0:00 SpaceXAI just spent sixteen billion dollars on AI in a single quarter.0:04 That’s the kind of number that changes the entire conversation about who can even compete.0:09 Last week on episode 133, we were tracking the conversation around the soaring costs of just training a frontier model.0:17 Today, we're seeing what it costs to build an entire ecosystem from scratch, and the price is astronomical.0:24 The spend from SpaceXAI is the headline, but it’s not the only story moving today.0:29 The threads on tech twitter show a landscape that's rapidly expanding in three directions at once.0:35 First, there’s a fundamental shift in ambition.0:38 The conversation is moving past just making models smarter.0:41 A thread from the researcher Hardmaru is getting a lot of attention for a simple, powerful idea: automating the scientific method itself.0:50 He argues that creating an AI scientist, one that can form hypotheses, run experiments, and learn from them in a recursive loop, will profoundly alter the trajectory of AI.1:01 This isn't just about better chatbots or code assistants.1:04 This is about using AI to automate discovery.1:07 That’s a whole different level of impact.1:10 Second, the geopolitical lines are getting sharper.1:13 The Japan Times is highlighting the launch of Moonshot AI's new model, Kimi K3.1:18 The key takeaway isn't just that it's another powerful model from China.1:22 The point is that Chinese AI is now so deeply embedded in the global technology industry that it’s impossible to ignore.1:30 This isn’t a separate ecosystem anymore.1:32 It’s direct, head-to-head competition with U.S.1:35 firms, and it's happening on the global stage.1:38 Kimi K3 isn't just a product launch; it's a statement of intent.1:42 And third, there’s a growing sense of anticipation for what comes next.1:46 A post from SciTech Era captures this perfectly.1:49 With all this focus on models and cloud infrastructure, the conversation reminds us that we haven't even seen the AI-native hardware wave yet.1:58 Nothing on the scale of the smartphone or the PC.2:01 All this software and intelligence is still running on devices that weren't designed for it.2:07 The consensus is that a tsunami of AI-powered hardware is coming, and it will change the user experience from the ground up.2:15 We just don't know exactly when it will hit.2:17 So you have these three major currents: AI as a scientist, AI as a geopolitical battleground, and the coming wave of AI-native devices.2:26 But they all point back to one thing.2:28 The incredible, eye-watering cost of admission.2:31 Which brings us back to SpaceXAI.2:33 Okay, let's go back to that sixteen-billion-dollar number.2:37 Because Tomasz Tunguz laid it out in a thread today, and the details are just staggering.2:42 SpaceXAI spent eighteen billion dollars in capital expenditures last quarter.2:47 Sixteen billion of that was on AI.2:49 To put that in perspective, Microsoft—a foundational hyperscaler—spent forty-one billion in total.2:55 Amazon spent fifty-three billion.2:57 So SpaceXAI, in a single quarter, spent nearly forty percent as much on AI infrastructure as Microsoft spent on its ENTIRE global operation.3:06 That is not a company dipping its toe in the water.3:09 That is a company trying to boil the ocean.3:12 The rate of spending is also on par with the giants.3:15 The sequential increase, from quarter one to quarter two this year, was about nine billion for Microsoft and Alphabet.3:23 For SpaceXAI, it was over eight billion.3:25 They are accelerating at the same pace as the biggest players on the planet.3:30 But here’s the turn.3:31 It’s where the money comes from.3:33 Microsoft funds its buildout from its operations, with plenty of cash to spare.3:38 Its operating cash flow is one hundred and fifty-five percent of its capex.3:43 Meta also clears the bar.3:44 They earn more than they spend.3:46 But almost everyone else is borrowing or issuing equity to keep up.3:50 And SpaceXAI is at the extreme end of that spectrum.3:54 Their operating cash flow covers just twelve percent of their spending.3:58 Twelve.3:58 The other eighty-eight percent is coming from investors, from debt.4:03 They are making a colossal bet with other people’s money.4:06 And those people are getting nervous.4:08 The stock, which IPO'd in June at one hundred thirty-five dollars a share, closed yesterday at one hundred and eight.4:16 It’s lost over a trillion dollars in market value from its peak.4:20 The bond market agrees.4:21 Every single tranche of the twenty-five billion dollars in bonds they issued in June is now trading below par.4:28 The long-term debt, the bonds due in 2056, are trading at just ninety cents on the dollar.4:34 The market is saying it does not believe this level of spending is sustainable.4:39 So what does it all add up to?4:40 You have a company spending like a hyperscaler, accelerating like a hyperscaler, but funded like a startup.4:47 It's an all-or-nothing gamble.4:49 The thing is, it might just work.4:51 They have the data centers.4:53 They have a model, Grok 4.5, that is genuinely competitive—it ranks fourth on the main Intelligence Index, just a few points behind the leader.5:02 And they have millions of developers using their tools, like Cursor.5:06 The IPO gave them a huge war chest and a great cost of capital.5:10 The question now is whether that advantage, that willingness to burn cash at an unprecedented rate, can last long enough for the AI business to actually start paying for itself.5:21 And this is where the other threads we saw today click into place.5:25 Why make a bet this big, this risky?5:27 Because of what Hardmaru is talking about.5:30 The prize isn't just a better search engine.5:32 The prize is automating science, automating intelligence.5:36 The potential payoff is society-altering.5:39 And why now?5:39 Because of what the Japan Times is reporting about Moonshot's Kimi K3.5:44 The competition isn't just coming from down the street in Silicon Valley.5:48 It's coming from Beijing.5:50 It's global, it's fierce, and it's accelerating.5:53 If you wait, you lose.5:54 This isn't a normal business decision.5:57 This is a strategic, almost military-grade mobilization of capital.6:01 It's a bet that being the third or fourth AI hyperscaler is an untenable position.6:06 The goal is to be number one or two, or die trying.6:09 This week's chatter shows the board has been redrawn.6:12 The era of clever algorithms winning the day is over.6:16 We are now in the era of brute-force industrial-scale investment.6:20 The conversation is no longer about who has the smartest model.6:24 It's about who has the deepest pockets and the strongest nerve.
