Newsletter · · Ashutosh Agarwal

Alphabet's AI Spending, the Agent Wars and a Nintendo Short Lead the Week's Stock Ideas - Weekly Podcast Idea Digest - Week of September 21, 2026

Weekly Podcast Idea Digest for the week of September 21, 2026 (episodes aired September 14 to 21). A cross-sector synthesis of single-name stock ideas from investor podcasts, led by a revisit of the Alphabet (GOOGL) thesis, the AI agent wars around Meta (META) and Booking (BKNG), a contrarian Nintendo (NTDOY) short, a Brazilian digital bank at 8x earnings (INTR), and Tom Lee and Dan Ives on Palantir (PLTR) and Apple (AAPL).

Weekly Podcast Idea Digest

Week of September 21, 2026: Alphabet's AI Spending, the Agent Wars and a Nintendo Short Lead the Week's Stock Ideas

Single-name stock ideas pitched on podcasts, September 14–21, 2026.

This was a busy week for people willing to name a ticker and defend it. Below are the concrete stock ideas, longs and shorts, that fund managers, analysts and investors actually laid out on podcasts over the past seven days. Each one includes who said it, the real numbers they used, and why they think it matters. I've grouped the meatiest debates first, then a run of quick-hit pitches at the end.

Big idea #1: Alphabet (GOOGL), the $4 trillion company that somehow still trades like a bargain

Two shows in the same podcast family, We Study Billionaires (September 17) and its sister show The Intrinsic Value Podcast (September 20), used the week to revisit an Alphabet pitch that has already roughly doubled, and to ask an honest question: is the company we own still the company we bought?

Here's the setup, from host Shawn O'Malley, who put Alphabet into his "Intrinsic Value" model portfolio about 18 months ago (it's now his second-largest holding, around 14% of the portfolio):

  • He bought it in early 2024 at about 17 times earnings, roughly $150 a share, when, in his words, "the market was pricing Google as if search was the only business segment that mattered and search was in trouble." The fear was that people would ask ChatGPT their questions instead of Googling them.
  • The stock has since roughly doubled. And the punchline: it still trades at about 17 times earnings today. The multiple didn't change; the earnings did.

What actually happened, in his telling, is that the scary story never showed up. As he put it: the narrative was that "AI would kill search, but it's just really the opposite is what has manifested." Google's "Search and Other" revenue has kept compounding at roughly 14% a year, and AI summaries turned out to be something Google can put ads against, not something that replaces ads.

The bull case is a reminder of how many world-class businesses sit inside one ticker:

  • YouTube, bought for $1.6 billion many years ago, now generates around $36 billion in annual revenue.
  • Google Cloud, growing fast with a large backlog.
  • Waymo, plus a venture arm (Google Ventures) running over $10 billion across roughly 400 companies.
  • A recent legal win: Judge Amit Mehta declined to force Alphabet to sell Chrome, and the roughly $20 billion a year Google pays Apple to be the default search on iPhones stays in place (now renegotiated annually). O'Malley noted the stock rose more than 50% in the year around that overhang clearing.

But here's why this made the deep-dive list rather than a quick mention: the hosts spent real time on what's changed for the worse, and it's the most interesting part.

  • Alphabet now plans to spend around $200 billion in a single year on data centers, has stopped buying back its own stock entirely, and has raised over $100 billion in fresh capital to fund the buildout.
  • Free cash flow (operating cash minus that capex) is getting crushed, and for some of the big cloud players is projected to swing negative. O'Malley pointed to a widely shared Bank of America chart showing the hyperscalers' cash flow turning negative while the chipmakers who sell them the gear see cash flow "skyrocket."
  • His real worry isn't competition anymore. It's that AI is simply more expensive to run than old-fashioned search, so "the company could actually become structurally less profitable as they integrate AI more and more into the business." Add customer concentration in cloud (Anthropic reportedly a large share of the cloud backlog) and you get a business that's better than ever and a balance sheet that's "a lot stranger."

The takeaway they landed on: the thesis got stronger, but the risk moved. It used to be "will search survive?" Now it's "will all this spending earn a return?"

Big idea #2: The agent wars, who gets richer and who gets disintermediated

Two investors spent the week on the same big theme, AI "agents" that go off and do tasks for you, and reached it from opposite ends. One is buying the winner. The other is mapping out the victims.

The buyer: Sean Emory is long Meta (META), and Meta Muse is his evidence

On his show Avory - Markets and Investing (September 20), Sean Emory, founder and chief investment officer of Avory & Co (he disclosed he owns Meta), argued that Meta's new personal-assistant app, Meta Muse, hitting #1 in the App Store and passing ChatGPT, is a genuinely big deal, and that the virality is building, not fading: "We've gone from top 10 to top three to top one."

His framework is simple and worth stealing. A useful agent needs three things:

  1. To know you (context)
  2. A way to reach you (a communication channel)
  3. The ability to act on your behalf

And his argument is that Meta is unusually good at all three:

  • Knowing you: in the consumer world, your context isn't emails and files, it's "how you communicate with your friends… WhatsApp, group chats, Facebook, Instagram." Meta owns that social graph, so Muse "already knows a little bit about you" the moment you open it.
  • Reaching you: Muse is its own app, but it also lives inside WhatsApp, and eventually voice, where Meta's Ray-Ban and Oakley glasses come in.
  • Acting for you: the technical breakthrough, per Emory, is reliable "computer use": the agent spins up a secure browser and actually navigates websites (like logging into your auto-insurance account) rather than needing a company to hand over a special connection. That's the hard, expensive problem, and "I'm actually pretty shocked that they were the first ones to figure it out in a way that is snappy, fun, playful and reliable."

He's tracking 30/60/90-day retention as the real test of whether this sticks.

The mapper: Joseph Carlson is grading his whole portfolio on "agent risk"

On The Joseph Carlson Show (September 18), the host walked through name after name asking a blunt question: if agents like Muse do the shopping and comparing for people, who loses their grip on the customer? His verdicts:

  • Most exposed, Booking Holdings (BKNG): it's essentially a gatekeeper that assembles a trip, and "this is exactly the type of thing that agents are good at." His key insight: the bear case "does not require Booking's volume to collapse. It requires the marginal traveler to become owned by a different interface." Even if Booking still does the work, being reduced to a back-end API changes its economics.
  • Also vulnerable: Zillow, the insurance middlemen (SelectQuote, GoHealth, EverQuote), OpenTable, and Intuit's TurboTax. He shared a striking anecdote: someone posted their insurance policy to an agent and, in 20 minutes, it found the same coverage for $600 a year less. Once shopping has no friction, people shop constantly. He said he sold Intuit over exactly this: TurboTax is, at heart, "a big form to fill out," and agents fill out forms.
  • His biggest holding, Google (GOOGL), a $204,000 stake, is defensible. He noted Google's Search and Other revenue was $224 billion in 2025, up from $198 billion in 2024, and argued Google's distribution (Android, Chrome, Gmail, Maps, YouTube) and intent data let it work with agents rather than get bypassed by them.
  • DoorDash (DASH), he's keeping it. Comparison shopping can reach outside its app, but "the delivery part of this is not a database lookup." It needs driver density, dispatching and routing that an agent can't replicate. The rule he offered: businesses that are just an interface are exposed; businesses that physically execute the transaction are protected.

Put the two shows together and you get the whole trade: buy the companies that own your context and execute reliably; be very careful with the ones that only sit in the middle.

Big idea #3: The week's best short, betting against Nintendo (NTDOY)

On Yet Another Value Podcast (September 18), a guest writing under "Accrued Interest" made the case for being bearish on Nintendo, a genuine contrarian short, since most people love the company.

His credibility pitch first: "I pride myself on my shorts. I went six for six in my shorts… for the first half of the year," with the average short down more than 30%. So he's "not afraid to go against consensus."

The core argument is that Nintendo never escaped its biggest problem: it's still a hit-driven, cyclical hardware company.

  • The "flywheel is not there." The whole bull thesis was that Nintendo would break free of the boom-bust console cycle. He doesn't see it. To believe you can own this for more than a year, he says, "they need to sell more Switch 2 software" to drag the audience onto the new console, and Switch 2's software attach rate (games bought per console) launched meaningfully below the original Switch's.
  • A weak holiday lineup. Nintendo is heading into "its second Christmas without any more flagship titles." The big September showcase (the "9-9" Nintendo Direct) leaned on a Legend of Zelda remake; the next real system-seller, Pokémon, isn't expected until spring 2027 and doesn't even have a date. Tellingly, "the stock started tanking during the presentation."
  • A pricing red flag straight out of the memory shortage. For the first time, Nintendo raised a console's price early in its life. His read: with memory chips in short supply (the same DRAM/NAND squeeze lifting Micron), Nintendo is eating roughly $150 of memory cost per unit and "implicitly told you that they don't see memory costs coming down anytime soon." So don't expect a cheaper "Switch 2 Lite" to reignite demand.

He was fair about the bull rebuttal, that Nintendo's library is a Disney-like "oil well" that can re-release a hit and top the charts, but countered that Disney kept buying new franchises (Marvel, Star Wars) to refill the well, while Nintendo is "just doing Mickey Mouse Part 10." His bottom line: it's a stock you rent into big releases, not one you can put away and forget.

(Same guest also touched on Spotify, arguing its advertising tier has quietly underperformed: the ad share of revenue has been falling even as bulls hoped it would grow.)

Big idea #4: Off the beaten path, a Brazilian bank at 8x earnings (Inter, ticker INTR)

On Chit Chat Stocks (September 16), investor Ian Bezek made the case for Inter & Co, a Brazilian digital bank most American investors have never heard of, the smaller, less-famous cousin of Nubank.

The numbers are the whole pitch, and they're eye-catching:

  • Market cap around $2.4 billion (Bezek: "this is definitely not JP Morgan").
  • Trading at about 8 times current earnings, and, on 2028 analyst estimates, around 5 times.
  • Earnings growing roughly 30% a year, which they've done more or less since the IPO.
  • Return on equity up from 1% three years ago to 16% today, with management targeting 30%.
  • Bad loans (non-performing loans) around 6% and stable, normal for a lender charging high rates in Brazil, where net interest margins run 15–20%.

The "why is it cheap?" answer is basically geography and interest rates. Brazil's benchmark rate is 14%, so a local investor can earn ~12% risk-free in the same banking app, a huge drag on the whole Brazilian stock market. Bezek lumps Inter in with other cheap Brazilian fintechs (PagSeguro at ~5x, StoneCo at ~6x) that the market throws in one "too hard" bucket, but argues Inter is quietly executing while weaker peers see bad loans explode.

Two catalysts and one quirk:

  • Catalyst 1: just keep hitting the ROE target. His bank-investor rule of thumb (take ROE and divide by 10 to get a fair price-to-book multiple) implies Inter is worth ~1.6x book now, heading toward 3x if it delivers the promised 30% ROE.
  • Catalyst 2: a U.S. launch in Florida this year, aimed squarely at the large Brazilian-American community, including a clever "credit portability" feature that lets people use their Brazilian credit history to buy in the States.
  • The quirk: Inter pays out ~25% of profits as dividends and has no buyback. Bezek thinks U.S. investors would prefer a buyback, "but I don't see one coming," a cultural preference for dividends in Latin America.

Big idea #5: The bull megaphone, Tom Lee and Dan Ives on Palantir, Apple, and "S&P 16,000"

The Compound and Friends (September 18) put two of the market's loudest bulls, Fundstrat's Tom Lee and Wedbush's Dan Ives, in one room. It's the most unapologetically optimistic hour of the week, and they named names.

Palantir (PLTR): Ives's headline call: "this is a stock that could appreciate 4 or 5x from here over the next 3 to 4 years given the cash flow… it's a trillion dollar name." His argument is that investors "massively underappreciate how advanced their technology is" and underestimate the free cash flow as Palantir moves from government into transforming enterprise software.

Apple (AAPL): the contrarian twist: the thing critics mocked (no giant data-center capex, no in-house frontier model) now "seems… to look more like an advantage than a deficit." Ives: "20% of the world is going to access AI through an Apple device," and Apple monetizes that install base without carrying the capex bill.

The macro backbone, mostly from Ives:

  • Chip demand-to-supply is running 13-to-1 ("13 orders for every one chip that can be produced"), and he doesn't see true balance until roughly 2029.
  • For every $1 of AI capex, he claims a "$5 to $6 multiplier across the rest of tech." He thinks the Street is underestimating earnings by 25–30% over the next few years.
  • His biggest fear isn't a bubble, it's politics: data-center projects getting voted down locally, because "every data center that gets voted down, China wins."

Tom Lee supplied the sober counterweight, and it's the most quotable bit for anyone who isn't a raging bull. Asked what actually ends a bull market:

"Two things will kill a bull market. One is a bona fide bubble… The other is monetary policy intervention. The Fed 80% of the time is the reason a bull market ends."

Lee's framing for why he's still bullish: this year earnings are up ~25% while the market is up ~10%, so "the market got cheaper." He floated S&P 16,000 as a target and argued "80% of what will get us there is already in your portfolio," meaning the existing giants, not yet-to-exist startups. All of this came during a week when semiconductors were in a ~19% drawdown, which both men framed as a buying opportunity rather than a warning.

Quick hits: more single-name pitches from the week

  • Tesla (TSLA), cautious bull, We Talk Money (Sept 17). The hosts, long skeptics, admitted the self-driving story is finally showing up: in Austin they're seeing a "step function change" in robotaxis. The catch is the price: a $1.4 trillion market cap with earnings that don't yet justify it, the most expensive of the Mag 7. Their bull math: only about 1.2% of U.S. cars are Teslas, and just 1 in 500 runs the latest "Hardware 4" self-driving. If adoption goes "parabolic" to 25–35% the way Nvidia or Shopify grew into their valuations, it works. A Grok sentiment scan of ~1,500 owner replies came back ~80% strongly positive ("life changer," "never going back"). Biggest risk flagged: cheap Chinese self-driving from the likes of BYD.

  • Credo Technology (CRDO), long, Chip Stock Investor (Sept 15). A fabless designer of networking chips and active electrical cables for AI data centers, with triple-digit revenue growth. The host ran a simple reverse-DCF: at roughly $171 a share, the market is baking in about 58% annual growth for three years, and the company is guiding to 70–80% revenue growth next fiscal year, so the price looks "within a somewhat reasonable range" after a big selloff. It's a full holding for them. The obvious risk: a serious slowdown in AI data-center buildout would hit both growth and the fat margins.

  • Kodan (ASX: KDN), long, Equity Mates with Ophir Asset Management's Andrew Mitchell (Sept 17). An Australian small-cap (just promoted into the ASX 100) with two engines. First, gold detectors, where it owns the market: its top model, the GPZ, had its price doubled from A$9,000 to A$18,000, and dealers told Mitchell the best-selling "Gold Monster 2" is so good "they should double the price again." Second, and the exciting one, drone communication systems selling into Ukraine: battlefield-proven, "low probability of intercept," scoring like Motorola and taking share as NATO, Taiwan and Europe pour money into drones. It trades near 40x earnings, which Mitchell concedes is rich, but he sees "significant earnings upside." Risks: the high multiple, a falling gold price, and, ironically, peace in Ukraine. His fund returned 22.5% after fees.

  • The memory trade: Micron, SanDisk, Silicon Motion, MoneyFlows with Jason Bodner and Lucas Downey (Sept 17). Three picks on the DRAM/HBM/NAND shortage feeding AI: Micron (MU) at ~5.9x forward earnings with a $1,575 target (~70% upside); SanDisk (SNDK) at ~6.9x with a $2,263 target (~46%); and Silicon Motion (SIMO) at ~16.6x with a $369 target (~56%), as the maker of the flash controllers that make all that storage work.

  • Deep-value screens: ADM, Delek, Super Micro, Value Investor with Zacks' Tracey Ryniec (Sept 18). Three cheap, top-ranked names: Archer-Daniels-Midland (ADM) at a 16.6 P/E and 0.5x sales, with ~52% expected earnings growth; Delek US (DK) at a 5.3 forward P/E, but flagged as a shorter-term refining-cycle play; and Super Micro (SMCI) at ~8x earnings and 0.6x sales with ~22% growth.

  • Two "weird" names from the Big Short crew, The Real Eisman Playbook with Porter Collins and Vincent Daniel (Sept 14). Alongside a lot of gold, they pitched Golar LNG (GLNG), which owns floating vessels that convert natural gas to LNG at sea, a "20-year toll road" contract with only ~$2.5 billion of capex versus $30–50 billion for an onshore plant, tied to their bullish view on Argentina's gas. And Glasshouse Brands (GLASF), a California cannabis grower that converted a giant vegetable greenhouse and is the low-cost producer at ~$90–100 a pound; the upside is the federal reschedule from Schedule 1 to Schedule 3 eventually letting it sell across state lines or overseas at far higher prices, a 50–65% EBITDA-margin business if it happens, which they peg as a likely 2027 event. They also revisited their long-running (and painful) Carvana (CVNA) short, admitting they got squeezed and that the "mystery buyer" of Carvana's loan paper, long rumored to trace to Mark Walter's Delaware Life, still can't be proven despite years of trying.

  • REITs getting a second look, Dividend Stockpile with a Hoya Capital analyst (Sept 17). Agree Realty (ADC), a net-lease landlord with a monthly dividend, ~7% earnings growth and raised guidance; and American Tower (AMT) as a contrarian, currently-out-of-favor pick with durable tower demand.

  • The "boring compounders," Money Tree Investing (Sept 18). A guest made the case for high-moat, predictable cash machines: Cintas (CTAS) for its steady dividend-and-share-price growth, Home Depot (HD) for "treasury-like" returns on capital, and Mastercard/Visa (MA/V) as long-term core holdings.

  • A copper name, unnamed but pointed, Monetary Matters / Other People's Money with Jérémie Boyer (Sept 20). A commodities fund manager talked up an unnamed copper producer trading at ~17x earnings that's entering its cash-generating phase, explicitly framing it as better value than the popular large-caps, which he'd steer clear of: Freeport-McMoRan (FCX), which he considers low-quality, and Southern Copper (SCCO) at ~40x earnings.