Here’s a preview of what we’ll cover this week:

Macro: Warsh: Learn From Experience; Inflation: The One Blemish

Markets: Strong Earnings, Scared Investors; What Fearful Investors Are Missing; Google’s Record Quarter, Record Selloff; Microsoft Struck the Best Deal in AI; Anthropic’s S-1: No Shoe Dropped

Lumida Curations: David Orr on Finding the Best Ideas; Dario Amodei on Why Most People Missed AI

I sat down with Michael Parekh, former Goldman Sachs lead tech analyst and one of our returning guests, to break down Anthropic’s S-1 and the arguments against the AI trade.

Here’s what we covered:

Watch the full conversation here. For more conversations like this one, subscribe to our YouTube channel.

The market was waiting for the shoe to drop on Anthropic’s S-1. It didn’t.

Overall, it’s fairly solid.

Start with growth.

Revenue went from $386 million in 2024 to $16.3 billion in 1H26.

Q2 alone brought in $11.6 billion, up about 15x from a year earlier.

That growth is coming from existing customers spending more.

Customers are spending 9x what they spent a year ago. About 250 customers now pay more than $10 million a year.

What We Like

The economics are getting better as Anthropic scales.

Every dollar Anthropic spends on compute now generates $2.08 of revenue, up from $0.57 a year ago.

And that happened while Anthropic cut customer cost per task by about 40% with Opus 5.5.

That’s the most tangible proof in the filing that AI can make money.

Revenue per employee is $6.6 million, up from $1.4 million a year earlier. That’s comparable to Nvidia.

The headline losses look scary, but most of them are accounting.

The $80.7 billion net loss includes $72.4 billion of fair-value markups on convertible notes. The actual operating loss was $693 million.

For every $100 of revenue, Anthropic spends $14 on cost of revenue, $60 on R&D, $23 on sales and marketing and $6 on G&A. This leaves $4 as an operating loss.

However, that $60 on R&D is Anthropic’s lever. If it cut R&D to 50% of revenue, it would turn an operating profit.

The balance sheet is strong.

Anthropic holds $132 billion of cash, more than most public companies. Its capex obligations are also far smaller than OpenAI’s.

And the funding is lined up.

Broadcom committed up to $40 billion in convertible notes. Google has $20 billion of remaining capacity, and Amazon $15 billion. A Morgan Stanley syndicate provided a $15 billion revolver.

However, this financing from its suppliers raises the argument of circular financing.

But most of it is clean financing in our view.

It is suppliers lending to their customer, the way a merchant offers an installment loan.

Lenders are underwriting against one to two years of visible orders, and they have visibility on what Anthropic is making.

So Anthropic has access to financing both from its own operations and from well-capitalized third parties.

And it’s still early.

Only about 10,000 customers pay Anthropic $100,000 or more a year, and about 1,900 pay $1 million or more.

That’s nothing. There’s much more room to grow.

What We Don’t Like

First, it’s hard to separate what Anthropic spends on training from what it spends on inference.

That’s the split that tells you whether the unit economics hold as the business matures.

Second, customer concentration is high.

About 48% of revenue flows through Amazon, Google and Microsoft, which all compete with Anthropic.

The coding companies built on Claude, like Cursor and Replit, can roll out their own models and are working on it.

Anthropic needs to keep the best technical edge so its customers don’t become its competitors.

Third, switching costs are low.

Developers use routers like OpenRouter that send each task to the most cost-efficient model. Good-enough models like DeepSeek 4.1 cost about 90% less, and they are taking away market share.

The next test is Anthropic’s October ARR update. The bar is roughly $78 billion, and it needs to point toward the year-end estimate of about $100 billion.

A miss would bring the profitability and funding concerns right back.

So how should you read this?

Anthropic will likely price before Thanksgiving, with expectations in the $1.5 to $2 trillion range. That’s roughly 10x next year’s expected revenue.

That’s not unreasonable for a company growing this fast with improving unit economics.

The S-1 is also bullish for semis. Every dollar of Anthropic’s growth runs through compute, and the filing proves the demand is real.

A new fight is breaking out across tech: when an AI agent does the shopping, the searching and the banking, who owns the customer?

Three examples from the past month.

(1) Amazon shuts out Muse.

Amazon blocked Meta’s Muse agent from shopping its site directly.

Amazon makes its money on attention. Sellers pay Amazon to show up at the top of the page, and shoppers scroll past sponsored listings on the way to checkout.

An agent doesn’t scroll. It reads the specs, compares the prices and buys.

If Muse does the shopping, the ad slot is worth a lot less, and Muse decides which product wins.

Amazon becomes a warehouse with a checkout page.

Watch what Amazon is willing to do instead. This month it joined YouTube Shopping, letting creators tag Amazon products in their videos. YouTube sends the traffic, and Amazon keeps the checkout and the customer.

That’s the line Amazon is drawing. It’s happy to take traffic from anyone, as long as it doesn’t hand over the customer.

Muse has its own prize in mind. If it owns the purchase decision, it can take a cut of every transaction, the way TikTok built TikTok Shop.

Amazon still has the leverage for now. Muse needs Amazon’s inventory and delivery network, and that’s nearly impossible to rebuild.

(2) Reddit loses traffic to Google’s AI.

People increasingly get their answers straight from Google’s AI and never click through to Reddit. Google keeps the relationship, and Reddit becomes a content supplier. More on this in the Google section below.

(3) Banks lock out AI agents.

Banks are moving to deny AI agents access to customer accounts.

They see the same threat. If an agent moves your money, pays your bills and picks your products, the bank is just plumbing.

The pattern is the same in all three cases.

Whoever sits between the customer and the decision captures the value. Everyone else becomes a supplier.

If an AI can own your customer relationship, you might be in trouble.

The incumbents are building walls. The agents are building doors.

Alphabet shares fell more than 12% on Thursday, the worst day in the company’s history.

The drop erased roughly half a trillion dollars of market value. Only Nvidia’s nearly $600 billion slide in early 2025 was bigger.

The reason was higher capex. (Google missed on revenue and earnings estimates as well, but we don’t think that led the drop.)

Alphabet spent $51.4 billion on capex in the quarter, and free cash flow went negative at $9.1 billion.

Anat Ashkenazi (CFO, Alphabet) said, “We continue to expect our CapEx to increase significantly in 2027”.

She also warned that “our sizable investments … will put pressure on free cash flows.”

Add a slowdown in Search, and the market didn’t need more reasons to hit the sell button.

But, Google’s negative free cash flow isn’t a sign of Nokia-fication of Google, as many Twitter gurus seem to suggest.

Start with the cash flow itself.

About $5.2 billion of the cash burn was a one-time payment for a 2018 EU fine. Another chunk went into inventory for TPU systems that Alphabet now sells to other companies.

That inventory isn’t speculative. “We only do that after we have signed contracts,” said Thomas Kurian (CEO, Google Cloud).

Ashkenazi put it simply: “it’s not the traditional kind of CapEx… but rather a different cycle of cash flow generation.”

Alphabet pays for the TPU systems upfront and gets paid when it delivers them.

That makes Google both a capex spender and a capex receiver.

Cloud Is the Prize

Underneath, the business is accelerating.

(1) Revenue grew 25%, putting Alphabet at a half-trillion-dollar annual run rate.

Operating income grew 30%.

(2) Google Cloud, the crown jewel, has revenue doubling.

And future demand is already contracted.

Cloud backlog stands at $561 billion. Management expects to recognize just over 55% of it within 24 months.

Capex today is buying future earnings.

Google’s Search Story

AI is changing Google search. It’s not dying, it’s changing.

People are searching more than ever.

Google’s year-over-year query growth “is the highest it’s been in the last 5 years,” said Sundar Pichai (CEO, Alphabet).

But they’re searching differently. More of them get their answer straight from AI Overviews and AI Mode, without clicking through to another website.

That shift has two costs.

The first is for Google. AI answers cost more compute to serve than a page of blue links, and Google isn’t fully monetizing them yet.

That will pinch margins for a while. But Google still owns the customer, and it’s already rolling out ads in AI Mode. It will figure out how to monetize that.

The second cost lands on everyone else.

Look at Reddit. It reported the same week, and revenue grew 53%. The stock still fell 18% after hours.

The reason was traffic. Reddit warned that search referrals dropped by more than a million a week in October. If that continues, daily users could fall by a low double-digit percentage this quarter.

When Google answers the question itself, nobody needs to visit Reddit.

And Reddit won’t be the last. Any business that depends on search traffic for customers is exposed.

We had the SaaS apocalypse in the first half of this year. We may be heading into a Search apocalypse next.

So how should you read this?

Google is still a long-term winner. Its returns on capital are positive.

Ashkenazi expects “an attractive ROIC that remains meaningfully above our cost of capital” throughout this cycle.

Today’s earnings growth is the payoff from last year’s spending, and next year’s growth will come from this year’s.

But it’s a time-horizon trade.

For now, we are underweight Google, and would rather own companies Google is paying.

The value in AI is moving from the LLMs to the orchestration layer.

The LLMs are becoming commoditized.

Look at DeepSeek 4.1. It’s 90% cheaper, it’s good enough, and it can run on US-hosted systems.

At that price, you can afford to spend tokens on evals. And for coding, which is the bulk of the market, you don’t need the latest frontier model.

OpenRouter has made all these LLMs interchangeable too.

So where does the value go?

To the harness.

The harness lets an engineer orchestrate multiple agents, much like a conductor leading a symphony.

Build agents write the code, eval agents test it, and an orchestrator agent runs the whole thing.

My own LLM stack has evolved to look like this:

Four models, four jobs. None of them owns the engineer.

Y Combinator recently funded Agent Relay. It is a startup that lets your agents talk to each other.

They also funded Omnara, another agent orchestration tool. We use Omnara at Lumida.

On a side note: YC’s edge in idea generation and distribution is unmatched. Its model gives it a pulse on what’s being built and what works in go-to-market.

If you’re Anthropic or OpenAI, you’re in a difficult position.

You can try to own the engineer’s desktop by building an AI-native GitHub.

Anthropic is already doing this.

But you won’t get traction unless you allow routing to your competitors. And once you do, you’re one option among many.

The recent FOMC meeting was Warsh’s first meeting when he didn’t say something to upset the markets.

Markets ended green on the 28th after declining after all 3 of his previous FOMC conferences.

The Fed held rates at 3.75% to 4% on Wednesday, as markets expected.

Warsh’s read on the economy was about as upbeat as a Fed chair gets.

First-half GDP was revised up to 2.4% from 1.8%. He expects Q3 to come in stronger, with consumer spending running above 3% and business investment at about a double-digit pace.

Based on everything we see, the economy is running hot - consumer, AI capex, and government are all contributing…

On jobs, he called conditions “quite positive.”

“Job gains have kept pace with the workforce, and the unemployment rate has changed little.”

Warsh also pushed back on the idea that the economy is just an AI story. The capex story, he said, is “not just robust, not just strong, but it is broadening.”

He added the FOMC is also seeing acceleration in economic activity in sectors other than tech.

The reason? Productivity.

Warsh says “The economy has been marked up for the first three quarters…Part of it is an improvement in productivity…productivity has been moving higher in the economy for a couple of years.”

We have highlighted the underlying American productivity boom since 2023.

It’s still improving, and AI helps.

AI is making each worker more productive. When each worker produces more, margins expand.

That’s partly why corporate profit margins are at record levels. (The other explanatory factor is government deficits which mathematically imply greater corporate profits as we explained here.)

Higher margins fund more investment. More investment makes workers more productive. And productive workers are worth keeping, so companies hire.

Hiring keeps the paychecks coming. Paychecks keep the consumer spending.

And spending feeds the next round of earnings.

That’s the cycle, and AI just added speed to it.

At the same time, consumers do feel stretched. Both statements can be true.

I was grocery shopping on a Thursday. The parking lot was packed. I asked the grocer why. “Payday.”

So long as Americans get access to the labor market and UE is low, the U.S. economy and the bull market will continue.

In September, headline PCE rose 0.5% m/m to 3.6% YoY.

The inflation is also spreading. Nearly half of PCE components are now rising faster than 3%, against a pre-pandemic average of about 32%.

We see two drivers, and neither is permanent.

The first is gas.

Energy prices have moved higher on geopolitics, and consumers feel it every time they fill up. But gas takes up about 5% to 6% of the average American’s budget.

The economy is less sensitive to oil than it has ever been.

The second is AI capex.

The buildout is pulling hard on chips, memory, power and construction. That’s real demand pressure.

But a lot of it is leaking abroad. That 19.7% jump in imports is mostly AI hardware arriving from Korea, Japan and Taiwan.

And capex is front-loaded by nature. You pay for the data center before it earns a dollar.

What matters is what happens once it’s built. That’s when AI turns deflationary.

Agents are going to do the work it used to take full-time employees to do. Companies will raise their output without adding more workers.

That’s productivity. More output per worker, without raising prices. That’s deflationary - but it will take years for that effect to outweigh inflationary capex.

Earnings season is in full swing, and the numbers are strong.

Of the 635 companies that have reported, 74% beat EPS estimates and 67% beat sales estimates.

Guidance is the standout. 13% of companies raised guidance, while only 5% lowered it.

All of these numbers sit higher than prior averages.

(You can track the latest earnings in the Lumida Feed on the Lumida Invest App. We provide curated insights from each earnings that matter for investors. Download Now)

Once you’re in the app, you can select the ‘Earnings’ section from the top menu to see the latest earnings.

Positioning

Much of the reason we’re seeing stocks decline despite strong fundamentals is everyone was ‘all-in’ on AI this summer.

That started to unwind after risks of datacenter cancellations heightened due to midterm risk in the summer.

Then levered players owning 3X semis and other names are forced to sell.

That unwinds momentum. Then latecomers start to sell.

Then the category catches a downdraft.

We are now in the over-correction part of this process.

Back in the 1H of this year, I was shouting at the over-valuation in industrials. It’s hard to say that about where we are now.

A lot of enthusiasm has leaked out.

Midterms play a role in this too.

We noted in 2023 and 2025 that there was a ‘Trump Bump’ boosting Financials, Defense Stocks, Industrials, Digital Assets, etc.

As betting markets price in a Dem victory, we are seeing healthcare sectors rally and Trump-linked themes fade.

Many studies show that these over-corrections are priced in and it’s best to own those themes that have sold off going into midterms.

The elevated risk indicators from the BofA flow show report have come back down from extreme levels.

Goldman’s risk appetite indicator has fallen to its lowest level since April 2025, right after the tariff correction.

Now, look at what happened after April 2025. Sentiment bottomed, and the market ripped.

Lower sentiment means investors are fearful of markets, and are staying away.

Post-midterm performance is uniformly strong, and it should help sentiment.

The VIX also went up 20% this past Friday and tagged the 20 level at the same time.

That’s bullish.

We ran a study on Friday after the VIX rose 20% in two days, while the S&P 500 is sitting within 3.5% of its all-time high.

The results were constructive for a week and longer timeframes.

That setup has appeared five times since 2013.

And, the results are consistent.

The S&P was higher two weeks later every single time, with an average gain of 1.9%.

One month later, it was again higher every time, up 4.0% on average.

So in today’s case, that one-month window runs straight through November, which strengthens the case.

November has been one of the best months of the year for the S&P 500.

Since WWII, it has averaged a 1.6% gain and been positive just over two-thirds of the time.

Over the last 50 and 25 years, it’s the only month averaging more than 2%, and the most consistent to the upside.

Since the 7.5% drop in 2008, November has been down only three times, each by less than 1%.

This November also comes with extra help.

The midterms clear on Tuesday, and post-midterm performance has been uniformly strong.

Buyback windows are now reopening, and corporate America is coming back with record authorizations.

Yes, there is less free cash flow to fund those buybacks, but still the flows incrementally are positive.

While sentiment soured, the market underneath got healthier.

In early October, we flagged breadth at its weakest since the Dot-Com Bubble. Tech was the only sector holding up the index.

Back then, we said the divergence can’t stay forever, and the average stock should catch up.

It’s catching up.

Breadth weakened for weeks in September while the S&P kept making new highs.

That divergence ended at the close of Q3.

Since then, breadth has improved even as the S&P traded sideways.

The cumulative A/D line has risen for the S&P 500 and all eleven sectors since the end of September.

That’s the opposite of the late 1990s back when breadth weakened for years before the top.

While more stocks are participating, prices haven’t run. That combination has made the market cheaper.

The S&P 500 now trades at about 18x forward earnings, near its April 2025 tariff low.

Now is a great time to shift fixed income allocations towards equities. Stocks are cheap.

This is the good kind of de-rating.

Prices haven’t collapsed. Earnings kept rising while higher Treasury yields capped the multiple.

The 10-year is now rolling over from its peak, which has historically supported a positive re-rating for the S&P 500.

Where is the opportunity?

The biggest discount lies in the AI leaders. They’ve given back all of their gains for the year, even though their earnings haven’t slowed.

The market is pricing them like the cycle is ending. The earnings say it isn’t.

Most of that discount came from data center fears heading into the midterms. That risk is now priced in.

That’s why we prefer the capex receivers over the Mag 7.

(We want true capex receivers, not names like Intel which are actually capex payers.)

Microsoft went up on earnings, and Google went down. OpenAI explains why.

OpenAI accounted for about 11% of Microsoft’s revenue this quarter. That’s roughly $10 billion, or $40 billion annualized.

Anthropic is on pace for a $100 billion run rate by year-end. Microsoft is collecting the equivalent of 40% of an Anthropic from a $12 billion investment.

And Microsoft keeps a revenue share from OpenAI through 2032, plus rights to its IP along the way.

What a great deal for Microsoft.

So OpenAI raises the capital, takes the dilution and develops the models. Microsoft gets the revenue and a free call option on the technology.

Our view: OpenAI took in capital from Amazon, violating its agreement with Microsoft. Microsoft used the contract violation and negotiated great concessions from OpenAI.

You have to give Satya Nadella credit.

Microsoft’s own AI products have been underwhelming, but its dealmaking has been world class.

That revenue share is also a weight on OpenAI.

Private investors value OpenAI on total revenue. Public investors will look at net revenue and earnings once its S-1 lands.

OpenAI’s S-1 will look considerably weaker than Anthropic’s S-1.

Elon Musk says “agentic demand is growing faster than I’ve ever seen anything grow.”

He’s right.

If you use these tools every day, you can see how much new value they can deliver for consumers and enterprises.

This tells me low-skilled white-collar jobs are at risk. (Yes, overall, this is productivity enhancing and a net win for the economy.)

But, we are under-estimating the pace of transformation.

An anecdote: I spoke with a partner at Latham & Watkins. His firm uses Harvey. He told me, “Look, I don’t need associates anymore. I work directly with AI. It’s the client, me and AI.”

The same thing is happening across domains:

CPAs >> are using Basis

Financial analysts >> are using Rogo or Model M

Wealth >> we’re building agents for Lumida

And the prize for these agentic startups is far bigger than software.

SaaS is roughly a $300 billion revenue market. Services is measured in the trillions.

For the first time, you can deploy AI service workers that actually do the job, faster and cheaper.

So, who gets hurt? The bottom of the corporate pyramid.

Professional services firms are built like pyramids.

A few partners sit at the top. Below them is a broad base of juniors doing the research, the first drafts and the document chasing. Expect a hollowing out of the lower ranks.

The top of the pyramid is going to do extraordinarily well, and you should expect more inequality as the top 1% triple their output.

If you’re the best at what you do, you’re going to love AI.

You’ll serve more clients, respond faster, and make even more money.

We are experiencing the shift first-hand at Lumida.

We have recently launched Iris, the world’s first agentic advisor.

Iris joins you live on a Zoom call, and you can ask her questions the way you would with a human advisor.

She is better than me at doing Discovery. She will draft an IPS on the fly. Share simulations. The future is coming fast!

We are open to demos. You can join the waitlist here.

I recently tested it at the Money20/20 event, and the reception was strong.

See some of the anonymized reviews below.

(Disclosure: Don’t rely on Iris for tax or investing advice, and using this agentic service does not establish a service relationship between Lumida and you. We are showing you a glimpse of the future.)

You can now view Lumida Curations in the Lumida Invest App. Curated insights delivered on time. Download the app here.

Dario Amodei argues that most people dismissed early GPT because each of its outputs looked unremarkable on its own, while the few who saw its breadth and generality understood it would reshape economies and nations.

View Curation

David Orr argues that once investors have a thousand or so hours of experience, they should make a call on an idea within an hour or two and move on if they can’t. Time spent agonizing over hard calls is time not spent finding the easy ones, where the real edge often sits.

View Curation

Not Subscribed Yet? Don’t miss out on future insights—subscribe to the newsletter now!

For real-time updates, follow us on:

X | Telegram | YouTube | TikTok | News | Ram’s X | Lumida Health | Lumida Tax | Lumida Memes

As Featured In

Disclaimer: Lumida Wealth Management LLC (“Lumida”) is located in New York, NY, and is an SEC registered investment adviser. SEC registration does not constitute an endorsement of the firm by the Commission nor does it indicate that the adviser has attained a particular level of skill or ability. Lumida only transacts business in those states in which it is registered, or qualifies for an exemption or exclusion from registration requirements. Any direct communication by Lumida with a prospective client will be conducted by a representative that is either registered or qualifies for an exemption or exclusion from registration in the state where the prospective client resides.

The information in this material has been obtained from sources believed to be reliable. While all reasonable care has been taken to ensure that the facts stated in this material are accurate and that the forecasts, opinions and expectations contained herein are fair and reasonable, Lumida, Inc. and Lumida Wealth Management LLC (collectively Lumida) make no representations or warranties whatsoever the completeness or accuracy of the material provided, except with respect to any disclosures relative to Lumida. Accordingly, no reliance should be placed on the accuracy, fairness or completeness of the information contained in this material. Any data discrepancies in this material could be the result of different calculations and/or adjustments. Lumida accepts no liability whatsoever for any loss arising from any use of this material or its contents, and neither Lumida nor any of its respective directors, officers or employees, shall be in any way responsible for the contents hereof, apart from the liabilities and responsibilities that may be imposed on them by the relevant regulatory authority in the jurisdiction in question, or the regulatory regime thereunder. Opinions, forecasts or projections contained in this material represent Lumida’s current opinions or judgment as of the day of the material only and are therefore subject to change without notice. Periodic updates may be provided on companies/industries based on company-specific developments or announcements, market conditions or any other publicly available information. There can be no assurance that future results or events will be consistent with any such opinions, forecasts or projections, which represent only one possible outcome. Furthermore, such opinions, forecasts or projections are subject to certain risks, uncertainties and assumptions that have not been verified, and future actual results or events could differ materially. The value of, or income from, any investments referred to in this material may fluctuate and/or be affected by changes in exchange rates. All pricing is indicative as of the close of market for the securities discussed, unless otherwise stated. Past performance is not indicative of future results. Accordingly, investors may receive back less than originally invested. This material is not intended as an offer or solicitation for the purchase or sale of any financial instrument. The opinions and recommendations herein do not take into account individual client circumstances, objectives, or needs and are not intended as recommendations of particular securities, financial instruments or strategies to particular clients. The recipients of this material must make their own independent decisions regarding any securities or financial instruments mentioned herein and should seek advice from such independent financial, legal, tax or other adviser as they deem necessary. Lumida may trade as a principal on the basis of its views and research, and it may also engage in transactions for its own account or for its clients’ accounts in a manner inconsistent with the views taken in this material, and Lumida is under no obligation to ensure that such other communication is brought to the attention of any recipient of this material. Others within Lumida may take views that are inconsistent with those taken in this material. Employees of Lumida not involved in the preparation of this material may have investments in the financial instruments or securities (or derivatives of such financial instruments or securities) mentioned in this material and may trade them in ways different from those discussed in this material. This material is not an advertisement for or marketing of any issuer, its products or services, or its securities in any jurisdiction.