🎭Big Money, Bad Rollouts, and Backroom Deals

📰 Top of Feed: This Week’s Focus

🚀Perplexity Makes a $34.5B Play for Chrome (Before Google’s Hand Is Forced)

The Story: AI search startup Perplexity has submitted a $34.5 billion offer to acquire Google’s Chrome browser… a move perfectly timed just ahead of an expected antitrust ruling that could force Google to sell. If reading that sentence made you ask - “Umm… with what money?”, given Perplexity’s own valuation of ~$14 billion, you’re head is in the right place. Funded by “multiple large investment funds,” Perplexity says if the bid is accepted it will invest $3B into Chrome/Chromium over two years, keep most Chrome employees, and make no “stealth modifications.” The pitch: stability for users and advertisers in an era where browsers are becoming prime AI battlegrounds.

Key Takeaways:

• Perplexity’s bid comes as the DOJ seeks to break Google’s search monopoly, with Chrome a key bargaining chip.

• The startup, valued at $14-18B (depending on reports) after its last raise, doesn’t have $34.5B on hand, but claims outside investors will fully finance the deal.

• Promise to keep Chrome unchanged aims to avoid user backlash and antitrust headaches.

• AI players like Perplexity and OpenAI see browsers as the gateway for “autonomous agents” that shop, search, and perform tasks.

• Perplexity also recently tried (and failed) to merge with TikTok’s US arm under similar regulatory pressure.

The Real Talk: Browsers are no longer just about utility… they’re essentially the front-line territory for consumer facing AI. Perplexity’s move is less about nostalgia for tabs and bookmarks, and more about locking in a MASSIVE distribution pipeline for AI-driven search and automation applications. And, though this news is HOT OFF THE PRESS🔥 and this angle is completely personal conjecture, I suspect that this deal would also heavily strengthen the offering coming out of Perplexity’s experimentation in ads. If regulators force Google’s hand and the DOJ pressures the Alphabet monopoly to break up, whoever controls Chrome controls a direct channel to billions of eyeballs AND an immense amount of user browsing data. This is Big Tech Jenga: pull out the wrong block and the whole structure wobbles. While I am skeptical of the likelihood of this sale, if nothing else Perplexity just made sure it’s a part of the conversation. Hats off to their teams for a very smart marketing play!

🧧Policy & Power Moves: Nvidia & AMD Agree to Pay Trump 15% on China Chip Sales

The Story: In a very unusual arrangement, Nvidia and AMD (two major players in the semiconductor industry/GPU manufacturing space) have struck deals with the Trump administration to pay the US government 15% of revenue from AI chip sales to China in exchange for export licenses. The move could bring over $2B to US coffers by year’s end… but critics warn it risks national security and sets a troubling precedent.

Key Takeaways:

• Nvidia expects $15B in H20 chip sales to China by year-end; AMD about $800M for its MI308 chips.

• The Commerce Department approved export licenses just two days after Nvidia’s deal.

• Supporters frame it as pragmatic industrial policy; detractors see it as pay-to-play.

• China hawks warn even “lesser” chips could still aid Chinese military AI capabilities.

• Raises questions about whether tariffs and export restrictions are about security or revenue generation.

The Real Talk: This deal really blurs the line between trade policy, protection, and extortion. On paper, the US gets billions while keeping a hand on the export tap… But in practice, it shows just how far the administration is willing to bend when the economic upside is tempting AND how much leverage top chipmakers have. It’s also hard to not question the potential “pay-to-play” scheme that appears to be getting set up. It’s the kind of arrangement where the government’s approval isn’t just about security, but rather about testing who’s willing to cut them in. Chipmakers like Nvidia and AMD hold enormous leverage in the global AI race (a la old adage about selling picks & shovels in the Gold Rush), but this move flips the dynamic: the administration is essentially saying, we’ll let you compete… for a fee. Nvidia’s Jensen Huang argued that banning sales would only boost Chinese rivals like Huawei so the additional “chip tax” is justifiable, but let’s see if other countries start to adopt similar playbooks where 15% becomes the price of admission to sensitive markets. Time will tell!

🫠ChatGPT-5’s Messy Rollout

The Story: OpenAI CEO Sam Altman has admitted GPT-5’s launch was “bumpy,” with faulty model routing, inconsistent performance, and a confusing upgrade that removed older models from ChatGPT. Complaints from users (including errors in basic math, logic, and code) led OpenAI to restore GPT-4o access and promise UI changes for greater model transparency.

Key Takeaways:

• New “auto-router” system failed for hours, making GPT-5 seem less capable than intended.

• Users report worse results than with older models in some coding and reasoning tasks.

• GPT-4o reinstated for Plus users while OpenAI gathers more performance data.

• Developer feedback includes comparisons showing rival Claude Opus 4.1 outperforming GPT-5 in certain cases.

• OpenAI plans to double rate limits and tweak infrastructure as API traffic spikes.

The Real Talk: The AI hype cycle is so fast moving that one bad launch day can open the door for competitors. I suspect GPT-5 will likely still prove to be a leap forward with time, but right now this roll out has been a useful reminder that bigger models don’t automatically mean better user experience. In the race for AI dominance, reliability may be the most underrated feature. The tech world loves to conflate “more parameters” and additional product upgrades with progress, but for most users, the scoreboard isn’t about intangible benchmarks… it’s about whether the AI in front of them works and feels better today than it did yesterday. If it seems slower, dumber, or more erratic, all the “most capable model yet” claims in the world won’t matter. OpenAI’s scale gives it room to stumble, but even tech giants shouldn’t take this headway for granted. If they would truly like to achieve global scale and to become the default AI partner across industries, I personally think OpenAI should be a little more careful with their marketing and growth strategy. No one likes to deal with the headache of a poorly received product roll out (speaking from personal experience😅).

🔍 The Decoder Lens: Power Shuffles, Public Fumbles, and the Price of Access

The gatekeepers of the digital age are being tested. And not always on their own terms.

In the browser wars, AI startups aren’t just building their own roads; they’re striving to buy the entire highway. Perplexity’s $34.5B bid isn’t about nostalgia for bookmarks. It’s a calculated attempt to insert themselves into the default way billions of people access the internet.

In geopolitics, chipmakers are discovering that “market access” now comes with a toll booth run by governments who are as interested in revenue as they are in security. Nvidia and AMD’s 15% payment deal to sell in China is either pragmatic industrial policy or a dangerous precedent that could make sensitive tech a bargaining chip (pun very much intended).

And in AI, the industry darling and loudest name in the room (OpenAI) is reminded that hype cycles can turn fast when real-world performance doesn’t match the keynote slides. GPT-5’s messy rollout shows how fragile dominance can be when users feel like they’ve been downgraded instead of upgraded.

In all three cases, an interesting foundation is in flux: who gets to set the defaults. Whether it’s a startup angling to own your entry point to the web, a government monetizing your ability to sell cutting-edge chips, or an AI giant deciding which models you’re “allowed” to use, the players who adapt quickly to this shaky environment will shape the next era of digital power. The ones who don’t? They’ll be playing catch-up in someone else’s ecosystem.

🌟 Editor's Note

Between spontaneous San Francisco trips, family obligations, and preparing for the upcoming tech networking event of the year (who will I be seeing at Burning Man? 👀) the newsletter may be a little inconsistent over the next few weeks! You also heard this hear first (newsletter EXCLUSIVE) but yours truly will be moving her home base out of NYC at the end of September. All this is to say - there have been a LOT of moving pieces going on these days, not all which I have been able to share🤭, but I appreciate you for being here and I can’t WAIT to be in a routine conducive to more consistency come October.

And on that note, I’m hiring!!! If you are interested in a Part Time editing opportunity, please apply here: https://forms.gle/wVw7sbApMnCv2ebq9

If you could take control of any tech product or platform for a day (just to see what’s really going on behind the scenes) what would you pick, and what’s the first thing you’d change? Reply and let me know - I’m so curious to hear how you’re thinking about the space!

-Feed Decoder

P.S. Somewhere, an AI is flubbing algebra right now. Wishing it well!