Thursday. Your humble servant spent the morning working for other people, so we’re condensing our format to maximize focus on the most important stories in technology today: Google’s second-quarter results and the current war in Silicon Valley over whether the major AI labs deserve protection or mockery.
- P.S. Yes, Tesla reported as well. Its shares went through the wringer after it reported mixed results(better-than-expected revenue, worse-than-expected profitability). Strong 26% revenue growth was beaten back by 47% higher operating expenses, leading to -57% operating income compared to the year-ago period. Tesla’s negative free cash flow in the quarter wasn’t as deep as expected, which was a win. Shares are off 14%, nearly pushing Tesla’s market cap under the $1 trillion threshold.
Google’s impressive quarter/Google’s admission of failure #
Alphabet, Google’s parent company, reported revenues of $119.80 billion in the second quarter, ahead of an anticipated $116.93 billion result. The company missed on adjusted earnings per share, reporting $2.85, under an expected $2.89.
With Google Cloud seeing revenue rise 82% in the quarter (compared to the year-ago period) and total growth coming in ahead of market expectations, you might expect Google shares to be up this morning. They aren’t. Instead, Alphabet stock is down nearly 7% in regular trading, a decline worth hundreds of billions of dollars in market cap terms.
What’s going on? Let’s break it down:
Google’s AI metrics didn’t impress:Announcingthat its “model APIs are now processing approximately 22 billion tokens per minute,” up from 16 billion in the first quarter, was a neat detail. But without a hard revenue figure attached to the number, it felt more vanity-ish than world-conquering.Google AI models are behind the curve: During its earnings call, Google talked up its recently released AI models. Those same modelsfailed to push Google’s AI offerings up the intelligence charts; further delays in Gemini 3.5 Pro were also not ameliorated by claims that Gemini 4 is seeing strong early training results. When you are already behind, saying you may catch up later is not a confidence-building exercise.Google’s capex forecasts rose: Google announced that its full-year 2026 capex guidance now rests in between “$195‑205 billion, up from [its] previous estimate of $180‑190 billion.” Investors are skittish about rising AI infra spend, concerned that it may not convert to revenue at a winsome rate, and that hyperscalers are starting to overspend their — admittedly — impressive financial foundation.- Even worse: Google told investors that it still expects “capex to increase significantly in 2027.”
Google’s cash flow went negative: All that capex powering second-tier AI models? It’s so costly that Alphabet flipped from generating cash to eating it, posting -$5.9 billion in free cash flow in the quarter despite generating $39.1 billion inoperatingcash flow*.*- It’s actually very difficult to spend that much money. You have to reallytry.
- It’s actually very difficult to spend that much money. You have to
All that sounds downright nasty (and we didn’t even get into depreciation!). What’s the other side of the coin? Well, there’s a lot to like in Google’s results:
Google Cloud growth is acceleratin g: Its 82% growth rate was up from 63% in Q1 2026, 48% in Q4 2025, and 34% in Q3 2025. That’s insane for a business now on a ~$100 billion yearly run rate. (And most TPU-based cloud revenues, ie selling chips, won’t come online until 2027, so we’re not mixing results in this case.)Google Cloud’s backlog continues to scale: The company now claims a cloud backlog of $514 billion, up $50 billion from the first quarter. Or about twice its current cloud quarterly run rate.Google is still compute-constrained: Anat Ashkenazi, Google’s CFO, saidthe company is still “in a supply‑constrained environment,” thanks to “very strong demand, both from external Cloud customers as well as across the business.” Ding ding ding, compute constraints! Today’s cloud giants could be growing faster if they had the chips. That’s bullish.
Our final point is not a pure win. Google shares a single compute base across its fleet of services, including bringing its AI products to bear in search, YouTube, and its enterprise products. Google (Alphabet) sells compute and inference to third parties and must apportion compute for both its internal use and the training of new models. I wonder if the company is more constrained internally than we might think; could Google’s recently slow AI progress indicate that it is making painful tradeoffs that may impact its ability to regain the frontier?
Maybe! Investors do not appear excited by its spending pace, so long as it sits behind three American AI labs (Anthropic, OpenAI, SpaceXAI) and two Chinese labs (Moonshot, Z.ai) in terms of model intelligence. Telling investors during the same earnings call you announce a flip to negative free cash flow that you intend to spend even more next year is a fat bolus of shit to swallow.
Folks critical of the great AI infra buildout view Google’s earnings as indication that it’s struggling — where is the pure AI revenue number? On the other hand, AI demand is up, quarterly revenue records are being broken, and Google Cloud is accelerating while still scrounging for compute.
What should you think? It doesn’t matter. You own a lot of Google shares in your index funds, 401ks, and the like. You are already long!
Everyone thinks distillation worries are bullshit #
The White House is beating the drum that Chinese AI models are unfairly distilling off American AI models, leading to an unfair competitive landscape. One that might, the implication goes, be suited to a bit of punitive regulation being applied to open-weight Chinese models.
- The argument here is pretty simple: If American AI labs are burning mountains of capital to train new models, only for Chinese companies to rip them off and resell the work at a lower token price, the United States could find itself subsidizing the decline of its own AI industry. Under a ‘Cold AI War’ framing, that’s bad!
Most folks in technology aren’t buying the argument. When Michael Kratsios — the President’s Science and Technology advisor — wrote that he has “information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model” thanks to a “sophisticated internal platform to conduct large scale distillation against U.S. models,” and that “large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable,” he got booed.
Over on Twitter, commentary from technologists and investors ranged from be quie*t**, *to * prove it, *to
.
general raspberriesThen Jacob Helberg, Under Secretary of State for Economic Affairs at the State Department and husband of investor Keith Rabois, jumped into the ring. Calling the alleged distillation “more than a heist of invaluable American Intellectual Property,” Helberg characterized the work as an “assault on every economy that prizes the entrepreneur, rewards private capital, and relies on fair and honest competition.” The undersecretary went on to argue that “lying, cheating, and stealing is not innovation,” but a substance “corrosive to productive economic activity and to the healthy economic development of nations.”
The response? Benchmark’s Bill Gurley argued that industrial research is something American companies take part in. Rippling’s Park Conrad wanted to know what law was broken, and why Anthropic isn’t suing the alleged thieves in question. The hoi polloi were not swayed.
Much criticism of the administration’s position that China is cheating in the AI game by distilling from American models rests on the argument that *Anthropic and OpenAI stole all our data already, so why should they get to cry foul when someone else does the same to them? *A few examples:
- “ I don’t remember letting Anthropic scrape my GitHub profile as well as all of my online writing!” - “ We have information that Anthropic distilled humanity’s entire internet for the development of Fable 5.“ - “ Please can you explain why Moonshot training on Anthropic’s output is stealing but Anthropic training on creatives’ output is not?”
To quote Mitt Romney, what’s sauce for the goose is sauce for the gander.
The subheadline to this section is wrong. Not everyone is throwing darts at the petitions of closed-source AI labs and the apparently bent ears of the White House. Here’s Founders Fund’s Mike Solana, quote-tweeting Kratsios:
what’s wrong, you don’t like the free market?*
*american startups competing against the second wealthiest nation state in human history, which is presently committed, from the top down, to destabilizing our industry and cornering the global intelligence market
Why bring up Solana’s complaint? Founders Fund lists several American AI labs on its portfolio page, including OpenAI and SpaceX (SpaceXAI). The company also took part in a recent Anthropic round.
As we wrote the other day: “To understand which tech folks are incensed or quiet at present, merely look where they have worked, invested or have friends. The investors in OpenAI and Anthropic are in favor of their backed horses.”
Before the current scrap over open-weight Chinese models, their alleged distillation from American AI models, and possible restrictions on use of China-built AI, there was a debate about what winning meant viz China. The argument went something like *‘we can’t worry about IP protections and such because we need to beat China in the AI race.’ * The prize was long-term economic dominance. I suspect that some American technologists still think along those lines.
But with lower-cost Chinese AI models keeping close to the frontier-leading American products, it may be that neither nation ‘wins,’ but instead we see healthy economies on both sides of the Pacific with each nation building ever-better products. And consumers everywhere reaping the benefits.
The national economic security — or just national security, I suppose — argument should still hold if you once found it compelling; in that case, Anthropic’s cries of cheating! may resonate. But Anthropic and OpenAI have found themselves with incredible growth rates and many detractors at just the moment when they could use some friends to help them protect their revenue bases.
I’m not pointing a finger at general AI critics, but technology founders and CEOs who have watched Sam and Dario expand horizontally into their markets — Cowork and ChatGPT Work are not small efforts — while charging them often-exorbitant token prices. To see those same companies work, allegedly, to close the door on lower-priced competition is about as welcome as an empty elevator shaft. No, thank you.
So, it’s a mess out there. Startups and technology companies worried about being eaten by the very companies they are wiring cash to want more, cheaper AI. Closed-source labs and their backers want to defend their market edge and ensure that they can continue to scale and go public.
You and I? We just want more and better technology products. And it seems that there are myriad voices arguing that trying to regulate Chinese AI models out of the domestic mix is verboten.
A fun question: How many investors in OpenAI and Anthropic are secretly annoyed at their alleged rent-seeking, but can’t speak out as it may harm their bags?