The Focus List: Four Verdicts Due While the Market Argues About Funding

The AI trade spent the week arguing about money, not machinery.

Reports this week said Amazon is exploring a plan to move roughly $8 billion of Nvidia chips into a special-purpose vehicle. That is a separate company, funded by outside debt, that would lease the hardware back to Amazon. The chip stays on the rack; the borrowing moves off the balance sheet.

Amazon also raised a £4.25 billion sterling bond and a $17.5 billion term loan, making it the largest bond issuer among the big cloud providers. Lease the silicon, bond the buildings, and let the credit market carry what the cash flow will not.

That is the argument underneath every AI stock right now. It is no longer about whether the demand is real; it is about who funds it and at what price.

When a market cannot price an uncertainty, it rarely debates it. It sells it. The week punished almost anything carrying a question mark, and it fell hardest on companies whose answers are not due until later.

Three of the four names below sit in the bottom fifth of their 52-week ranges, and one sits near the top. All four report inside four weeks: Netflix on October 20, SoFi on October 27, Amazon expected around October 29, and Constellation expected in early November.

The financing has become the story, and these four are about to replace an argument with a number. Constellation owns the scarce fuel. Amazon makes its own chips and rents them out. SoFi sells the deposits that fund its loans. Netflix sells a habit, the only product here that survives a bad quarter.

Last edition put Meta, AMD, Zscaler and Axon on the list, each with a stated proof point. Not one has produced it, because every catalyst was a future quarter.

Meta has not reported. Its proof was a late-October print with free cash flow, the cash left after paying for buildings and equipment, re-accelerating off $784 million. Shares are down roughly 3% since the edition, and the $10 billion legal charge we flagged lands in the third quarter.

AMD has not reported. Its proof was revenue at or above $13 billion, data-centre growth no slower than June, and adjusted gross margin at 56%. Shares are roughly flat, and the MI450 ramp into 2027 is still the swing factor.

Zscaler's proof was a late-November print with net new recurring revenue re-accelerating above 17%; the kill number was another surprise in the sales organisation. Reports have since surfaced of a change at the head of its revenue organisation, ahead of its October 6 investor day. Shares are up roughly 2%.

Axon has not reported. Its proof was an eleventh straight quarter above 30% growth with margin at or above 25.5%. Reported talks naming Axon a possible buyer of Flock Safety keep the consolidation thread alive. Shares are down roughly 4%.

One of the four has produced the risk. None has produced the proof.

1. Constellation Energy (CEG) — The Nuclear Landlord

Think of a landlord whose buildings come with their own power station, in a city where every new tenant runs a factory.

Constellation is the largest operator of nuclear plants in the United States, running 55 gigawatts of generation. In plain English, it signs long contracts to sell that electricity, the capacity and the clean-energy credits.

That makes it the firm, round-the-clock supplier an AI campus cannot do without, because a data centre cannot run on sunshine alone. It signed 920 megawatts of new long-term power purchase agreements, contracts to sell electricity at a set price for a set period, starting in 2029.

The June quarter proved the machine still runs. Adjusted operating earnings were $2.55 a share, ahead of the roughly $2.37 expected, on revenue of $7.504 billion, up about 23%. Management raised full-year guidance to $11.50 to $12.50 a share.

Nuclear output hit 44,160 gigawatt-hours at a 93.0% capacity factor, a measure of how much a plant produced against the most it could have.

Constellation sits near the bottom of its range, roughly 36% below its one-year high and about 30% lower this year, and the argument is about price, not business. It trades near 19.3 times forward earnings, the price against next year's expected profit, against roughly 13.5 times for its closest peer, Vistra.

On EV/EBITDA, which compares the whole business including debt to cash profit, it is about 14.5 against Vistra's 10.5. The catch is the cash: over four quarters Constellation generated roughly $0.3 billion of free cash flow, and the June quarter alone was negative $118 million on $1.25 billion of capital spending.

It carries about $24.7 billion of debt and depends on refinancing at good rates. On the day it announced the flagship Amazon deal at Calvert Cliffs, the stock opened down 5.04%.

What proves the bears right is full-year earnings at or below the $11.50 floor, free cash flow staying negative, or peers re-rating lower and dragging the multiple down.

What we're watching: third-quarter results expected in early November, though the company has not confirmed the date. The proof is full-year adjusted earnings tracking inside $11.50 to $12.50. The kill number is earnings below the floor, or another decline in nuclear output.

Our take: Wait for a better entry. The asset cannot be replicated, but you are paying a premium multiple for rent that starts in 2029 while free cash flow sits near zero. A pullback toward the peer multiple changes that call.

2. Amazon.com (AMZN) — The Everything Store With Its Own Chips

The store that sends you a parcel is also the landlord of a computing empire, and it now designs the processors inside it.

Amazon is the world's largest online retailer and its largest cloud provider. In plain English, it sells you goods and rents computing power to companies through AWS, the arm that hosts other businesses' software and AI.

In the AI buildout Amazon sits on both sides of the trade. It is the largest buyer of computing on the planet, with roughly $220 billion of capital spending guided for 2026. Its AI and custom-chip businesses are each past a $25 billion annual run-rate.

The June quarter showed the engine pulling. Net sales were $200.6 billion, up 20%, and AWS grew 37% to $42.2 billion, its fastest in eighteen quarters.

Operating income rose 43% to $27.5 billion, with the AWS margin near 39.3%. Advertising grew 26%, and management guided the current quarter to $197 billion to $202 billion of sales.

It is the one name here near the top of its range, up roughly 11% this year. At about 24 times forward earnings it trades in line with Microsoft near 21.9 and Alphabet near 22.8, not at the discount people assume.

Here is the catch. Trailing free cash flow is negative, a $7.6 billion outflow against an $18.2 billion inflow a year earlier, because purchases of property and equipment rose $66.1 billion.

And the June profit headline is not what it looks like. Net income of $62.6 billion was dominated by a $53.4 billion non-operating gain on the Anthropic stake, so valuing Amazon on that number is valuing a mark-to-market.

Rivals grew faster, with Google Cloud up 82% and Microsoft Azure up 43%. What proves the bears right is AWS growth dropping below 30%, or 2026 capital spending above $220 billion while free cash flow stays negative.

What we're watching: third-quarter results expected around October 29, not yet confirmed by the company. The proof is AWS growth holding at or above 35% with the segment margin in the high 30s, financed without another stretch of the balance sheet.

Our take: Accumulate. AWS is accelerating while the rest of retail funds it, and negative free cash flow is the price of a runway rather than a warning. An AWS quarter below 30% growth ends that.

3. SoFi Technologies (SOFI) — The Deposit Machine

The cheapest money in banking is the money customers hand you to look after, and SoFi has turned collecting it into a machine.

SoFi is a digital-only bank with a real national charter. In plain English, it takes deposits, makes personal and student loans, issues cards, and offers investing and crypto in one app. It has 15.8 million members.

Its advantage is funding. Deposits reached $45.5 billion, up $5.3 billion in the quarter, and money from depositors is far cheaper than money borrowed in markets. The mix saved roughly $712.6 million of annualised interest expense and widened the net interest margin, the gap between what the bank earns on loans and pays on deposits.

The growth is not in question. June-quarter revenue was a record $1.2187 billion, up about 42.5%, with net income of $156.6 million, or $0.12 a share.

Adjusted EBITDA, a rough proxy for cash profit, rose 44% to $357.8 million at a 30% margin. Members rose 35%, and fee-based revenue, the part that does not depend on lending, was 39% of the total.

The stock is not loved. It sits in the bottom 5% of its 52-week range, roughly 51% below its one-year high and about 43% lower this year. On the full-year guide of $0.60 it trades near 26 times forward earnings, or 19 times next year's estimate.

The catch is that profit is not keeping pace with revenue. On July 29 the company raised its full-year revenue guidance but held earnings per share at $0.60, because its assumed tax rate jumped about 700 basis points. The stock fell roughly 9% on the print.

Read plainly, the extra revenue is absorbed by funding costs, credit provisions and tax before it reaches the shareholder. Financial-services contribution margin slipped to 46% from 52%, and the chief technology officer sold shares on September 21 under a preset plan.

What proves the bears right is a spike in credit provisions, or another year when revenue rises and per-share profit does not.

What we're watching: third-quarter results on October 27, confirmed by the company. The proof is earnings per share tracking toward the $0.60 goal, with the quarter near the roughly $0.17 expected. The kill number is a provision for credit losses well above the roughly 37% rise seen in June.

Our take: Accumulate. The cheapest funding base in consumer fintech with a 42% growth rate is worth about 19 times next year's earnings. A credit miss on October 27 ends it.

4. Netflix (NFLX) — The Habit

The strongest business in media is not the one with the best shows; it is the one whose cancellation you keep postponing.

Netflix is the world's largest subscription streaming service, with more than 325 million paid memberships across a cheaper ad-supported tier and premium ad-free ones. In plain English, it sells monthly memberships and increasingly sells advertising inside them.

The cash machine still works. June-quarter revenue was $12.56 billion, up 13.4%. Operating income was $4.193 billion at a 33.4% margin, and free cash flow was $1.525 billion.

Over four quarters the company has produced about $11 billion of free cash flow. Watching hours grew 2% in the first half.

The stock is priced for decline. It closed the week at its one-year low, in the bottom 3% of its 52-week range, roughly 46% below its high and about 26% lower this year. Forward earnings sit near 17.6 times, down from roughly 46 times a year ago.

The catch is engagement, and it is specific. Growth has decelerated from 16% last year to a guided 11.7% this quarter, and the stock has fallen on the session after each of the last four earnings reports.

Wells Fargo's Steven Cahall cut the shares to Underweight and reduced his target, citing engagement trends and a weak second-half content slate. His base case is a roughly 21% decline in viewing hours for the platform's biggest originals.

The company has also been showing less: no quarterly subscriber numbers, and a viewing report moving to annual from 2027. Its earlier earnings beat leaned on a one-off termination fee.

What proves the bears right is a flat-to-down view-hours number on October 20, or a cut to the full-year margin guidance. The most-watched number into that print is engagement, not revenue.

What we're watching: third-quarter results on October 20, confirmed by the company. The proof is full-year revenue in the $51.0 billion to $51.4 billion band with an operating margin at or above 31.5%. The kill number is a margin below 31.5%, or engagement growth turning negative.

Our take: Wait for a better entry. A business earning about $11 billion of free cash flow is not broken, but you would be catching a falling knife before a print the market is braced to sell.

The bottom line

The week asked who funds the AI buildout and at what price. The honest answer is that nobody knows yet, so the market sold whatever carried a question mark.

Constellation holds reactors the hyperscalers cannot route around. Amazon rents the cloud and prints its own chips. SoFi collects the deposits that fund its loans. Netflix owns a habit that outlasts a weak quarter.

Three sit near the bottom of their ranges, where the questions live; the fourth sits near the top, where the confidence is. All four answer inside the month.

We will track all four, tell you when our call changes, and grade every one of them in public.

Analysis, not advice. We may own these names and we may sell them at any time.