How AI is Draining Google’s Finances — Transcript
Full transcript
- 0:00On Wednesday, Google's parent company
- 0:01Alphabet published its financial results
- 0:03for the third quarter of this year, and
- 0:05well, they weren't well received.
- 0:08Spending came in hot, revenue came in
- 0:10soft, and the stock duly slumped by 10%
- 0:13in after-hours trading, wiping about
- 0:15$400 off Alphabet's market cap in barely
- 0:19a few hours. So, in this video, we're
- 0:21going to explain why Alphabet stock just
- 0:23slumped, and why, shock, it's all to do
- 0:25with AI.
- 0:33Sick of slop? Well, Tail Digger's
- 0:35content is entirely AI-free, with us
- 0:37having strict rules when it comes to the
- 0:39use of AI in our company. So, while we
- 0:41might not be perfect, at least you know
- 0:44we're human. Find our full AI policy
- 0:46linked in the description.
- 0:48So, simply put, the fundamental problem
- 0:50for Google is that spending came in
- 0:52higher than expected, and revenues came
- 0:54in lower than expected. And well, you
- 0:56don't need an MBA to know that's not
- 0:58good business. Let's start with
- 1:00spending. In the third quarter alone,
- 1:02which is July to September this year,
- 1:04Google spent a whopping $87 billion,
- 1:07which is up 22% from the same period in
- 1:102025. This follows a broad trend, with
- 1:13spending in the year-to-date up 21%
- 1:16reaching $236 billion in the first three
- 1:19quarters. Obviously, the culprit here is
- 1:21AI, but what does that mean? Well, we
- 1:24can break down Google's soaring AI spend
- 1:26into two main categories: capital
- 1:28expenditure and TPU inventory build-up.
- 1:31If those things mean nothing to you,
- 1:33don't worry, we'll explain each in
- 1:34detail. And when we're done, you should
- 1:36have a clear idea of what hyperscalers
- 1:38are actually spending money on as they
- 1:40fund the AI build-outs. First, capital
- 1:43expenditure, or capex. This refers to
- 1:45money being spent up front by Google
- 1:47towards its own infrastructure. As you
- 1:50can see on their balance sheet, this
- 1:51value of the company's physical assets
- 1:54has leapt up since last year, which
- 1:56reflects growing capex investments.
- 1:58Right there, you can see $51.4 billion
- 2:01spent in the last quarter alone, which
- 2:03we're told is funding Google's AI build
- 2:05out. But, what on earth is that money
- 2:07being spent on? Well, according to CFO
- 2:10Anat Ashkenazi, it's a 65/35%
- 2:13split between active computing hardware
- 2:15and physical data center infrastructure.
- 2:18The computing stuff is essentially GPUs,
- 2:20CPUs, and TPUs, along with the tech that
- 2:23allows huge chip clusters to communicate
- 2:25with each other. The data centers costs
- 2:27go towards physical buildings, their
- 2:29energy supplies, and cooling systems.
- 2:32So, that's CapEx. Next is TPU inventory.
- 2:35Google's AI build out strategy is
- 2:37two-pronged. On the one hand, they're
- 2:39building out their own computing power
- 2:40so they can reap the rewards of the AI
- 2:42boom. On the other hand, they're acting
- 2:44a bit like Nvidia and designing their
- 2:47own specialized AI chips called TPUs, or
- 2:50tensor processing units, which they plan
- 2:52to sell on to clients like Anthropic,
- 2:54Apple, and Oracle. If you look at that
- 2:56inventory line on the balance sheet, you
- 2:58can see that since last year, an
- 3:00additional $10.9 billion
- 3:02has been spent building up Google's
- 3:04inventory. A few years ago, that line
- 3:06would have reflected products like Pixel
- 3:07phones sitting in warehouses waiting to
- 3:10be sold. Now, we're told that the figure
- 3:12is almost entirely accounted for by TPU
- 3:14building. But, before you leap to
- 3:16conclusions, there's an important point
- 3:18here to keep in mind, which illustrates
- 3:20that Google's AI spend isn't without
- 3:22justification. Google's Cloud CEO,
- 3:24Thomas Kurian, pointed out in
- 3:26Wednesday's earnings call that the
- 3:27company only produces components and
- 3:29builds TPU inventory after signing
- 3:32contracts with customers. In other
- 3:33words, while this multi-billion-dollar
- 3:35figure is sticking out on their books
- 3:37like a sore thumb, those processors are
- 3:39going to be shipped soon and turn into
- 3:41revenue.
- 3:42In broad terms, this speaks to a dynamic
- 3:44in the AI build out which benefits
- 3:46hyperscalers like Google, Amazon, and
- 3:48Microsoft over companies further down
- 3:50the food chain like OpenAI and
- 3:52Anthropic. Sure, the hyperscalers are
- 3:54spending huge amounts building out
- 3:56America's computing power, but they also
- 3:58have significant commitments from
- 4:00enterprise customers, which, if all goes
- 4:02well, means huge revenues will follow.
- 4:04Of course, we'll be the first to admit
- 4:06that if all goes well is doing a lot of
- 4:08heavy lifting in that sentence. That
- 4:10brings us to the other side of the
- 4:12ledger, revenues. At first glance,
- 4:14revenues look pretty good. Total revenue
- 4:17for the quarter came in at $127.6
- 4:20billion,
- 4:21only a tenth below analyst forecast of
- 4:23$128 billion,
- 4:25and up 25% compared to a year ago, i.e.
- 4:29to the third quarter 2025. The thing
- 4:31that slightly spooked the markets,
- 4:33however, was where that revenue came
- 4:35from. For context, Alphabet's revenues
- 4:37are often broken down into three
- 4:39segments. Google services, which covers
- 4:41stuff like ad revenues from Google
- 4:43Search and YouTube, as well as
- 4:45subscription fees, Google Cloud, which
- 4:47basically rents out computing power and
- 4:49software to businesses and other
- 4:50organizations, so they don't have to
- 4:52build and run their own data centers,
- 4:54and other bets, which covers the
- 4:57loss-making more experimental
- 4:58businesses, and is usually a relatively
- 5:00small slice of the overall pie.
- 5:03Historically, Google services has
- 5:04accounted for the vast majority of all
- 5:06revenues, and has been by far and away
- 5:09the most profitable part of the
- 5:10business. The thing that spooked the
- 5:12markets this time round is that Google
- 5:14services revenue came in below forecast
- 5:17at $96 billion.
- 5:19That's up 11% year-on-year, but down
- 5:21slightly compared to the previous
- 5:23quarter. This was in part because
- 5:24revenues from Google Search,
- 5:26historically by far the largest and most
- 5:28profitable part of Alphabet's business,
- 5:30came in at $63.3 billion, about $1.5
- 5:34billion below forecast, and basically
- 5:36flat on the previous quarter. The only
- 5:38reason that headline revenues came in
- 5:40line with expectations is because of an
- 5:42above-forecast print from Google Cloud,
- 5:44which generated about $30 billion in
- 5:46revenues for the quarter, roughly $2
- 5:49billion more than forecast. Now, at this
- 5:51point you might be asking, "What's wrong
- 5:53with all this?" It sounds like Alphabet
- 5:55is essentially diversifying, becoming
- 5:57less singularly reliant on the services
- 5:59part of the business, and offsetting any
- 6:01shortfalls over there with more revenues
- 6:04elsewhere, like Google Cloud. And yet,
- 6:06while diversification is generally a
- 6:08good thing, Google services has been a
- 6:10supremely reliable money printing
- 6:12machine for Alphabet for years now. And
- 6:15markets apparently aren't convinced that
- 6:17other parts of the business, like AI,
- 6:19can truly fill its shoes.
- 6:21Google Search, for instance, has been a
- 6:23money printing machine in large part
- 6:25because it's had a near monopoly on the
- 6:27search engine market for years, if not
- 6:29decades now. It's nice that Google Cloud
- 6:32is making money, but it only became
- 6:34consistently profitable in around 2023.
- 6:37And it's not obvious, at least to the
- 6:38markets, why it would ultimately be as
- 6:41profitable as, say, Google Search, given
- 6:43the lack of any obvious moat.
- 6:46At the same time, the trend lines look
- 6:47like they're squarely headed in the
- 6:49wrong direction for search at the
- 6:50moment, as people move away from
- 6:52traditional search towards AI chatbots.
- 6:55Alphabet themselves admitted on
- 6:57Wednesday that the growing popularity of
- 6:59Google's AI mode was taking a toll on
- 7:01search revenues, because AI queries
- 7:03apparently have lower monetization
- 7:05rates.
- 7:06All in all, this combination of lower
- 7:07revenues and lots more spending meant
- 7:09that Alphabet's free cash flow came in
- 7:12at negative $9 billion, down from
- 7:14negative $5.9 billion in the previous
- 7:17quarter, and only the second quarter in
- 7:19the red since Alphabet went public in
- 7:212004. So, that's Alphabet's financial
- 7:23situation right now. They're in a
- 7:25somewhat risky position, which explains
- 7:27the stock dip this week. But there's
- 7:29certainly room to recover if their bet
- 7:32on AI pays off. Exciting announcement.
- 7:35I'm happy to reveal the cover of the
- 7:36next issue of Two Long, our 10th. The
- 7:39main story is all about the ascendancy
- 7:41of China, spanning energy, artificial
- 7:43intelligence, global trade, and an
- 7:45increasingly assertive posture towards
- 7:47its neighbors. The country's trajectory
- 7:49points towards total dominance. Yet
- 7:52beneath the surface of this meteoric
- 7:53rise, cracks may be forming. You can
- 7:55pre-order your copy right now, and if
- 7:57you use code 10, you'll get £3 off any
- 8:00new subscription, including the Two Long
- 8:02and TLDR party bundle subscription. Also
- 8:05also mentioning, as always, there's much
- 8:07more than just the cover story. In fact,
- 8:09and I know you'll love it when I say
- 8:11this, your talk of our longest issue
- 8:13ever is so repetitive.
- 8:15>> I don't want to hear about your longest
- 8:16issue ever.
- 8:17>> It's only one or two pages more than the
- 8:19other issues anyway. But this time it
- 8:21truly is longer at 100 pages to
- 8:24celebrate our 10th issue. That means
- 8:26that we also have tons of space to
- 8:27discuss the US midterms, the return of
- 8:29company towns, the future of Taiwan, the
- 8:32seeming rise of the AFD, Burnham
- 8:34reversing Thatcherism, Ukraine's
- 8:35corruption scandals, and much more. Find
- 8:38out more about the latest issue and
- 8:39subscribe to Too Long by heading to our
- 8:41website. It's linked in the description.
About this transcript
This page contains the full transcript of How AI is Draining Google’s Finances by TLDR Business, generated from the public captions YouTube serves with the video. The transcript has 1,508 words across 244 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.