Apple's Growth Costs More

If You Only Read One Thing

Apple's customers are absorbing price increases, and its suppliers are still getting the better deal. Apple forecasts shrinking margins after a record quarter. Oracle's $20 billion financing puts a different cost on growth: interest for lenders and ownership for new shareholders. Both businesses have demand. The harder question is how much of the reward reaches the investors who already own them.

Apple's Higher Prices Don't Cover Higher Costs

Apple can raise prices and grow sales without protecting its margins. That's the finding in John Ternus' first earnings report as CEO: customers are still buying, but component suppliers are taking more of each sale.

The September-quarter release puts revenue at $114.5 billion, up 12%. iPhone revenue rose 17% and Mac revenue 24%. Apple is keeping its customers while forecasting less gross profit per dollar of sales.

Apple forecasts December-quarter gross margin of 45%–46%, against September's reported 47.9%. But September included roughly one percentage point of tariff refunds. Remove that benefit and compare the adjusted 46.9% with December's 45.5% midpoint: the underlying decline is about 1.4 percentage points, rather than the headline 2.4. Apple would retain roughly 1.4 cents less gross profit per dollar of sales on that comparison.

On the earnings call, CFO Kevan Parekh attributed most of the expected decline to memory. Ternus said higher product prices and previously purchased inventory provide partial relief. Inventory bought at older prices helps until Apple consumes it; the next shipment brings the higher cost into the accounts.

Apple's scale still helps it secure components. It can't make scarce components cheap by negotiating harder. That's how the memory boom reaches consumer hardware: suppliers can earn more even when the device brand has already charged customers more. October 28's supplier-contract story showed producers securing future buyers. Apple's results reveal the other side of that bargaining relationship, inside a consumer company's forecast.

Apple has another cushion: its services business. Yet Siri AI adds cloud costs as usage expands. Owning the customer relationship gives Apple ways to monetize demand over time; delivering the included service costs money immediately.

The bullish case is credible. Apple still expects December revenue growth of 7%–10%, and strong product demand could support further price increases. We don't know how much customers will tolerate before they postpone upgrades.

For the next several quarters, I'd favor memory suppliers' ability to retain price increases over a rapid recovery in Apple's hardware margins. That follows the broader pattern of scarce inputs collecting the first returns from expanding demand. Apple's next results would weaken that view if product margins recover while management still reports rising memory purchase costs.

Oracle has found money for its AI expansion. Existing shareholders will share future profits with new owners, while the bonds commit Oracle to recurring interest payments. Whether the expansion earns enough to justify those claims is a separate question.

Oracle's November 2 pricing announcement specifies about 86.2 million shares at $116, raising $10 billion before expenses. It also prices $10 billion of bonds, with coupons ranging from 7% to 8.65%. Both are expected to close November 4, independently of each other.

Those five bond tranches imply roughly $777 million in annual coupon payments: each tranche's principal multiplied by its coupon. That's the gross obligation while all $10 billion is outstanding. If Oracle uses proceeds to repay old debt, some coupons replace existing interest. New shareholders take a different claim: part of the upside previously belonging to existing holders.

The share price also matters. Oracle's preliminary prospectus lists October 30's closing price at $130.67. Selling at $116 means accepting an approximately 11% discount to that close. A lower price requires more shares to raise the same amount.

This advances October 29's comparison of cloud growth and cash generation. Oracle has moved from describing its funding needs to setting the price investors receive for meeting them. A lender can receive interest while shareholders are still waiting for completed data centers to earn back their cost.

Oracle's defense deserves weight. Funding early reduces the danger of needing money during a worse market. If demand and delivery follow the plan, today's expensive capital could still finance valuable assets. The interest bill alone can't tell us whether Oracle has overpaid for growth.

But the offering proceeds are available for general corporate purposes. They aren't a separate account reserved for profitable AI projects, and raising them doesn't tell us what the completed capacity will earn.

Over the next year or two, equipment and construction suppliers have a clearer route to payment than Oracle shareholders have to excess returns. Oracle's creditors receive contractual coupons, although they still bear credit risk. Shareholders need enough operating profit to cover those coupons and justify the additional shares. The capacity boom can reward suppliers while leaving the builder's owners disappointed.

Oracle's next quarterly filing can test that judgment: operating cash flow catching up with capital spending would weaken the case that expansion keeps outrunning its ability to fund itself.

The Contrarian Take

Everyone says: Securing the money for an AI buildout resolves the main risk to its shareholders.

Here's why that's wrong: Oracle has priced both shares and bonds, but securing the money creates new obligations. Existing owners have to give up part of the reward to obtain it. A company can remain liquid and keep building while its shareholders receive a smaller fraction of future earnings. The roughly $777 million gross annual coupon bill makes the lenders' claim concrete, although refinancing old debt could offset some of the added interest. The financing can succeed even when the eventual investment disappoints. A business can grow faster than its owners' returns.

Under the Radar

China is making share-buyback support permanent. Alongside another stock-support swap operation, regulators are converting buyback relending into a long-term arrangement and cutting its rate from 1.25% to 1%, Reuters reports. That changes how companies can finance purchases of their own shares. It can support share prices without increasing household spending or fixing weak business demand. The more persistent intervention is the cheaper recurring funding, rather than a single day's rally. (Reuters)
Nvidia is pressing offshore clouds to reduce Chinese customer exposure. The FT reports a requirement that at least 51% of services go to non-Chinese firms; The Information describes nonmandatory guidance for future contracts (both first reports are paywalled). The scope and enforceability remain disputed. Either version moves scrutiny beyond where chips are installed toward who rents them, potentially reducing the value of offshore capacity built around Chinese demand. (Reporting)

Quick Takes

The Supreme Court rejected emergency intervention in California's ballot count. The unsigned order denied DOJ's request to segregate ballots associated with roughly 20,000 alleged noncitizens; no dissents were noted. This materially advances the dispute covered yesterday. California can continue processing ballots under the existing rules. Denial of emergency relief settles neither every individual's eligibility nor the underlying appeal, and the list isn't a count of proven unlawful votes. (Source)
US strikes complicate the Hormuz diplomacy. US officials told Reuters that forces struck Qeshm Island, describing the action as limited and defensive rather than a resumption of major combat. That is fresh military activity alongside the conditional reopening proposal covered yesterday. Shipping companies need a dependable operating environment; negotiations can continue while aircraft and vessels remain exposed. A diplomatic headline alone can't remove that operating risk. (Source)
Marvell is buying the supplier that moves data between chips. Its $115 billion all-stock Lumentum deal would combine chip and networking design with lasers and optical components. Existing Marvell owners would hold about 67% of the combined company. Bringing those technologies under one owner could improve joint design; the purchase price also gives Lumentum owners a claim on the gains. Regulatory and shareholder approvals still stand between agreement and completion. (Source)
South Africa and Rwanda face withdrawal of AGOA preferences. Washington notified Congress of its intent to end their eligibility on January 1, 2027, citing failures against the program's criteria. The notice creates a dated threat to qualifying duty-free trade; it doesn't impose a fresh charge on every shipment immediately. Exporters with US-dependent production face a narrower window to renegotiate contracts or find customers elsewhere. (Source)

The Thread

Scarcity can make a supplier more profitable without making its shares a better investment at any price. Marvell's purchase of Lumentum adds that complication to the case for owning the companies that supply this expansion. When a buyer pays for expected future profits upfront, the seller's shareholders receive part of the reward before the combined business delivers it. An all-stock deal makes the buyer's existing owners share those future profits. Oracle is also selling ownership to fund growth, while Apple is trying to keep more of each sale from its suppliers. The next question is what each company paid for its position. A scarce input can command a premium, yet buying the company that owns it can transfer much of that premium to the people selling. Correctly identifying the bottleneck is only the first half of an investment judgment.

Predictions

I predict: Oracle will announce completion of both priced offerings by November 6, raising approximately $20 billion before expenses. The transactions have specified terms and an expected November 4 closing; this prediction tests execution, separately from the longer cash-flow question. (Confidence: high; Check by: 2026-11-06)

November 3, 2026 · 3:21 AM ET