Newsletter · 2026-09-28 · Ashutosh Agarwal

Contrarian Value Ideas and AI Bets Dominate a Busy Week for Stock Pickers - Weekly Podcast Idea Digest - Week of September 21, 2026

A cross-sector roundup of the stock ideas making the rounds on investing podcasts for the week of September 21, 2026, as the 10-year Treasury yield hit 5.11 percent: contrarian value calls on Shift4, Devon Energy and Lennar, dueling models on Meta and Broadcom, Tom Hayes on non-AI names like GXO and Hormel, and a warning that Micron is cheap only at peak earnings.

Weekly Podcast Idea Digest

Week of September 21, 2026: Contrarian Value Ideas and AI Bets Dominate a Busy Week for Stock Pickers

This was a busy week for stock ideas on podcasts. That's worth noticing, because the backdrop was rough: the 10-year Treasury yield hit 5.11% on Wednesday, its highest since July 2007, and traders now price roughly a 70% chance of another Fed rate hike in October (The Rundown, Sept 24).

The ideas that stood out mostly went against the crowd:

• Cheap and unloved: a payments company at 6.5x cash earnings, an oil producer at 4.5x, and a homebuilder that Berkshire kept buying straight through a bad quarter.
• AI, argued on the numbers: Broadcom as a growth stock priced like a value stock, and two careful looks at Meta that end up on opposite sides.
• Everything except AI: Tom Hayes laid out long cases on GXO, Estée Lauder and Hormel, and warned that Micron is cheap only because its earnings are at a peak.

The rest of this issue takes the ideas one at a time: who made the case, what they own, and why.

A quick glossary for terms that come up a lot:

• EBITDA: a company's profit before interest, taxes and some accounting charges. It's a rough stand-in for the cash the business produces. A stock at "6x EBITDA" means the whole company (stock plus debt) is valued at six years of that figure.
• Free cash flow (FCF): the cash left over after the business pays for itself and its investments. "23x free cash flow" means you are paying $23 for every $1 of that cash each year.
• P/E: share price divided by earnings per share.
• Activist investor: a fund that buys a stake and then publicly pushes management to change something, such as selling the company.

1. Shift4 ($FOUR): a payments stock at 6.5x EBITDA that one manager says is "paying for zero" growth

Who: a portfolio manager at Emeth Value Capital, interviewed on Yet Another Value Podcast (Sept 22). Direction: Long.

Payments stocks have been, in the host's words, "a value investor graveyard" for the past year. Even Shift4's own CFO told a Goldman Sachs conference that the company's share price is "a gift."

The Emeth manager's argument is that the market has missed a real growth engine sitting inside Shift4: its restaurant point-of-sale (POS) system, the checkout terminals now sold as Shift4 Dine / SkyTab. Shift4 doesn't report the product separately, so he tracks it through a balance-sheet footnote, "gross equipment under lease," which grows as more terminals go into restaurants:

• About $40 million before the 2021 launch (legacy HarborTouch systems)
• $70M (2021) → $110M (2022) → $180M (2023) → $250M (2024) → $340M (2025)
• About $400 million run-rate by mid-2026, more than 4x in five years

His unit economics: each terminal costs Shift4 about $2,000. A restaurant workstation handles roughly $300,000–$350,000 a year in sales. Shift4 earns about 0.8% on those payments plus $29 a month in software fees. As he put it: "I'll leave it to the listeners to kind of do the math on that."

The bear case, and his answers. Host Andrew Walker pushed hard. He pointed out that PAR Technology and Toast pitched similar stories, and the stocks disappointed. He also asked whether the easy wins are already gone, so that organic growth (growth without acquisitions) slides from about 11% to 2–3% once competitors fight back on price. The manager's replies:

• PAR goes after big chains like McDonald's and Wendy's. Those customers insist on picking their own payment processor, so PAR can't earn much from payments. Shift4's mid-sized restaurants can't negotiate that way.
• "The vast majority, call it 80% plus, of Shift4 Dine POS installs are 100% net new merchants to Shift4." In other words, growth isn't just moving old customers onto new hardware.
• About 60,000 restaurant locations still run legacy software that Shift4 owns. Shift4's plan is to add its payments to those systems, not replace them ("we'll charge you 50 bips" versus 70 at Fiserv), and it does that through roughly 7,000 independent sales-agent relationships.

On valuation, he framed it this way:

"Being at $42 a share solves a lot of issues... I don't know if organic growth is 5% or 10%, could be 12... You're paying for zero or less in my view right now."

The open risk: trust in management. Walker said many investors he spoke to distrust the company after it set medium-term cash-flow targets and later withdrew them, and after what bears call a "self-inflicted mistake" in buying Global Blue.

2. Devon Energy ($DVN): an activist wants it sold at 4.5x EBITDA

Who: Toms Capital Management (over $4 billion in assets), which says it is one of Devon's top five shareholders. It is working with the well-known litigator Alex Spiro. Reported on Squawk on the Street (Sept 23). Direction: Long, with an activist push for a sale.

In a letter sent earlier this month, Toms urged Devon to "review strategic alternatives, including a sale." Its argument:

• Devon's merger with Coterra, completed last May, added the Marcellus, Eagle Ford and Powder River basins to its core Delaware Basin (Permian) acreage. Toms says the resulting mix is so complicated that it costs Devon "at least one multiple point" of valuation versus peers.
• That discount matters a lot when the stock trades at only about 4.5x next year's expected EBITDA.
• Toms first pushed Devon to simplify by selling assets. It now argues a buyer of the whole company could do the selling afterward, "shifting the execution risk of such sales away from Devon holders."
• It isn't alone. Energy investor Kimmeridge has also publicly pushed Devon to streamline, and both firms object that five of the board members are former Coterra directors.

CNBC's reporter was careful about the odds. Exxon, Chevron, Conoco and Diamondback "certainly might have interest" in the Delaware acreage, but negotiating a deal "given the volatility in oil prices, that is not a given." The window for a proxy fight doesn't open until next year. The stock barely moved on the news.

Toms' record includes Kellanova ("a big success"), Kenvue and Denbury.

3. Lennar ($LEN): Berkshire buys through a miss and a guidance cut

Who: Berkshire Hathaway's buying, analyzed by Marcus Graham on Telltales Weekend Update (Sept 27). Note that Telltales says it is produced entirely with AI tools and uses AI-generated voices. Direction: Long (Berkshire's position).

The quarter was bad on every measure:

• EPS of $1.23 vs. $1.29 expected, revenue down 8%
• New orders down 9%, on 20,840 homes delivered
• Full-year delivery guidance cut by about 2,000 homes, to 80,000–81,000

Meanwhile, Berkshire bought $212 million of stock over three September sessions, taking its stake past 10% of the company.

Graham's view: "Nobody buys 10% of a home builder for a delivery guide." At a stake that size, "you are buying the land and the balance sheet." He calls it "a duration call rather than a quarter call," meaning a bet that high mortgage rates are cyclical while the housing shortage is structural.

The numbers he uses: Lennar at about 16x free cash flow on $1.8 billion of trailing FCF, with $10.5 billion of net debt. In the same week it cut guidance, Lennar also declared its regular 50-cent dividend and opened a new division around Des Moines, Iowa. "You don't stand up a new division in a market you're leaving."

What would change his mind: "If orders are still falling high single digits two prints from now with rates unchanged, the structural half of the argument is weaker than it looks."

The same episode compared the week's two big drugmakers:

• Merck ($MRK), at about 23x FCF on $18 billion of trailing FCF, with $44 billion of net debt. It is buying pipeline assets because "the market has already marked down the back half of the decade."
• Eli Lilly ($LLY), at about 50x FCF. The stock's price depends on the Medicare GLP-1 group. CEO Dave Ricks said 700,000 new seniors have started GLP-1s since Medicare coverage began in July, and 70% of them picked Lilly. "Lose it and nothing else on this week's list closes the gap."

4. Broadcom ($AVGO): "Best Buy" three months running at about 21x forward free cash flow

Who: Simon Erickson, founder and CEO of 7investing, on The 7investing Podcast (Sept 25). Direction: Long. This is his firm's top pick for the third month in a row.

Erickson's point is simple: Broadcom's revenue growth is speeding up, even though it's already one of the biggest companies in the market. Here is fiscal 2026 so far:

He points out this is the reverse of what normally happens as a company grows: "It's supposed to go the opposite direction... Not happening for Broadcom."

The main driver is custom AI chips, headlined by the TPUs (Google's in-house AI chips) that Broadcom designs for Google and that are now sold to other customers too. Broadcom had targeted $100 billion of AI revenue by 2030. It now says $120 billion in 2027, "three years ahead of schedule." Erickson's rough math: every gigawatt of data-center capacity means about $60 billion of chip spending this year, up from $40 billion last year. Both SpaceX and Meta say they want five-gigawatt sites.

The valuation argument:

"Broadcom is a company that right now is priced at about 21, 22 times its forward free cash flow. And it's going to grow its revenue at about 93% in this quarter, I think at least 60% plus for the year 2027... that's the market saying... We don't think you're really going to do that."

He thinks the recent weakness in the stock comes from worries that Broadcom won't stay Google's only TPU supplier.

5. Meta ($META): two careful looks, two different conclusions

Meta rose 11% in one day and more than 30% in a month after launching its Muse AI agent. Two podcasts built models to check whether the rally makes sense.

The Investor's Podcast Network, with Daniel Mahncke and Shawn O'Malley (We Study Billionaires TIP848, Sept 24; The Intrinsic Value Podcast TIVP099, Sept 27). Their conviction is split.

Mahncke has built a position. His reasoning is that the market sees Meta's AI spending as all or nothing:

"Either Meta is super intelligence and cloud vision all work out and it's a huge success... or it's a huge mistake... I think the truth is somewhere in the middle. And if that's the case, I think Meta is attractively valued today. Although... it's clearly not even close to as much of a bargain as it has been in 2022."

His model is sobering on cash. In the base case, Meta produces "basically no free cash flow between 2026 and 2028": slightly negative in 2026, "massively negative in 2027," around break-even in 2028. Free cash flow comes back to about $90 billion only in 2030. The scenarios:

• Bear: capital spending (capex) climbs to about $230 billion a year, and margins stay stuck near 30%.
• Base: ads and messaging work, capex levels off around $200 billion, and margins recover to the high 30s by 2030.
• Bull: the "every pixel turns into an ad" idea works. Growth runs near 20%, margins go back to about 40%, and capex peaks at $200 billion in 2027 before falling $10–20 billion a year.

Meta's 2026 capex guidance has been raised several times and now stands at $130–145 billion. Analysts expect close to $200 billion in 2027.

Their blended base-case fair value is about $650–750 per share. That is roughly where the stock trades now. O'Malley's hesitation: "Is this the most obviously compelling moment to buy Meta? Probably not." He sees Meta's spending as "so much more speculative than Alphabet and Amazon."

Chip Stock Investor Podcast (Sept 25). The hosts keep a full position that they have held for nearly 15 years, but they aren't adding. At more than $740 a share after the Sept 21 close, they calculate Meta needs 16% yearly EPS growth for five years to be fairly priced. That's "not as reasonable as it was six months ago and 11 months ago when we took our last two little nibbles."

Their estimate of direct Muse revenue:

• Subscription plans are $7.99 a month (core) and $19.99 (premium), with business tiers priced higher.
• Bear case: fewer than 1% of Meta's 3.6 billion users sign up, which works out to about $3.5 billion a year.
• Base case: about 2% adoption.
• Their call: somewhere between bear and base, "less than $20 billion per year in revenue after a year or two."

The main risk they flag is walled gardens. Amazon has already blocked Muse agents from its site, and they expect other big retailers to do the same.

6. Tom Hayes: buy what isn't AI (GXO, Estée Lauder, Hormel), and be careful with Micron

Who: Tom Hayes, Hedge Fund Tips with Tom Hayes, Episode 362 (Sept 24). Direction: Long GXO, Estée Lauder and Hormel. Cautious on Micron.

Hayes's big-picture view: consumer staples and health care make up a smaller share of the market than at any time in 27 years. That reminds him of energy in 2020 and of Buffett in March 2000, "buying Procter and Gamble, Coca-Cola, Disney, all the stocks that you couldn't give away."

GXO Logistics ($GXO), the world's largest pure-play contract logistics company, and a position he added to:

• The factory cycle is turning. The ISM manufacturing index was below 50 (contraction) for 36 of the 38 months before January 2026, but has now shown expansion for eight straight months, with August at 54.6. The Cass Freight shipments index rose 2.1% year over year in August, its first annual gain since January 2023.
• The business is steadier than the market thinks: multi-year contracts, customer retention above 95%, and 80–90% revenue predictability. It has no clean public comparison, so it "gets lumped in with broader transport."
• There's a lot of room in North America. The US is only 23% of revenue, versus about 49% from the UK. North America is now 41% of the sales pipeline, up from 27% a year ago. First-half North American wins are up 85%. The new CEO, Patrick Kelleher, used to run DHL's North American supply chain business.
• The Amazon sell-off was, in his view, overdone. Amazon sells standard access to its spare capacity, while GXO builds custom operations, and management "has yet to see Amazon in a competitive bid."
• Catalyst: the November 16 investor day, the first long-term roadmap in years. GXO's AI warehouse platform, GXOIQ, is on track for 50 sites by year-end, with 20,000 robots deploying this year.

Estée Lauder ($EL): a $350 stock at its COVID-era peak that fell to $50 and has since doubled to close to $100. The new CEO cleared out excess inventory, cut "billions of costs," and moved to where shoppers are now: Sephora, TikTok Shop, Amazon. Digital sales are up 34%, free cash flow has doubled, and margins are up 320 basis points (3.2 percentage points). "You're probably going to see another 50-point move over the next couple of years."

Hormel Foods ($HRL): "your best GLP-1 play in terms of food." Spam, Planters and sausage are protein, and people on weight-loss drugs want protein. The company has raised its dividend for 65 straight years. "The last time that Hormel was this out of favor... was the last tech peak in 2000."

The warning, on Micron ($MU): people think they're buying it "at nine times earnings. And what they're missing is... they're buying it at nine times peak, never-happen-again earnings." Memory chips run on roughly four-year boom-and-bust cycles. In his view, Chinese and Korean competitors are "going to flood the... market like they do every single cycle." (Micron reports Sept 30.)

7. Natural Gas Services Group ($NGS): an investor who bought "a dollar for 25 cents," then became CEO

Who: Justin Jacobs, formerly of Mill Road Capital, now CEO of NGS, on the Oilfield 360 Podcast (Sept 22). Direction: Long. He owns it and runs it, so this is an insider's case.

Jacobs was not an energy specialist. In Q2 2020, Mill Road put about 40% of its fund to work, and NGS was one of the buys:

"This is a business with about $400 million of units, book value. It's trading for $100 million. It's got long-term contracts. It's production-oriented. It's got no debt... it was just a straight value investment of buying, you know, a dollar for 25 cents."

After board turmoil in 2022, including a plan to spend $150 million on growth in 2023 against about $30 million of EBITDA, funded with the company's first real debt, Mill Road (about 8% owner) nominated him to the board. He joined in April 2023. In February 2024 he moved his family from Miami Beach to Midland, Texas to become CEO.

His industry view: rental gas compression (renting out the equipment that pushes natural gas through pipelines) now shows "much greater capital discipline" in a capital-heavy business, with longer lead times, very high utilization and strong demand. Data-center power needs and LNG exports add to that. He called NGS "an absolute diamond in the rough. I don't think the market understands where this industry is going."

8. Capstone: a microturbine maker bought out of bankruptcy near $1, now about $5

Who: Dean Pernas, who runs a public-markets fund with his brother Deiya, on Buyers and Builders (Sept 21). Direction: Long, held for a little under a year.

Capstone makes microturbines, small gas-fired generators that produce 50 kilowatts to a couple of megawatts. The industry lost money for 40 years (Capstone's tax-loss carryforwards "exceeded a billion dollars at one point"), and most competitors gave up. Capstone went through a pre-packaged bankruptcy and came out just as demand for power off the grid ("behind the meter") took off because of AI and an aging grid.

What changed was a new CEO with heavy engineering experience from GE, who:

• brought key technology and manufacturing in-house instead of relying on distributors and outsourced production
• ran a Six Sigma cost program across hundreds of parts
• lifted gross margins from "sub 10% when we initiated the position to now in excess of 25%"

"We got it around $1 or so. It's now around $5, $5.50 a share... the potential energy is still building." The brothers call their approach "motor investing": buying potential that hasn't yet shown up in the financial statements.

9. Alibaba ($BABA): a long thesis built on Moonshot AI and cheap Chinese models

Who: Tommy Thornton, a long/short hedge fund veteran, on Maggie Lake Talking Markets Ep. 72 (Sept 22). Direction: Long. The position has been a winner.

Thornton's case:

• Alibaba owns 35% of Moonshot AI, valued at $50 billion. It also has stakes in Chinese chip companies, so it's "taking the full stack rather than just taking one lane."
• Alibaba's open-source Qwen model is one of the most downloaded AI models in the world.
• Cheap Chinese models are cutting token prices (what customers pay per unit of AI output), which "have been cut in half recently." He thinks that threatens OpenAI and Anthropic, which "depend on high token costs." He compared it to Microsoft giving away Internet Explorer and killing the $40 Netscape CD-ROM.

He was frank about the price: "Alibaba's not cheap right now," because it's putting its cash flow into AI just like the US giants.

His other positions from the episode:

• Recently bought Alphabet ($GOOGL) around $340. It's now about $355.
• Still owns Exxon ($XOM) and Chevron ($CVX): "$95 oil... Crack spreads are nuts on the upside." (A crack spread is a refiner's margin between crude oil and the fuels it makes.)
• Short the transports ETF ($IYT) because of fuel costs.
• Closed shorts on DoorDash and Airbnb at a profit, and covered the rest of his Tesla position around $300.

10. CRISPR Therapeutics ($CRSP) and Intellia ($NTLA): Cathie Wood's "most inefficiently priced" corner of the market

Who: Cathie Wood, ARK Invest, on the Equity Mates Investing Podcast (Sept 21). Direction: Long.

Asked for "the next Nvidia," Wood pointed to gene editing. Her argument is about analysts: tech analysts avoid health care because it's slow and regulated, and health-care analysts don't value cures properly because a one-time cure doesn't bring repeat revenue the way a daily pill does. "That is a ridiculous answer." She added that insurers "have no problem paying $2.2 million for a cure" when it replaces decades of emergency care.

• CRISPR Therapeutics is already curing beta thalassemia and sickle cell disease. Its pipeline includes a one-time gene edit that cut blood lipid levels by roughly 50% in trials. She called it "the category killer."
• Intellia is earlier-stage and "has been demolished." Wood said Intellia and CRISPR were each worth about $200 billion in 2020 and that Intellia is now under $5 billion. (That $200 billion figure is well above what most published market-cap data shows for 2020, so take it as her claim, not a confirmed number.)

"These stocks are priced as though, really, they're not going to go anywhere... Doubt, doubt, doubt. And then, boom."

11. Deep Sail Capital's microcap book: Supercom, ClearPoint Neuro, NameSilo, Arista, Kraken

Who: Sean Westropp, founder and PM of Deep Sail Capital, on the Planet MicroCap Podcast (Sept 26). Direction: Long on all current positions.

Westropp grades every stock from A-plus to D-minus on four "pillars." Each current position illustrates one:

• Business quality, Arista Networks ($ANET): gross profit margins around 60%, returns on capital around 20%, and "a pretty solid moat" in networking.
• Management, NameSilo: a Canadian domain-registry business run by CEO Paul Andriola, which uses the registry's cash flow to make "venture-style growth" investments.
• Growth, ClearPoint Neuro ($CLPT): more than 60 strategic partnerships as the main supplier of the device used to deliver gene therapies into the brain. "If you believe in cell and gene therapy... they have a long, long, long growth path ahead."
• Valuation, Supercom ($SPCB): he pitched it at Planet MicroCap Las Vegas. It trades "at around five times next year's EBITDA" after "a ton of contracts recently that is just not appreciated by the market."
• Kraken Robotics supplies batteries and sonar/lidar for autonomous underwater vehicles. It has "been a wonderful investment for the fund," helped by the shift toward drone warfare.

He also discussed a finished trade: Precipio ($PRPO), bought at about $6.50 in 2024 and sold at $22–23 after it reached about 30% growth and positive EBITDA.

His rule: valuation is the pillar he bends on most, and management is the one he never bends on.

12. CCL Industries (TSX: CCL): a "very unsexy" Canadian compounder

Who: the hosts of The Canadian Investor (Sept 21). Direction: Constructive, but they call it fairly valued rather than cheap. One host owned it in 2018–19.

CCL makes labels, packaging and polymer banknotes. The track record:

• Five-year average return on invested capital of 11.1%
• Operating margins between 12% and 16% for a decade (15% last year)
• Free cash flow per share growing about 8% a year since 2017, and nearly 10% a year since 2016
• Share count down about 4% since 2021
• Dividend yield of 1.4%, but paying out only 20–35% of free cash flow

The balance sheet is solid: net debt of about $1.5 billion, which is about 1x EBITDA, and interest covered more than 16 times. $500 million of 3.25% debt comes due this October and will probably be refinanced at 6–7% or simply paid off.

Valuation: about 17x forward earnings and 19x forward free cash flow, "around the average it trades for" historically, on a market value of about $16 billion.

Risks they named: acquisition integration (the Innovia deal took a while), resin and energy cost inflation, customer concentration, currency swings, and succession. CEO Geoffrey Martin has led the company for 18 years.

13. The "light bottleneck": Corning, Marvell and MaxLinear

Who: the hosts of The MoneyFlows Show Ep. 43 (Sept 24). One host personally owns MaxLinear. Direction: Long all three.

Their premise: memory was AI's bottleneck in 2026, and in 2027 it will be the optical links that move data between chips using light instead of copper wire.

• Corning ($GLW), "the highway": fiber and connectivity, at about 37x earnings. It has a deal with Nvidia to expand US optical capacity roughly tenfold, and a multi-billion-dollar data-center interconnect agreement with Verizon. Analyst estimates for its optical segment: $8.5B (2026) → $11.1B (2027) → $14.6B (2028).
• Marvell ($MRVL), "the brain": silicon photonics and signal-processing chips. It is 23% below its high at 41.8x earnings, with expected EPS growth of 61% and sales growth of 51% next year. Data-center segment estimates: $9.8B (2027) → $15.7B (2028) → $23.6B (2029). It recently signed an agreement with Google.
• MaxLinear ($MXL): chips that clean up and prepare signals inside optical modules, aimed at 1.6-terabit connections. It is 37% below its high, at about 33x forward earnings.

This episode leans more on fund-flow charts and analyst estimates than on each company's own reported numbers, so treat it as a list of ideas to research.

14. Short takes

• Dave Iben, Kopernik Global (The Meb Faber Show #652, Sept 25). He took Korea to about 19% of his portfolio after the attempted martial-law crisis, buying Hyundai and LG "at a third of book." He thinks Korea Telecom and LG Uplus are "every bit as good as Verizon and AT&T. We're just paying... two thirds less." He owns neither Samsung nor Hynix, which now make up about half of the Korean market. He likes platinum at about $1,800 an ounce, when producers say it would take about $2,500 to justify building a new mine. He also owns Ukrainian farmland companies valued at 80–90% discounts per hectare to US Midwest land.
• Knife River ($KNF) (The Rundown, Sept 24). Starboard Value took a stake and told the construction-materials company to improve or sell itself. The stock rose 5% on the news and is still down about 30% this year.
• Main Street Capital ($MAIN) (Wealthion, Sept 21). Mark Skousen likes this business development company (a fund that lends to and invests in private companies). He cites a 7% yield, the fact that it pays both monthly and quarterly dividends, and defaults of about 1%. "You're lucky if it's up 10, maybe 20% in a year, including dividends. And I'm satisfied with that."
• XPeng ($XPEV) (Bitcoin Takeover, Sept 25). Crypto investor Marc De Mesel has started buying Chinese EV makers. He says they are "worth 10 times less than their U.S. counterparts" while growing "20, 30, 40%" a year. He also holds Tesla call options through a small hedge fund, betting on robotaxis.

What to watch this week

• Micron's earnings (Sept 30). Tom Hayes expects a memory-cycle peak. Telltales notes that Micron's two-year ban on buybacks, a condition of its CHIPS Act funding, ends in December.
• Nike (Thursday). It reports after Bank of America cut its price target from $47 to $30.
• Lennar's new orders, the number Berkshire's bet depends on.
• Devon's response to Toms Capital and Kimmeridge.
• GXO's investor day on November 16, which Hayes calls "a clearing event."