Kerry Grinkmeyer - Best of Us Investors

10/27/2026

Seeing a lot of DeepSeek posts on X that feel incomplete or just off.

One part is usually fair: their list API prices really are much lower than OpenAI and Anthropic for similar work. Cache-miss prices around $0.28 in / $0.42 out per million tokens. That part holds up.

Where it goes sideways is when people mash three different things into one viral chart:
- assumed GPU lease cost (like that $2/H800-hour in the Day 6 write-up)
- theoretical profits if every token sold at old R1 list prices
- what customers actually pay today

That famous "huge margin" math was theoretical. DeepSeek said actual revenue was much lower. And old R1 charts from early 2025 keep getting treated like today's reality.

So: "DeepSeek is dirt cheap on list price" — mostly real.
"They crushed OpenAI on cost and profit because of that one report" — usually oversold.

This is education, not investment advice.

Learn more at the Best of US Investors website: www.bestofusinvestors.com

10/27/2026

Just finished a new paper called "Why the Winners Won." It covers retail and consumer stocks.

The big question: did these stocks go up because the companies actually made more money, or because people were just willing to pay more?

Some of each. Five Below, Ross, Dollar General and Dollar Tree grew earnings 24-51%, and their valuations stayed flat or dropped. They earned it.

Others? Abercrombie's P/E doubled from 6.9x to 13.8x. Cheesecake Factory went from 15x to 27x. Macy's and Delta went up even though their earnings went down.

A few more things that jumped out:
- One-time tariff refunds boosted recent margins at Ross, Dollar Tree and Target by 4-6.5 points. That won't repeat.
- Marriott, Hilton, Starbucks, Walmart and Costco trade at 40-59x earnings, and they score lowest.
- Carnival and Royal Caribbean grew earnings 16-18%... and their stocks fell about 8%.

Top Scorecard names were Abercrombie, Chili's parent Brinker, Five Below, Airbnb and Ross. 13 of 22 passed.

The full report is for Best of US Investors subscribers. Powered by 1880Q. This is education, not investment advice.

Learn more at the Best of US Investors website: www.bestofusinvestors.com

10/27/2026

New podcast just dropped: Anthropic's S-1 IPO.

We went through the filing with one big question — what's real operating performance, and what's mostly marks on convertibles?

Quick take: the revenue is real (about $16.3B in H1 2026 vs about $4.6B for all of 2025). The giant GAAP loss number is mostly fair-value accounting. On an adjusted basis they showed about +$1.6B operating income in that half.

Also in the mix: ~$131.7B cash and short-term investments, 11,900+ customers at $100k+, NRR around 900%, proposed ticker ANT.

Full research is mine, powered by 1880Q, for Best of US Investors subscribers. This episode is the audio overview.

Here it is: https://youtu.be/ianz3Autzdk

This is education, not investment advice.

Learn more at the Best of US Investors website: www.bestofusinvestors.com

**14 Day Free Trial! Test Drive the BUS!! ** https://buy.stripe.com...

10/26/2026

Anthropic just filed its S-1. Proposed Nasdaq ticker: ANT. Share count and price aren't set yet — still early.

Why I'm paying attention: if you watch AI, capex, or enterprise software, this is one of the big labs stepping into public markets. The filing is a clearer look at how this stuff gets financed and who the customers are.

Highlights from the S-1 and company notice:
- Goldman Sachs and Morgan Stanley named as underwriters
- Public benefit corporation, with Long-Term Benefit Trust / Class T directors and Founder LLC Class F at 50.1% voting power
- About $16.3B revenue in H1 2026 vs about $4.6B for all of 2025
- About $9B on AI infra in that same half
- Cash and short-term investments around $131.7B; $141.1B raised in total
- GAAP loss is mostly fair-value marks on convertibles — adjusted operating income was about +$1.6B in H1 (~10% margin)
- More than 11,900 customers at $100k+ TTM; net revenue retention around 900%
- Top five customers ~25% of revenue (nine months ended Sep 30, 2026); no single customer over 10%
- Dario Amodei as CEO, Daniela Amodei as President

Not a buy or sell call. Just a useful window into the AI economy.

This is education, not investment advice.

Learn more at the Best of US Investors website: www.bestofusinvestors.com

10/26/2026

Last week oil went up. Stocks were mixed — S&P and Nasdaq down, Dow a little better.

The bigger story to me: how does all this AI buildout actually get paid for? You're seeing that financing stress around names like CoreWeave. Jensen Huang said neocloud demand is off the charts, but there isn't enough capital out there to fund all the capacity people want right now.

This week several of the Mag7 report. Microsoft and Google Wednesday. Amazon Thursday. I'm watching the guidance and the capex plans more than anything.

Bond yields matter too. The 10-year fell last week. Bounce higher this week, or keep grinding lower?

Fed decision Wednesday. Markets expect them to hold. Inflation still feels high for regular households though. If they did raise rates, yields would probably follow. Base case is no change. Either way — be patient.

This is education, not investment advice.

Learn more at the Best of US Investors website: www.bestofusinvestors.com

10/24/2026

Kind of a strange week in the market. 9 of the 11 sectors went up... and the S&P 500 still went down 0.44%.

Why? The big guys. Communication Services dropped 4.31% and Tech dipped 0.19%, and they carry a lot of weight in the index.

Energy (+1.58%) and Health Care (+1.52%) led the week, and they're also #1 and #2 on my 62-day ranking. Industrials had a rough one. It's last on the board and slowing down fast.

Couple things I'm watching. Consumer Discretionary jumped from #8 to #5 and is still picking up speed. Health Care is still strong, but it's starting to slow.

And the S&P 500's momentum just flipped to "accelerating"... by a hair (+3.38% vs +3.34%). That could be a short bounce, not a real trend. Trends have to keep showing up.

This is from my Morning Edge Sector Scorecard, powered by 1880Q for Best of US Investors subscribers. This is education, not investment advice.

Learn more at the Best of US Investors website: www.bestofusinvestors.com

10/22/2026

New episode is live now! This one's all about who's winning at the checkout right now.

Watch it here: https://youtu.be/7XjSqzl3jNw

Retail sales in September were up 6.3% from a year ago. Sounds great, right? But the control group, the core number, only rose 0.1%. People are saving just 4.1%, sentiment is at 46.3, and mortgage rates are 7.40%. Folks are being careful.

You can see it in the stocks. Over the past year, Staples are up 13.2% and Discretionary is down 1.4%. Walmart, Target, Amazon, Ross and Five Below are winning. Home Depot, Lowe's, Nike, Lululemon and Wingstop are lagging. And the NRF expects holiday sales up 4.2%.

I wrote the report behind this episode, powered by 1880Q for Best of US Investors subscribers. Give it a listen!

This is education, not investment advice.

Learn more at the Best of US Investors website: www.bestofusinvestors.com

10/21/2026

You've probably heard someone say AI debt is "the next subprime." A new report from Americans for Financial Reform, called "Brace for Impact," says this debt is being repackaged "in a manner eerily reminiscent of the subprime bubble." So I looked into it.

Here's how it works. Banks make big loans to build data centers, then sell them off. Meta's Hyperion project sold $27.3 billion in A+ bonds, and PIMCO bought about $18 billion of it. Banks took more than six months to sell a $38 billion loan for data centers leased to Oracle. They're even paying funds and insurers to take the first hit if loans go bad.

Who ends up holding it? Insurers, with about $512 billion in privately rated bonds. And regular folks, through private credit funds. Blue Owl had to limit withdrawals at two of its funds.

Is it 2008 all over again? Not quite. Most of the tenants are big, highly rated companies, the debt is backed by leases and guarantees, and there's none of that bundling-the-bundles stuff. But it's concentrated, a lot of it is hard to see, and the chips wear out fast.

What I'm watching: Oracle's cost to insure its debt hit a record 2.88%, CoreWeave bonds yield about 15%, and OpenAI makes up about half of Oracle's $638 billion backlog.

This is education, not investment advice.

Learn more at the Best of US Investors website: www.bestofusinvestors.com

10/19/2026

New episode is up! This one's on CoreWeave, and it's a fun fight: the credit market is screaming while analysts are cheering. So who's right?

Watch it here: https://youtu.be/xIvkb4a6rms

Here's the setup. CoreWeave's bonds traded at roughly 15–16% yields last week, and the cost to insure its debt against default hit record levels. Meanwhile, 38 analysts have an average target more than double the stock price as of Oct 19.

The bulls have a ~$104B backlog and contracts where 98% of revenue is take-or-pay. The bears have $51.6B of debt plus leases and interest eating about 25% of revenue.

I wrote the 10th Man Report on this, powered by 1880Q research, and I argue against both sides. Best of US Investors subscribers get the full thing. Give the episode a listen and tell me which side you're on.

This is education, not investment advice.

Learn more at the Best of US Investors website: www.bestofusinvestors.com

10/19/2026

Everybody talks about AI. Not many people talk about how it's getting paid for. More and more, the answer is debt.

Take Oracle. Last quarter it spent $28.5 billion building data centers. It brought in $23.1 billion in operating cash flow. It has about $125 billion in borrowings, plus $288 billion in data-center lease commitments that aren't on its balance sheet yet.

The bond market is paying attention. Oracle's 6.55% bonds due 2046, issued back in February, traded around 83 cents on the dollar on Oct 15. That's about an 8.4% yield. Some of that is rising Treasury yields, and some of it is worry about credit. The cost to insure Oracle's debt against default hit 288 basis points that day, a record in data going back to 2008.

It's not just Oracle. A $38 billion construction loan for data centers Oracle leases in Texas and Wisconsin went to the developer, Vantage Data Centers. Months later, banks were still trying to sell the last pieces, and some tried selling at a discount. The FT reported JPMorgan and Morgan Stanley were looking at ways to shed some of their data-center and AI loan risk. And when Goldman Sachs and JPMorgan arranged $5 billion for GPU cloud startup Volta Infrastructure Holdings, they guaranteed only $1.3 billion. The other $3.7 billion, backed by chips, went to investors with no guarantee at about an 11% yield.

My take? The build-out keeps rolling, but lenders are starting to charge for the risk. Keep an eye on it.

This is education, not personalized investment advice.

Learn more at the Best of US Investors website: www.bestofusinvestors.com