Sponsored by
In today’s post:
😳 Even Google Ran Out of AI
🐂 Even The Bulls Are Scared Now
🐉 China Caught Up. Now What?
The most trustworthy AI admin agent for busy executives.
Busy executive. Packed calendar. An inbox that never empties.
Everyone is talking about AI agents. Every week brings another promise of an assistant that can do it all.
Catch is the real deal.
A smart, proactive AI admin agent focused solely on taking administrative work off your plate.
It schedules meetings, triages your inbox, drafts emails in your voice, resolves conflicts, sends follow-ups, and handles the countless small tasks that consume your day.
Available wherever you work — Gmail, Outlook, Slack, WhatsApp, and even over the phone.
No setup. No training. No learning curve.
Catch learns how you work, takes action when it's confident, and keeps things moving without constant supervision.
From swamped to sorted in seconds.
Get started with Catch and have your assistant ready before your next meeting.
😳 Even Google Ran Out of AI
Turns out even trillion-dollar companies have to wait in line.
Google just throttled Meta's access to its Gemini AI models. The reason? Not enough computing power to go around.
Yes, even Google is running low.
Alphabet capped usage for several big customers because it's hitting capacity limits. Meta got hit the hardest.
Some of Meta's internal projects stalled. Now they're telling staff to use AI resources more "efficiently." (Corporate speak for stop hogging the bandwidth.)
Why was Meta using a rival's AI anyway?
Because Gemini was just better at certain jobs. Things like content moderation and scam detection.
When your competitor's tool beats your own, you swallow your pride and rent it.
But now the tap's been tightened. So Meta's leaning harder on its in-house model, Muse Spark, to cut the dependency.
Here's the part investors should care about.
The AI boom is shifting. The model isn't the prize anymore. The compute is.
Think of it like the gold rush. Everyone's hunting for AI gold, but the real money's in selling shovels, picks, and the only road into town.
Companies that own the data centers, cloud, and advanced chips? They hold the leverage now.
Companies without enough compute? They face higher costs, slower products, and an awkward reliance on the very rivals they're trying to beat.
Who do you think actually wins the AI race?
The weird twist: in this race, your fiercest competitor is also your supplier.
Meta has bet the house on AI. Zuckerberg's poured billions in, cut jobs, and shuffled thousands of staff into AI roles.
But all that ambition still needs somewhere to plug in.
TL;DR
Google limited Meta's access to Gemini AI models because of compute capacity constraints.
Meta was the hardest hit, with some internal projects disrupted.
Meta used Gemini because it beat its own systems at moderation and scam detection.
It's now leaning on its in-house Muse Spark model to reduce reliance on Google.
The bigger signal: computing power is becoming more valuable than the AI models themselves.
In the AI race, rivals are also each other's customers, and whoever owns the infrastructure holds the leverage.
1. Buy the Shovel Sellers
Compute is the new oil. The companies renting out AI capacity win regardless of who builds the best model.
📌 Action: Build a position in infrastructure names like $NVDA ( ▼ 1.64% ) (chips), $GOOGL ( ▼ 1.84% ) (cloud capacity it can now ration), or $MSFT ( ▲ 5.71% ) (Azure). Own the toll roads, not the cars.
2. Tilt Toward the Capacity Kings
Cloud giants with their own data centers and chips have pricing power now. Google just proved it can throttle a trillion-dollar rival. That's leverage you want on your side.
📌 Action: Overweight integrated players $GOOGL ( ▼ 1.84% ), $AMZN ( ▲ 2.5% ), $MSFT ( ▲ 5.71% ) that own infrastructure end to end. Trim pure-play AI developers who rent everything.
3. Watch Meta's Self-Reliance Story
Meta's pivot to its in-house Muse Spark model is the tell. If it cuts dependence and keeps spending billions, the AI bet stays intact. If projects keep stalling, that's a crack.
📌 Action: Track $META ( ▲ 1.36% ) through earnings for AI capex and model progress updates. Add on confirmation it's scaling its own compute successfully.
Short BTC. Long ETH. Hold as Long as You Want.
Kalshi is the first CFTC-regulated perps market in the US. No expiry, no rollover, up to 5.8x leverage on BTC, ETH, SOL, XRP, and more. Trade price direction without touching the asset.
Using leverage increases risk of loss. Leverage is subject to the Firm's review and the customer's risk profile.
There's a creative-software giant that just posted double-digit revenue growth, an AI business growing over 200% year-over-year, and an 89% gross margin that's the best in its entire industry. And yet the stock is down 48% in the last year.
Not because the business is broken. Because the market got spooked by AI fears that the actual numbers just flattened.
Here's the part that stops you cold: it trades at 7.4x earnings. Microsoft trades at 19.3x. Oracle at 13.6x. The sector average is 13.4x. Same software world. Wildly different price tag.
This is a company growing EPS 13–17% a year, sitting on a $25 billion buyback that could retire nearly a third of its shares, priced like it's going out of business.
Our analysis says a re-rate to a normal multiple implies…. 👀 Let’s just say it’s up to 90% upside from here.
In today's Premium+ deep dive, we break down:
Why this 48% selloff is a market mistake, not a warning sign
The exact valuation math behind our price target (and the number that would flip us bearish)
The one risk that could crack the whole thesis
What to watch as the AI business scales
🐂 Even The Bulls Are Scared Now
Remember that ceasefire that was supposed to calm everything down? Yeah… about that.
The U.S. and Iran traded fresh military strikes over the weekend. The fighting around the world's most important oil chokepoint is now on day three.
So prediction market traders did what traders do. They panicked, politely.
The odds got ugly, fast.
On Kalshi, the market betting on shipping returning to "normal" before September 1 cratered.
From 62.5% Thursday down to 45% now
"Normal" here means a seven-day average of more than 60 ship transits, based on IMF PortWatch data.
The longer-dated bets slipped too:
Return before Oct. 1: 64.8% → 53%
Return before Nov. 1: 75.9% → 64%
When every timeline gets more pessimistic at once, that's not noise. That's the market changing its mind.
What actually happened?
U.S. forces hit Iranian communications, air defense, drone storage, and mine-laying sites. Washington says this was payback for an attack on an oil tanker in the strait.
Iranian state media reported explosions in port cities Sirik and Bandar Lengeh, plus Qeshm Island. All conveniently home to military facilities.
Then Trump turned up the volume online, warning the U.S. could be "forced to militarily complete the job that we very successfully started."
Subtle.
The optimists are still out there. Barely.
Polymarket traders still give it an 80% chance traffic normalizes by the end of 2026.
But that's down from 90% Thursday. Even the optimists are quietly lowering their expectations.
The real fight isn't just military. It's about who's in charge of the water.
Iran now claims it has exclusive authority over traffic through the strait, citing a preliminary peace deal with Trump.
Foreign Minister Abbas Araghchi said it plainly: managing the strait is Iran's job, and "no other country or entity has any responsibility or authority in this matter."
The U.S. position? The strait is international waters and ships move freely, full stop.
Two countries reading the same agreement and arriving at opposite conclusions. What could go wrong?
Why should you care about a waterway you'll never sail?
Because roughly one-fifth of the world's oil moves through it.
Squeeze that pipe and you get higher oil prices, pricier shipping, and a fresh dose of market volatility. The kind that shows up in your portfolio uninvited.
It keeps escalating.
This round kicked off when Iran attacked a vessel using a route near Oman it doesn't like. Iran wants ships hugging its own coastline instead.
Then came strikes, counterstrikes, and Iranian attacks on targets in Kuwait and Bahrain. Planned U.S.-Iran talks in Switzerland, including on Iran's nuclear program, got delayed.
U.S. Ambassador Mike Waltz made the threat level clear on Fox News, saying Iran would be "sadly mistaken" to think Trump will just stand by.
The strait has partially reopened. But traffic is still well below where it was, and Iran can turn the threat back on whenever tensions spike.
TL;DR
Fresh U.S.-Iran strikes pushed fighting around the Strait of Hormuz into a third day, breaking a fragile ceasefire.
Kalshi traders slashed near-term "return to normal" odds from 62.5% to 43% for September 1, with later dates falling too.
Polymarket's end-of-2026 optimism dropped from 90% to 80%. Even the bulls are nervous.
Iran claims exclusive authority over the strait; the U.S. insists it stays open international water.
The strait carries ~20% of global oil, so disruption means higher oil prices and more market volatility.
Switzerland nuclear talks were delayed, and Iran can still threaten shipping whenever things heat up.
🐉 China Caught Up. Now What?
Remember when everyone assumed America would own the AI economy forever? China didn't get the memo.
Chinese developers are closing the gap fast, especially in cybersecurity. That's the takeaway from a Wall Street Journal report this weekend.
The star of the show: Zhipu AI's new GLM-5.2 model.
It's gotten scary good at one specific job: finding software vulnerabilities. Think of it as an AI bloodhound that sniffs out bugs in code.
It still trails the big US names like Anthropic and OpenAI on overall smarts.
But in bug hunting? It's punching way above its weight.
Why investors should care
The whole "US companies will dominate AI" thesis just got a stress test.
Here's the problem. Chinese models are getting better while staying cheaper and easier to access.
That combo squeezes the pricing power of American AI giants. Less pricing power means thinner margins, shakier market share, and valuations that suddenly look a little rich.
Who might win instead? Cloud providers and software firms that happily run any model, not just the premium American ones.
The open-weight twist
Unlike most top US models, GLM-5.2 is open-weight.
Anyone can download it, tweak it, and run it themselves. No permission slip required.
Businesses love that because it's cheap and flexible.
Security pros hate it for the same reason. The same tool that helps companies patch holes can help hackers find them. It's a lockpick that doesn't care who's holding it.
GLM-5.2 has already rocketed up the usage charts on OpenRouter, a platform that hosts hundreds of models. With some tuning, it reportedly creeps toward Anthropic's Mythos model on finding flaws.
China's AI is catching up. Where's your money?
The policy headache
Here's the irony nobody loves.
Washington is tightening the leash on US AI. OpenAI recently capped access to a new model pending a security review. Anthropic temporarily pulled back a high-end model to satisfy regulators.
Meanwhile, the Chinese alternatives stay wide open for anyone to grab.
See the trap? Lock down American models too hard, and global businesses just shrug and switch to the cheaper Chinese option.
Restrict your own team and the competition signs up your customers.
The Pentagon isn't sitting still, though. It's inking deals with US open-weight firms for government and classified work.
So the big question for policymakers and investors is the same.
Do export controls protect America's lead? Or do they just hand the world a reason to try China's stuff?
TL;DR
China is closing the AI gap fast, with cybersecurity as the breakout area
Zhipu AI's open-weight GLM-5.2 rivals top US models at finding software bugs
Cheaper, accessible Chinese models threaten the pricing power and valuations of US AI leaders
Open-weight means flexibility for businesses but a free toolkit for hackers
US restrictions on OpenAI and Anthropic models may push global users toward Chinese alternatives
The Pentagon is countering with deals for domestic open-weight AI