Good morning. Nvidia has a $35 billion queue of Chinese chip orders. GM has a fresh $20 billion buyback. Both numbers need a little unpacking before they go into anyone’s earnings model.

This morning’s results also show where the money is going: aircraft maintenance, defense, Magic cards and the equipment inside data centers. Netflix gets its turn tonight. First, the difference between demand, deliveries and dollars actually earned.

Nvidia’s $35B queue has a supply problem.

Nvidia headquarters, archive. Coolcaesar; edit by Jacek Halicki / Wikimedia · CC BY-SA 4.0.

ByteDance, Alibaba and Tencent have ordered more than three million Nvidia RTX PRO 5500 cards, with a combined sticker value above $35 billion, Reuters reports, citing people familiar with the matter.

The catch: Nvidia has reportedly told customers it cannot guarantee filling those orders. Demand exceeds planned production. The report points to early samples in December and commercial deliveries starting in the first quarter of 2027.

These are workstation chips, below Nvidia’s most advanced data-center processors. Chinese buyers still want them because they need computing capacity for AI services. Export restrictions have made the menu smaller; they have not removed the appetite.

For shareholders, this is evidence of demand that could become revenue over several quarters. It is not $35 billion of sales already booked, nor a company-confirmed backlog. Manufacturing allocation, delivery timing and trade rules still determine how much reaches the income statement.

That makes the next useful update fairly specific: units Nvidia can actually ship, and what supplying China means for capacity allocated elsewhere. The order headline is impressive. The delivery schedule will tell us what it is worth.

GM’s $20B buyback is bigger than its guidance change.

2023 Chevrolet Silverado LT, archive. Deathpallie325; crop by SouthernResidentOrca / Wikimedia · CC BY-SA 4.0.

GM reported $50.0 billion in revenue and $4.1 billion in adjusted operating profit, then authorized another $20 billion of share repurchases. It has finished the previous $6 billion program. The new authorization runs over multiple years and has no expiration date.

The important distinction sits in the first row. GM narrowed its profit range around the same $15 billion midpoint. The more meaningful upgrade is automotive free cash flow: the midpoint rises from $10.5 billion to $12 billion.

Extra cash gives the company more flexibility to retire shares while funding the business. GM had 849 million shares outstanding at September’s end, down from 904 million at the end of 2025. A smaller share count can lift earnings per share even when total operating profit changes little.

That is why the buyback deserves attention without turning every guidance line into a growth story. Repurchases create more value when the purchase price is sensible and the underlying cash generation lasts. The authorization itself is permission to buy; GM can slow or suspend it.

For the next quarter, watch the cash conversion behind the program. A larger repurchase budget is useful, but recurring cash generation is what makes it sustainable.

GE and RTX have the orders. Now comes the delivery.

GE90-115B engine, archive. Chris Light / Wikimedia · CC BY-SA 4.0.

GE Aerospace booked $19 billion of orders, up 49%, while RTX ended the quarter with a $340 billion backlog. Airlines and defense customers are keeping both businesses busy. Turning that demand into cash remains the more useful test.

GE’s adjusted revenue was $13.3 billion, up 18%, and adjusted EPS rose 36%. Its service business benefits when engines stay in use and need maintenance. Management now expects low-double-digit commercial-services growth in 2027, supported by a services backlog exceeding $170 billion.

RTX raised its full-year adjusted EPS outlook to $7.65–$7.75 from $7.10–$7.25. But this quarter’s $2.26 adjusted EPS includes a 25-cent recovery of IEEPA tariff costs. That benefit deserves separation from the recurring earnings power of its engines, systems and defense programs.

Cash offers another check: quarterly free cash flow fell 14% at GE and 47% at RTX. Timing can move cash around substantially in these businesses, so one quarter does not settle the argument. It does explain why a record backlog alone is an incomplete investment case.

The next reports need to show delivery progress, service growth and cash collection moving together. GE’s release provides the commercial-engine detail; RTX’s tables separate its businesses and adjustments.

3M raises its outlook. Amazon adds another angle.

3M headquarters, archive. Tony Webster / Wikimedia · CC BY 2.0.

3M posted $6.6 billion in sales and $2.50 in adjusted EPS. Adjusted organic sales grew 3.3%, and management raised its full-year adjusted EPS range to $9.10–$9.15.

There is an accounting wrinkle before calling an earnings beat: 3M now excludes intangible amortization from its adjusted results. The change adds about four cents to this quarter’s adjusted EPS, with prior periods recast. An estimate prepared on the old basis is not a clean comparison.

The separate Amazon agreement adds a longer-term data-center angle. Under the filed terms, 85% of the warrant vests with product payments, reaching full vesting for that portion at $6 billion in cumulative payments. The remaining 15% depends on specified milestones.

That $6 billion is a vesting threshold, not a guaranteed order. The arrangement gives Amazon a potential ownership incentive as it buys products. For 3M, the commercial question is whether those purchases become meaningful, profitable revenue; for shareholders, there is also potential dilution.

Today’s results and the agreement answer different questions. The quarter shows current operating progress. The warrant creates an incentive for a customer relationship to expand. Neither establishes the eventual sales volume or return on the investment.

More AI revenue. Almost no extra profit?

Here is a small piece of our Adobe research. Its latest quarter included $4.65 billion of Creative & Marketing Professionals subscription revenue. Annualize the more precise reported figure and the base is roughly $18.6 billion.

Now test a scenario: $1 billion of new annual AI revenue at a 40% contribution margin, alongside 3% erosion of that subscription base at a 70% contribution margin.

These are C&C assumptions, not Adobe’s forecast. The point is to test whether AI adds spending or replaces existing subscriptions. Our sample also examines valuation, cash flow and the evidence that could undermine the thesis.

Would you pay $20/month for four in-depth C&C company research reports? We are testing interest before launching a paid subscription.

Where big tech finished.

Monday, October 19 close · USD · Change from Friday’s comparable close. Mag 7 plus AMD, Broadcom and Micron: our editorial watchlist.

Amazon led this group, up 4.98%, followed by Tesla and Micron. Eight of the ten rose. These are completed-session prices, not Tuesday premarket quotes.

Hasbro’s best growth is still in the cards.

Hasbro’s Pawtucket offices, archive. Kenneth C. Zirkel / Wikimedia · CC BY 4.0.

Magic: The Gathering generated more than $600 million in quarterly revenue for the first time. Hasbro’s total revenue rose 15% to about $1.6 billion, with adjusted EPS of $2.90.

The split matters. This is a much stronger gaming story than a broad toy-market boom. Wizards’ growth can improve the group’s earnings mix, while licensing lets partners do some of the manufacturing and distribution work.

There is a boost to separate out: results include roughly $55 million of tariff refunds. That cash benefit is welcome, but it is not the same as customers buying more products.

Management raised its annual outlook and doubled its share-repurchase target. Into the holiday period, the useful question is whether the Magic release schedule and consumer-products demand can sustain growth without another refund doing part of the work.

Sea wants to sell its cloud—and potentially list Monee.

Sea’s Shopee offices in Singapore, archive. Chainwit. / Wikimedia · CC BY 4.0.

Sea is preparing a cloud business and considering a public listing of its financial-services arm, Monee. Dow Jones reports the plans from CEO Forrest Li’s staff memo.

Sea Cloud would take technology developed inside the group and offer it to outside businesses. That is a plausible starting advantage: an existing platform and experience operating large consumer services. It still has to win customers against established cloud providers and support them commercially.

A Monee listing could provide dedicated capital and a clearer market valuation for the financial business. Investors would need the ownership structure, offering terms and credit performance before judging what it means for Sea shareholders.

For now, there is no completed IPO or verified offering valuation to plug into a sum-of-the-parts calculation. The development worth following is whether management turns these proposals into disclosed economics: revenue, investment requirements and the stake Sea would retain.

Netflix’s earnings bar is almost exactly its own forecast.

Netflix offices, archive. Coolcaesar / Wikimedia · CC BY-SA 4.0.

Netflix reports after today’s close. The MarketBeat consensus is about $0.82 EPS and $12.87 billion in revenue. Management’s July forecast was $0.82 diluted EPS, $12.86 billion in revenue and a 33.2% operating margin.

That leaves just $10 million between the two revenue numbers. Meeting the company’s forecast would land very close to that estimate; the fourth-quarter outlook may provide more information than a tiny revenue beat.

Last quarter, management expected slower content-amortization growth in the second half and roughly $3 billion in annual advertising revenue. Those assumptions help explain its 31.5% full-year operating-margin target.

Tonight, compare the actual margin with that 33.2% promise, then check whether advertising and content costs still support the annual plan. More viewing or more ad customers only becomes a stronger shareholder story if the added revenue outruns the cost of serving them.

Housing gets the next check.

September housing starts and permits are due this morning. The previous release put August starts at a 1.275 million annual rate and permits at 1.394 million. These are prior readings, not today’s results.

For homebuilders and building-products stocks, watch single-family permits and revisions alongside the headline. A construction rebound matters more if buyers can finance the homes and builders can preserve margins. Follow the Census release.

The market still leans toward a Fed pause.

At roughly 7:45 a.m. ET, Polymarket’s October contracts priced no change at 85.5% and a 25-basis-point increase at 14.5%. These are separate contract prices, not a Fed forecast or an exhaustive probability total.

The useful signal for stocks is how that balance moves after new data. View the contracts and resolution rules.

The Street’s changes.

BlackRock: TD Cowen upgrades to Buy with a $1,405 target. Rating upgrade.

Robinhood: JPMorgan raises its target to $115 from $103. Target change.

Lam Research: Evercore ISI cuts its target to $315 from $355. Target cut.

The latter two are target revisions; neither is being presented as a rating change.