Apple's foldable iPhone is expected to cost about $2,000, and it arrives the same week John Ternus takes over as chief executive from Tim Cook.
For a decade and a half, Apple's big product events mostly refined the iPhone. MG Siegler says the slate now in the pipeline — a folding phone, a touchscreen MacBook, a smart-home hub with a screen, and AirPods with cameras still to come — is Apple finally willing to ship products that could fail.
"That feels more like co-CEOs to me than it does like one person's the CEO and the other person's the executive chairman."
Siegler spent over a decade as a general partner at GV, Google's venture-capital arm, before leaving to run the newsletter Spyglass — a vantage point on both the Apple product cycle and the Google turmoil this episode works through.
I listened to the full interview so you can skip it. 1 hr 2 min of audio, 14 minutes of reading.
Here are the 8 takeaways that matter.
👤 Guest: MG Siegler, who runs the newsletter Spyglass and spent over a decade as a general partner at GV, Google's venture-capital arm, after covering tech for VentureBeat and TechCrunch
🎙️ Host: Alex Kantrowitz, who wrote BuzzFeed News's Tech Giant Update, authored the book Always Day One, and now runs Big Technology, a newsletter and podcast about Amazon, Apple, Google, Meta and Microsoft
📰 Published: 9 September 2026
🔴 YouTube | ⏱️ 1 hr 2 min | ✅ Time saved: 48 min
Key Takeaways
Apple's foldable "Ultra" is expected to cost around $2,000, and Siegler is buying one anyway
He's owned three generations of the Pixel Fold and calls its shape "a little bit awkward"
Tim Cook's pay package looks more like a co-CEO deal than a retirement plan
Ternus gets a $55M RSU target for 2027 with performance conditions attached; Cook gets $45M with fewer
A touchscreen MacBook and a smart-home hub were reportedly held back for years because Siri wasn't good enough
Apple's 2027-and-beyond roadmap adds AirPods with cameras, smart glasses, a home security system, and a tabletop robot
Siegler frames it as Apple using its hardware and retail advantage instead of chasing AI software alone
Google's Gemini 3.5 Pro, promised for June, still hasn't shipped — and Siegler thinks it's dead
A SemiAnalysis post he cited argues Google Cloud, not Gemini, is now winning the internal fight for compute
Google may be choosing to sell AI compute rather than keep chasing the AI frontier
Its roughly $4T market cap trails Nvidia's $5.5T, and the strategy mirrors Microsoft's and Amazon's cloud-neutral playbooks
OpenAI's AGI claims around GPT-6 and Astra are, in Siegler's words, "full-on marketing now"
He ties the shift to OpenAI no longer having an AGI clause governing its Microsoft relationship
Both Anthropic and OpenAI may want a splashy moment before their IPOs
Chatter includes the idea that Anthropic could time a scientific breakthrough to its offering
1. The $2,000 Foldable iPhone
Bloomberg's Mark Gurman previewed Apple's September 9 event as the biggest product slate the company has assembled in 15 years, centered on a foldable iPhone many inside Apple reportedly call the iPhone Ultra: a large interior screen, a smaller external display, Touch ID, split-screen multitasking, and a price around $2,000 — about the size of a passport when folded.
Siegler thinks it will be a hit. He's tested three generations of Google's Pixel Fold as his own Android backup device, and prefers Apple's reported passport-style shape to Google's narrower, taller one: "I think the Pixel Fold to me felt a little bit awkward in that it feels like it's very sort of narrow when you're holding it folded straight on." He expects the unfolded size to land close to an iPad mini — good for reading and video — and coined a nickname for the split personality on the spot: "the mullet iPhone," business up front when it's folded, a party once it's open. Kantrowitz agreed, adding his own line: "I've always said that people with folding phones they just look happier than people with phones that don't fold."
The business case leans on precedent. Samsung's 2021 foldable push lifted its phone sales 6% and revenue 11%; Motorola's foldable launch pushed its sales up 21% and, Kantrowitz recalled, its average selling price up around 17%. Apple's own iPhone sales had been roughly flat through the iPhone 17 generation before climbing — by Kantrowitz's account, up about 21% year-over-year last quarter and 20% the one before. A $2,000-plus phone raises the ceiling on how much revenue a single device can generate, which Kantrowitz suggested may be exactly the story Ternus wants to tell Wall Street as he takes over.
Both hosts expect durability headaches at Apple's scale — Samsung's early foldables drew backlash over screen and hinge failures — and Siegler predicted new versions of the same complaints, plus opportunistic lawsuits, once tens of millions of units are in the field, alongside a busy AppleCare pitch given how much more screen there now is to break. Apple is also adding a feature to transfer a phone number between devices in software, which Siegler read as a hint that some buyers might keep both a foldable and a regular iPhone rather than switching entirely — though the foldable is expected to lose out on camera quality to the Pro line, purely because of how little space a folding body leaves for optics. Asked directly which one he's buying, Siegler committed to the fold, even though he'll miss his current phone's color. Kantrowitz is holding off, sticking with his iPhone 17 for at least one more year.
2. The Ternus Pay Package
Kantrowitz framed Ternus's central job simply: sell iPhones, since the phone is more than half of Apple's total sales. That won't be easy at the start — a memory-chip shortage has already pushed Apple to raise laptop prices, with phone price increases still to come — and Ternus, a hardware engineer by background, isn't the supply-chain specialist Cook is. Cook is staying on as executive chairman and, Kantrowitz said, remains especially important for geopolitics, including whatever relationship the White House wants to keep with Apple's leadership.
On Apple's own AI position, Siegler said the company is in a fairly good spot after a rough stretch: recent iOS and macOS betas are stable, and Siri — long mocked for failing basic tasks — can now handle things like reciting sports scores. He doesn't expect it to match OpenAI or Anthropic's Claude on raw capability, but said it will be "good enough for the masses," calling functional on-device AI table stakes heading into the new hardware cycle rather than a real differentiator yet.
Then came the pay numbers. Bloomberg reported Cook's salary is dropping from $3 million to $2 million, but his 2027 restricted-stock-unit target is $45 million against a $55 million target for Ternus — figures Kantrowitz said "feels more like co-CEOs to me than it does like one person's the CEO and the other person's the executive chairman." Siegler's read is that the two packages aren't equivalent: "So in other words like Ternus has major targets to hit in terms of the share price and different milestones that sort of unlock a lot of those." Cook's award, he said, carries fewer performance conditions — mostly he just has to stay to collect it — a contrast Siegler ties back to the previous executive chairman, Arthur Levinson, who by Kantrowitz's account received well under $1 million a year and wasn't nearly as active day to day. Siegler thinks the size of Cook's package reflects real, ongoing work, especially geopolitical relationships Apple would rather not discuss publicly. He closed the point by sharing a meme he couldn't stop thinking about: "Steve did the job. He let Tim cook. Now it's John's turn."
3. The Rest of the Launch
Apple's iPhone 18 Pro and Pro Max, by Gurman's preview, look similar to their predecessors apart from two new colors, but pack new A20 chips, upgraded modems, better battery life, and vapor chambers to fight overheating, plus a camera bump — a mechanical aperture and, on the Pro Max, added depth control. Siegler is torn about giving up that camera for the fold: as a father who uses his phone mainly to photograph his kids, he doesn't want to trade down, even for a device he otherwise wants.
A touchscreen MacBook is also coming — Apple's biggest MacBook Pro redesign in five years, with an OLED display and the dynamic-island element carried over from the iPhone. Apple has spent years publicly insisting it saw no need for a touchscreen Mac, so Siegler expects an awkward marketing pivot on top of a harder problem: macOS was built for a mouse and keyboard, and adapting its interface for touch is a real design challenge, not just a hardware one. He wondered whether Apple brands it a "MacBook Ultra" and whether it starts eating into iPad Pro sales.
Apple is also expected to ship its first home hub with a screen — a device with a roughly 7-inch display, Siri, and facial recognition for personalized content, sitting on a nightstand or mounted to a wall. Siegler said the product had been shelved for years because Siri wasn't ready: "Siri was so bad that they had to postpone the release of this and maybe a few other devices, including new versions of the HomePod and Apple TV." Nicknamed the "HomePad" in early reports, the device is expected to finally add visual user-switching — letting different family members get personalized content on a shared screen, something Apple has offered on the Mac for years but never on an iPad. Siegler sees it as the base layer for a broader push into the smart home, including rumored security cameras.
4. The 2027 Wearables Bet
Beyond this week, Kantrowitz laid out Apple's reported pipeline for 2027 and after: AirPods with cameras, an in-home security sensor system, smart glasses, a tabletop robot version of the home hub, and a camera-equipped pendant. He called it a sharp break from the Apple of the last 10 to 15 years — a company now willing to ship things that might not work.
Siegler drew a parallel to Microsoft's transition from Steve Ballmer to Satya Nadella, when Office for iPad shipped almost immediately and signaled a culture shift that had actually been years in the making internally. He credited Ternus specifically for a lot of the work now surfacing — the foldable iPhone, the touchscreen MacBook, and what Siegler called the MacBook Neo — as products Ternus has personally overseen. The unreleased wearables are the bigger unknown, since they're new categories for Apple rather than new versions of something it already sells, and Siegler said "there's a lot of ways that those can go wrong."
Kantrowitz said this is the direction he'd been hoping Apple would take: "We're no longer living in a slowmoving technology ecosystem. Things are faster." Siegler agreed the shift is tied to Apple's own AI stumble — an early attempt at AI-first products that didn't work out, forcing something closer to a reset — after which the company concluded its real edge is elsewhere: in-house hardware expertise and a retail channel no other AI company has, both of which point toward shipping more physical devices rather than trying to out-build OpenAI or Google on software alone.
5. Google's Frontier Retreat
Last time Siegler was on the show, the two discussed reports that Google was quietly giving up the frontier AI race. Days after they finished recording that episode — before it had even published — Demis Hassabis stepped down as CEO of Google DeepMind. Kantrowitz read from a SemiAnalysis post arguing the company is no longer a frontier lab, pointing to departures from Google's reinforcement-learning teams and poor compute allocation, and predicting Google Cloud is the real winner: "Whereas Gemini and Google Cloud Platform used to desperately fight for compute allocation, it's now clear that Thomas Kurian, the head of Google Cloud Platform, has won."
Siegler, who spent over a decade at GV, Google's venture-capital arm, before leaving a couple of years ago, called the take somewhat overzealous — Google hasn't directly disputed it, he said, though he no longer has any inside information. He does think one specific promise is broken: Gemini 3.5 Pro, unveiled on stage at Google's I/O conference and promised for release the following month, has never shipped. "We still have no 3.5 Pro. They haven't come out and said that it's dead, but I think it's dead," he said. Google's public focus has shifted to Gemini 4, still in training, while its Flash-tier models are well reviewed but not frontier-class. He drew a parallel to Meta's Llama, which stumbled publicly before Meta reset the entire team — though he doesn't expect Google to go that far — and said reports suggest DeepMind's leadership had bet on "world models" as a different direction from the pure large-language-model race, a bet that may simply have come too early.
6. Google Bets on Compute
If Google isn't racing for the frontier anymore, Kantrowitz asked, is the company making that trade on purpose — swapping frontier AI research for compute sold as a service? He compared it to Microsoft's own posture: after renegotiating its OpenAI deal, Microsoft has let Mustafa Suleyman's team chase frontier AI without releasing anything at that level yet, while chief executive Satya Nadella talks about being neutral ground for any AI model, so long as it runs on Azure. Amazon runs a similar playbook with AWS.
Siegler agreed Google is uniquely positioned to make that bet work: it has in-house TPUs and chip design, and, critically, a search business profitable enough to fund the buildout — an advantage he said even Amazon doesn't have despite AWS being a major profit center. He credited Sequoia Capital's David Cahn, who raised the underlying idea on the show in August, with the logic: since AI models may become interchangeable while compute stays scarce, a company that owns the compute earns money every time any model is used, no matter whose model it is. Kantrowitz put a number on the opportunity: Google's market cap, near $4 trillion as they spoke, still trails Nvidia's roughly $5.5 trillion, and Anthropic's expected IPO alone could value it near $2 trillion. Siegler named bureaucracy as Google's real risk, but he said not to count the company out: "it feels like they'll be back at some point soon," arguing that these companies tend to cycle between being written off and roaring back.
7. OpenAI's AGI Marketing
OpenAI released its GPT-6 model, branded Astra, and Nvidia chief executive Jensen Huang called it AGI — months after making the same claim about an earlier model with, Kantrowitz noted, little notice. Siegler's read: "My general read on AGI of Greg Brockman proclaiming it with Astra is that it's full-on marketing now." He said OpenAI's cofounder and president can make the claim more freely now that the company no longer has an AGI clause governing its relationship with Microsoft, so declaring AGI carries no contractual consequence — only a public-relations one.
That carries its own risk, Siegler said: years of hype that AGI would transform everything have set up a letdown, since capability hasn't obviously caught up. Only half-joking, he said, "I still can't get grandma home from the airport." He did point to one genuinely interesting demo — a builder who set up a simulated world where AI agents began talking to each other and building their own simulations inside it — but called the overall situation "a little bizarre." He wondered whether the AGI claims are timed to Anthropic's approaching IPO, since OpenAI's Astra push doubles as competitive marketing against a rival about to go public first.
8. Racing Toward Two IPOs
Kantrowitz sees OpenAI's momentum as a comeback: the company was in what he called "code red" roughly nine months earlier, playing catch-up to Google and then losing ground to Anthropic as Anthropic pulled ahead on coding tools. Astra, in his read, is OpenAI clawing some of that back right as both companies approach the public markets — Anthropic's S-1 could land within weeks, with an offering as soon as next month. Anthropic, for its part, just shipped new Mythos and Fable releases ahead of OpenAI's own launch.
Siegler expects both companies to want a splashy moment before they file — he mentioned chatter that Anthropic could time some kind of scientific breakthrough, even a Millennium Prize-level result, to the IPO window. He also flagged a tension neither company has resolved: publicly declaring AGI is here while also publishing detailed warnings about AI security risks, which he said makes it harder by the week to know what to trust. Only half-joking, he added: "I want to know who is going to choose AGI as their stock ticker symbol and who gets stuck with RSI."
By the end of the conversation, Siegler's view was that Apple has decided the safest move is to start taking risks again, while Google may have decided the safest move is to stop chasing the AI frontier and instead sell everyone else the compute to get there.
Bonus Insights
Kantrowitz opened by flagging that the conversation might already be stale by air time — the episode was recorded the Monday before a Wednesday release — and joked that "none of this will hold for Wednesday."
Siegler predicted the news would outrun the show anyway, joking that the iPhone Ultra would probably turn up "left in a bar or something" before Apple's own event.
Extending the stock-ticker joke about OpenAI and Anthropic's IPOs, Siegler suggested a smaller company could squat the AGI ticker itself and charge OpenAI to use it, "like a jersey on a sports team."
The two skipped over Apple's prior "Liquid Glass" software redesign entirely, with Kantrowitz waving it off as not worth revisiting.
Products, Companies & Tools Mentioned
Apple (Previewing its foldable "Ultra" iPhone, a touchscreen MacBook, and a Home Hub with a screen at the September 9 event, plus AirPods with cameras, smart glasses, a security system, and a tabletop robot further out)
Samsung (Its Galaxy Z Flip aspect ratio is the comparison point for Apple's foldable; Samsung's own 2021 foldable push lifted sales 6% and revenue 11%)
Google (DeepMind CEO Demis Hassabis stepped down; Gemini 3.5 Pro, promised for June, still hasn't shipped, and Siegler thinks it's dead)
The Pixel Fold, which Siegler has owned across three generations and calls "a little bit awkward" folded
Google Cloud Platform, run by Thomas Kurian, which a SemiAnalysis post called the "biggest beneficiary" of Hassabis's exit
Microsoft (Renegotiated its OpenAI deal and is positioning Azure as neutral ground for any AI model — a "Switzerland" strategy Google may now be copying)
The Surface Duo, an early foldable phone Siegler called "probably ahead of its time"
Amazon Web Services (Runs a similar model-agnostic cloud strategy, building its own AI chips without chasing the frontier)
Nvidia (Jensen Huang called OpenAI's GPT-6/Astra "AGI"; its roughly $5.5 trillion market cap is the gap Google's compute strategy is aimed at closing)
OpenAI (Released GPT-6, branded Astra, and no longer has an AGI clause limiting its Microsoft relationship — freeing it, Siegler said, to declare AGI more freely)
Anthropic (Expected to file for an IPO that could value it near $2 trillion; released new Mythos and Fable updates just ahead of OpenAI's Astra launch)
Meta (Its Llama model line stumbled the same way Gemini has, a parallel Siegler drew before saying he doesn't expect Google to reset as drastically)
Books & Resources Mentioned
Spyglass – M.G. Siegler (His own newsletter on technology and media, mentioned repeatedly as the source of the Apple and Google analysis discussed in this episode)
Gemini is Cooked but GCP is Cooking – SemiAnalysis (The post Kantrowitz read from at the top of the AI segment, arguing DeepMind is no longer a frontier lab)
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: