can still be a bubble.

    The question is not whether AI is real. The question is who has a cash engine today, and who only has a story about tomorrow.

    This is too big for one post. I will do more posts on this topic soon, one layer at a time. The spenders, the chips, the builders, and the software on top. Then I will come back to the bubble question with the whole map in front of us.

  • My Mom Asked Me What AI Is

    My mom asked me to explain what AI is.

    I tried a few answers in my head, and none of them sounded right. Then it hit me: if I can explain it so my mom gets it, that is the best explanation there is.

    AI is badly named. "Artificial intelligence" promises something it is not.

    Don’t get me wrong. I am amazed by it every single day. It is very smart software. But smart is not the same as intelligent.

    Intelligence is judgment. Knowing what matters, what is right, and what to do when no rule covers the situation. That is a human thing.

    What AI actually is: software that is exceptionally good at pattern matching. Finding patterns in mountains of information, faster than any person ever could. It is better at that than me, any day and every day.

    Once you see it that way, you stop expecting the wrong things from it — and start using it for the right things. It finds patterns faster than I can, finds precedents I would never find alone, and compares things better than I can on my own. That is why it is an amazing co-pilot, advisor, sparring partner, coder, and assistant.

    So here is how I think about it now: AI is the best pattern-matching machine ever built. I bring the intelligence — the judgment about what to do with the patterns it finds.

    That worked for my mom. I think it works for everyone.