Meta's new AI agent, Muse, added an estimated $80 billion to $90 billion to the company's market cap in a single day this week.
Rivals had already written Muse off as a copycat of xAI's Grokbot. Hamid Shojaee argues being second barely matters in AI right now, because most of the people who will ever use these tools haven't signed up for any of them yet.
"So immediately this product just added$80 or$90 billion worth of market cap to Meta's market cap."
Shojaee sold two software companies, Axosoft and Pure Chat, before turning full-time to investing and building Earnings Hub, the stock-analysis tool he pulls up live on the show to score names he's never looked at before; co-host Dustin Alper is Earnings Hub's co-founder and director of product. Meta is now 18% to 19% of Shojaee's own portfolio.
I listened to the full episode so you can skip it. 1 hr 7 min of audio, 16 minutes of reading.
Here are the 7 takeaways that matter.
🎙️ Hosts: Hamid Shojaee, founder and CEO of Savvy Trader and Earnings Hub who previously built and sold Axosoft and Pure Chat, and Dustin Alper, co-founder and director of product at Savvy Trader and Earnings Hub
📰 Published: 9 September 2026, on the show's own feed
🔴 Episode page | ⏱️ 1 hr 7 min | ✅ Time saved: 51 min
Key Takeaways
Meta's Muse added $80 billion to $90 billion to the company's market cap in a single day
Hamid used the pop to add to his Meta stake, which is now 18-19% of his portfolio
Meta doesn't need to be first in AI, because most of its eventual users haven't signed up for anything yet
Hamid points to Anthropic, which entered late and still became the AI revenue leader within a year
Dustin sold another 17% of his Robinhood position at $123, mostly to cut crypto exposure
A full Tesla robotaxi rollout would barely dent Tesla's income statement
60,000 cybercabs driving 100 miles a day at $1 a mile works out to about $1 billion, roughly 1% of Tesla's revenue
Apple's $2,000 foldable iPhone Duo might not make it to a fifth birthday
Dustin thinks the odds of Apple canceling it within five years are higher than it becoming a meaningful revenue line
ChatGPT's Astra scores almost the same as Anthropic's Opus on intelligence for close to half the price
$3.26 per task on Astra versus $5.86 on Opus, per Artificial Analysis' benchmarks
Micron's forward P/E of 7 is why Hamid owns it over Nvidia and Broadcom
Broadcom's forward P/E is 22, Nvidia's is 18 and growing faster — Micron is still the cheapest of the three
1. Meta's Muse and the S-Curve
Hamid added to his Meta position twice this week: once right after Muse launched, before the stock had reacted, and again the next day once Meta said adoption was beating its own expectations. Between the two purchases he added just under 10% to his position, bringing Meta to roughly 18-19% of his portfolio — his single largest holding.
Dustin framed the reaction in market-cap terms:
"So immediately this product just added$80 or$90 billion worth of market cap to Meta's market cap."
He thinks that undersells it: Meta's older businesses — Instagram, Facebook and WhatsApp — are worth more on their own than the company's current total valuation, in his view, which would make the AI business essentially free at today's price.
Hamid's framing for why he's excited runs through what he calls the S-curve of market adoption: every new technology category grows from zero users to near-saturation along the same curve, and the best time to enter is on the early, steep part of that curve — after a pioneer has proven the category exists, but before most of the eventual users have signed up for anything. "Tesla essentially built the EV market, OpenAI has built the AI market." Anthropic, he said, got in late and still became the AI revenue leader within about a year, because it didn't need to convert an existing user away from a rival — it only had to win a larger share of the millions of new people coming online for the first time. Meta, in his read, is entering AI in that same window, with a product already getting good reviews and a user base — 3.6 billion daily actives — that no AI-only competitor can match.
Being a "me-too" product, in other words, isn't fatal this early in a market's life:
"Nothing that can't be fixed, but right now Grokbot is very much focused on the professional user, and MetaMuse is focused on the personal use case of a user."
Hamid's own criticism of Meta's launch is the audience it's chasing. "And while there are definitely personal use cases, my issue is there are way more digital tasks that need to be done in a work setting than there needs to be in a personal setting." He'd rather Meta had opened with the assistant handling triage, bug-catching and Slack management than trip planning — a use case he thinks nobody actually needs often enough to matter. He expects Meta to bring enterprise billing and integrations later, and once it does, Meta's distribution — he estimated roughly seven to ten times the reach of X, xAI's own distribution channel — should give Muse a real edge even arriving second.
Muse's design also drew some real-time debate on the show. It gives users a persistent "goals and tasks" checklist and a customizable news feed alongside the core chat, plus separate side chats for different topics — a structure Dustin hadn't seen an AI product try before. Hamid was ambivalent: individual chats make it easier to find old conversations now, but he isn't sure how the system will handle search and continuity if users mostly stay in one long-running conversation instead. He also likes that Muse can act on a purchase directly — recommending a product and then buying and shipping it — rather than stopping at a recommendation the way ChatGPT does today, calling that extra "agency" a plausible reason casual users switch over, especially while the free tier lasts.
2. Dustin Trims Robinhood
Dustin made his first portfolio move in about two months, cutting Robinhood by roughly 17% at $123.
"So I sold a little bit of Robinhood, specifically around 17% when it was at$123."
The trade was mostly about reducing crypto exposure rather than a call on Robinhood's business, and it followed a roughly 15% one-day jump in the stock that he said felt like a natural point to take some profit — he isn't aware of major news behind the move beyond a possible analyst upgrade. He's already sold most of his position lower, in the low $100s, so most of his cuts came before this week's run-up. With an election year ahead and volatility likely, he said he wouldn't be surprised to start buying Robinhood back if it pulls back, though he isn't in a hurry — his next real signal will be the company's earnings.
3. Tesla's Cybercab Math
Tesla rolled its Cybercab out in Austin this week in what Hamid called an oddly muted event — not livestreamed, journalists under embargo, and little actually revealed on stage. The bigger question the co-hosts dug into is who ends up owning and operating these vehicles: Tesla itself, individual owners who opt their personal cars in and out of the network, or regional partners similar to Amazon's delivery-driver network.
Dustin expects a hybrid, weighted toward Tesla: in dense cities Tesla will likely run its own fleet because charging and cleaning infrastructure is a real barrier to entry, but "So it might be dynamic from that standpoint, but most of the time, I think Tesla's gonna have it covered." He sees more room for individual or small-fleet owners in areas Tesla doesn't build out its own charging and cleaning depots.
Hamid is more skeptical that individual owners factor into Tesla's plans at all, and doesn't think the popular Airbnb comparison holds up:
"So none of those things would be possible for these cyber cab vehicles if the only thing you can do with it is to enter and exit the Tesla network."
Unlike an Airbnb host's house, which has value with or without the platform, a cybercab that can only earn money inside Tesla's own network has no independent use — closer to a depreciating Uber car than an appreciating Airbnb asset, and one an owner is buying purely to put to work rather than driving themselves. Each vehicle also needs its own city-by-city regulatory approval, the same bottleneck that already caps how many robotaxis Waymo can put on the road, so an individual can't simply buy a cybercab and start earning without clearing that hurdle. Hamid's guess: within ten years, Tesla runs the network itself rather than dealing with the liability and pricing-control headaches of outside partners, since it has more than enough of its own capital — a company with a $1.5 trillion valuation, he noted, could raise the cost of half a million cybercabs by selling roughly 1% of its own stock.
He also thinks the hype around robotaxi revenue is running well ahead of the math. If Tesla was able to deploy 60,000 cybercabs by the end of 2027 — "if it was able to reach 60,000 cybercabs by the end of 2027, that would be phenomenal. That would be, by the way, 20 times more than Waymo already has, right?" — and each one drove 100 miles a day at a dollar a mile, that comes out to roughly $1 billion in incremental revenue.
"That's not a lot of revenue for a company that already has$100 billion of revenue. That's a 1% increase in revenue."
By his math, robotaxis don't meaningfully move Tesla's numbers — or take a meaningful bite out of Uber's — until well past 2028 or 2029.
4. Apple's $2,000 Foldable
Apple's event brought new iPhone 18 Pro and Pro Max models, AirPods 5, a new Apple Watch and Watch Ultra, and price increases on the older iPhone 16 and 17 models. Hamid flagged the increase directly:
"They increased all the prices by$100."
The headline product was the foldable iPhone Duo, unveiled as Apple's traditional "one more thing" — even though, as Dustin pointed out, the name had already leaked before the keynote even started. Starting price is roughly $2,000.
"And I understand the starting price here is like$2,000."
Neither host is convinced by the pitch. Dustin's problem is the format itself: the unfolded screen is wider but not taller, so video content doesn't actually fill more of the display, and Android's own foldables — running $1,200 for the flip-phone style and $1,800 for the tablet style — haven't found much of an audience beyond enthusiasts. Still, he doesn't rule out Apple succeeding where Android failed:
"Because typically, whenever Apple does it, people just kind of come."
"But Apple could have easily asked for more." He compared it to the original $1,000 iPhone X, a price that looked shocking at launch and became the new normal within a couple of years — though he isn't convinced the Duo has an equivalent case for existing at all once the price passes what a separate iPad plus a normal iPhone would cost. His base case: the odds Apple cancels the Duo within five years are higher than the odds it becomes a meaningful line of revenue.
Hamid is a bit more open to it doing decently, betting that novelty — a half-folded mode usable like a small laptop — and status-symbol appeal will move some units regardless of the format's practical case. He also flagged a bigger structural shift: "The other call out is they did not announce a new iPhone 18." Only the Pro line got refreshed this cycle. Dustin thinks this confirms rumors that this year's memory shortage pushed Apple onto a new release cadence — a Pro-focused event in September, and a separate standard-iPhone event roughly six months later.
5. Astra Beats Opus on Cost
OpenAI's Astra shipped last week, faster than expected — Dustin joked their prior episode's skepticism may have added pressure: "So I think our podcast added some pressure. At least that's what I like to think."
Hamid ran the numbers on Artificial Analysis' intelligence-and-cost charts, comparing his current daily driver, Anthropic's Opus, against Astra.
"The intelligence score for Astra is 53 versus Opus 51."
"Claude Fable 5.1 is$7.63, and Astra is$3.26. So significantly cheaper with basically the same intelligence."
By that comparison — Opus runs $5.86 per task on the same index — Astra edges out Opus on intelligence at close to half the price, which Hamid said means he should probably be switching his default model. On the cheaper end of his usage, he compared xAI's Grok 4.6 against Meta's new Muse Spark 1.3: "But muse is cheaper than grok and smarter at$1.60 versus$1.86." That leaves him planning to switch both of his everyday models — Opus to Astra, and Grok to Muse Spark.
Speed told a different story, Hamid said: "So Muse kills it here, but this is where Astra lags, at least compared to Fable." Muse ran 226 tokens per second against roughly 52-67 for Astra, Fable and Grok, which were bunched together on the slow end. "The fastest model on this chart, if you're not watching, is Gemini, which I can attest is very fast. But it doesn't have as good of an intelligence as Muse." Taken together, Dustin summed up the whole comparison: "So like it almost looks like Muse might be the best cost speed intelligence combo." Hamid agreed, adding, "That's impressive, considering they basically had nothing a month ago" — a reminder, as he put it, that this market is still early enough that the leaderboard can flip in weeks.
6. Mailbag: Iron to Rivian
Listener questions covered five separate names.
Iron. Dustin's view on the Bitcoin-miner-turned-AI-neocloud is unchanged from a deeper breakdown he did the week before: "Right now their target annual run rate by the end of this year is$4 billion. Their last year of revenue is$707 million." Extrapolating from the pace of the company's per-megawatt contract signings, he sees a plausible path to much more: "So my own math is next year, end of year, we could be seeing an annual run rate of 15 billion to 19 billion dollars." He flagged that estimate as his own, caveated math, contingent on the AI buildout continuing.
Options. Asked about his own use of options, Hamid said he learned the hard way not to trade them for speculation: "There was a time when I thought that I could time the markets well, and options slapped me around enough to realize that I can't time the markets well." He now only buys calls on stocks he already owns, as a way to defer committing full capital, and avoids selling covered calls after getting burned selling shares he'd have preferred to keep once the stock later ran past his strike price.
Sterling Infrastructure (STRL). Pulling the stock up cold on his Earnings Hub tool, Hamid found: "It has a current PE of 31, a forward PE of 23, which means it's growing pretty rapidly. 61% on top line revenue, and 100% on EPS is its growth rates for trailing 12 months." Despite the growth, he wasn't drawn in — construction and civil-infrastructure fundamentals aren't his area of interest next to names like Micron or Meta, and the next couple of quarters are projected to show declining revenue.
Apple's Crash Call. Responding to a listener worried that Apple Intelligence isn't working properly in Europe, Dustin noted the feature barely works well anywhere yet and suspects Europe's data rules, not Apple's product, are the real constraint. Hamid pushed back harder on the premise that Apple is about to crash: "Apple hasn't been growing that fast for the past several years," yet the stock has kept performing regardless — a pattern he's seen the market sustain for years before it corrects. His example: "AOL at one point had a higher market cap than Microsoft." He bet against AOL with puts in the late 1990s and lost for years before finally being proven right: "And for the next two, three years, I lost many thousands of dollars on buying puts on AOL, which expired worthless every time as AOL just reached new highs, and it's just unbelievable."
Rivian and Optimus. On whether Rivian should build its own humanoid robot to compete with Tesla's Optimus, Dustin disagreed with the premise — Rivian isn't yet profitable and should stay focused on vehicles and self-driving. He noted Rivian already spun its robotics ambitions into a separate company: "And Mind Robotics is going to be focused on manufacturing robotics." Hamid added a wrinkle: "I believe RJ is also the CEO of Mind Robotics, so it might be a little bit of a distraction for him, but not necessarily at the company level." On autonomy specifically, Dustin said Rivian is already building toward robotaxis: "Oh, well, they are working on autonomy and they've partnered with Uber as potentially one of the first customers for their autonomous vehicles," though full self-driving is still a ways off.
Asked to weigh Uber against Tesla directly, Dustin exited his own Uber position about a year ago, in the low $90s, over the coming robotaxi threat, and draws a direct historical parallel:
"I view Uber very similar to how I viewed Blackberry in 2000, let's say seven, 2008 timeframe when the iPhone was first released."
BlackBerry kept posting record revenue for a few years after the iPhone launched before the business actually cracked, and Dustin expects a similar lag for Uber — record profits could continue for years even as the long-term threat builds, which is exactly why he no longer wants to hold it as a long-term investment. Hamid agreed Uber isn't in immediate danger, circling back to his cybercab math: even a wildly successful Tesla robotaxi rollout would only pull about $1 billion away from Uber's revenue by 2027 — a rounding error against a company already growing on its own.
7. Why Micron Beats Broadcom
Asked whether to open a position in Broadcom after CEO Hock Tan projected $115 billion and $230 billion of AI revenue in 2027 and 2028, Hamid ran the same earnings-hub comparison he uses on every stock.
"So a forward PE of 22 is actually pretty solid. It's fantastic."
But measured against the alternatives, Broadcom doesn't win: Nvidia trades at a lower forward P/E of 18 while growing faster and holding the top position in AI compute, and Micron is cheaper still.
"Well, in Micron's case, it has a forward PE of seven."
That means, in Hamid's framing, that Micron would pay back an investor's money in profit within about seven years even before counting dividends or buybacks — a better potential return, in his view, than either chip giant, and the reason Micron remains his position over both Nvidia and Broadcom despite its slower headline growth rate.
"What matters to me is how much revenue are they making, how much profits are they making, and what are they projected to make."
Not chart patterns, not technical setups — just the revenue, the profit and whether he believes the company will hit its own projections. That test is what keeps him in Micron over Broadcom, and it's the same lens he applied earlier in the episode to Meta, Sterling Infrastructure and everything else the two of them traded this week.
Across a week that ran from Meta's market cap to Tesla's fleet math to a $2,000 iPhone, the two hosts kept coming back to the same test: not the chart, not the hype cycle, but whether the revenue and profit are actually there and headed where the company says they're going.
Bonus Insights
Hamid admitted he had to be told who Broadcom's CEO was before running the numbers on the stock. Asked about Hock Tan's AI revenue projections, he said, "I haven't been following Broadcom very closely." "And I don't know who Hock Tan is," he added, before Dustin told him: "The company's CEO."
Hamid flagged that his market-adoption S-curve explanation plays better on video than audio: "So if by the way, if you're listening to the podcast, this portion might be worth turning on the video," which he said is available on Apple Podcasts and Spotify.
The two closed the episode on their usual running joke about wrapping up "in the books," with Dustin calling it the "book of life."
Products, Companies & Tools Mentioned
Meta (Muse.ai added an estimated $80-90 billion to its market cap in a day; Hamid added to his position twice and it's now 18-19% of his portfolio)
Apple (Launched the $2,000 foldable iPhone Duo, new iPhone 18 Pro models, AirPods 5 and Apple Watch Ultra, and raised prices on older iPhones by $100)
Tesla (Rolled out Cybercab in Austin; the co-hosts debate whether individuals, partners or Tesla itself will own the robotaxi fleet)
OpenAI (ChatGPT's Astra scores 53 on intelligence versus Opus's 51, at roughly half the cost per task)
Anthropic (Claude, benchmarked in the episode as "Fable 5.1" and "Opus," is Hamid's current daily-driver model)
xAI (Grokbot is Muse's closest rival on the professional side; Grok 4.6 is Hamid's current low-cost model, soon to be replaced by Muse Spark)
IREN (Referred to on the show as Iron; transitioning from Bitcoin mining to an AI neocloud, targeting a $4 billion annual run rate by year-end against $707 million last year)
Robinhood (Dustin sold another 17% of his position at $123, largely to cut crypto exposure)
Sterling Infrastructure (A listener's pick; Hamid found 61% revenue growth and 100% EPS growth but passed, citing projected declining revenue next quarter)
Rivian (Spun its robotics ambitions into Mind Robotics rather than build an in-house Optimus rival, and has partnered with Uber on autonomous-vehicle deployment)
Uber (Dustin exited his position a year ago on the coming robotaxi threat, comparing it to BlackBerry's slow-motion decline after the iPhone launched)
Broadcom (CEO Hock Tan's $115B-$230B AI revenue targets weren't enough to win Hamid over at a forward P/E of 22)
Nvidia (Trades at a lower forward P/E of 18 than Broadcom while growing faster, but still loses out to Micron in Hamid's comparison)
Micron (Hamid's preferred AI-hardware pick over Nvidia and Broadcom, at a forward P/E of 7)
Waymo (Used as the yardstick for Tesla's cybercab ambitions — 60,000 units would be 20 times Waymo's current fleet)
Google (Gemini is the fastest model on Hamid's speed chart, though he says it trails Muse on intelligence)
Earnings Hub (Hamid's own analytics tool, used live on the show to score Sterling Infrastructure, Broadcom and Micron)
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: