CVS Plans Its Pharmacy Split

If You Only Read One Thing

Breaking up a company can preserve its strongest grip on a market. CVS's separation plan leaves insurance and drug purchasing together while sending roughly 9,000 pharmacies elsewhere. Meanwhile, AI builders are falling as Big Tech rises. Both developments challenge the idea that owning more assets automatically creates more value: what matters is which assets strengthen the business that remains.

CVS Will Separate Pharmacies but Keep Aetna and Caremark Together

CVS is shrinking the scope of its ownership without surrendering the combination that negotiates over healthcare spending. Its planned pharmacy spinoff leaves Aetna, Caremark and specialty drugs together. The company is separating the storefront from much of the machinery that determines how prescriptions get paid for.

The November 4 announcement puts roughly 9,000 pharmacies, mail pharmacy, MinuteClinic and Health100 into CVS Pharmacy. Aetna, Caremark, CVS Specialty, Oak Street Health and Signify remain in CVS Health. Completion is expected in the first half of 2028, subject to board approval, registration and tax conditions. This is a plan, with the businesses still operating together.

That dividing line tells us more than the breakup headline. A pharmacy sells a prescription; a pharmacy benefit manager negotiates drug benefits and pharmacy payments for health plans. Keeping Caremark beside Aetna preserves the connection between insurance coverage and drug purchasing. CVS isn't abandoning integration across healthcare. It's deciding that retail ownership adds less than owning those two functions together.

Separate ownership could also make the pharmacies a more acceptable partner to rival insurers and benefit managers. CVS explicitly says the independent pharmacy company can pursue wider partnerships. Those partners would no longer be directing business into a retailer owned by an insurance competitor. Whether they'll actually send more prescriptions is the commercial question behind the promise of greater focus.

The risk is that contracts become more expensive than internal coordination. Separate companies need their own capital structures and must negotiate services they previously shared. CVS expects both to retain investment-grade ratings, but that expectation leaves unanswered how debt, shared costs and commercial agreements will be divided.

Over the next two years, I'd give the retained insurance-and-drug-services business the stronger bargaining position. It keeps the relationship with the payer and the purchasing function. The pharmacy company gets a chance to win customers from outside CVS, but also has to earn adequate reimbursement from organizations with buying power. A simpler organizational chart alone won't improve that bargain.

The market's initial rejection of the deal, described in Reuters Breakingviews, is a reasonable warning about execution rather than proof that separation can't work. The first Form 10 should show whether CVS Pharmacy receives enough cash-generating capacity to cover its assigned debt and standalone costs. An overloaded pharmacy balance sheet would undermine the case for two stronger businesses.

Slower AI Construction Can Favor Existing Cloud Owners

Investors are drawing a line between companies that already sell computing and companies that must build facilities before collecting rent. Wednesday's election trade sent Amazon higher while data-center developers fell. The AI spending boom can lose speed without every participant losing value.

Closing-price data put Hut 8 down 23.2%, against Amazon's 6.1% gain. The local ballots covered outright bans and advisory questions, as Ballotpedia's measure tracker shows. Voters weren't deciding a national construction ban, and advisory votes don't enact restrictions. A governor's campaign promise also isn't an enacted moratorium.

The financial difference is straightforward. Amazon can keep serving customers from operating infrastructure while reconsidering its next expansion. A developer awaiting its first tenant payment keeps carrying construction and financing costs if delivery slips. Both own physical assets, but only one may already be collecting the cash needed to support them.

Hut 8's quarterly announcement makes that timing concrete. It reports approximately $26.6 billion in expected base-term AI lease value. Initial deliveries are targeted for the second quarter of 2027 at River Bend, the third quarter at Beacon Point's first phase, and the second quarter of 2028 at its second phase. Contract value isn't current rent.

Hut 8 is also a reason to resist indiscriminate pessimism. The company reports construction underway and expects project-level funding without new parent equity or recourse debt. An existing project with approvals and a committed customer may become more valuable if competitors struggle to build. Political exposure needs to be assessed campus by campus; a Texas address alone doesn't prove a delay.

Yesterday's briefing explained which federal powers change after the election. The additional evidence is local opposition and the market's split between owners. This extends the capacity-constraint pattern we've followed: permission can restrict new supply, but the resulting benefit goes first to usable capacity, not every supplier with a large order book.

For the next year or two, established cloud operators look better placed than developers dependent on successive financing rounds. They can retain customers and potentially spend less before earning more. That advantage disappears if restrictions obstruct their own committed expansions or customers' computing demand weakens. Hut 8's next delivery guidance matters more than another day's share price: postponed 2027 handovers would show political or construction risk entering the revenue timetable.

The Contrarian Take

Everyone says: A company needs to own the whole chain to coordinate it effectively. Selling a major business means giving up the integrated strategy.

Here's why that's wrong: Ownership is only one way to keep a commercial relationship. CVS can retain insurance and benefit management while its independent pharmacies seek business from competing payers. Cloud customers can contract for computing without owning the developer that builds the campus. The harder question is what each side can demand when the contract is renewed or something goes wrong. A spinoff can broaden a supplier's customer base while leaving its buyers with most of the negotiating power. Counting subsidiaries won't reveal whether competition improved; reimbursement terms and customer alternatives will.

Under the Radar

Pliant wants to keep the payment relationship when agents take over the task. The Berlin company raised $80 million to expand its payments business and agent connections. Its offer combines regulated infrastructure with permissions, spending controls and audit records. The less obvious opportunity is recurring payment activity after the conversational interface changes hands. Funding and a working connection aren't yet evidence of substantial agent-driven payment volume.
Cloudflare lets publishers distinguish agents by name. Its personal-agent controls let sites allow or block supported agents separately, including blocking access on advertising-supported pages. That makes the commercial choice more specific than blocking all automation: a publisher can welcome one intermediary and reject another. Cloudflare gains a stronger position in administering those choices, although the release supplies no evidence that publishers have recovered lost advertising revenue.

Quick Takes

Iran: the administration's public split hasn't become a peace agreement. JD Vance blamed the war for Republican losses while Trump defended launching it. Axios's reporting describes advisers divided among negotiations, renewed strikes and continued economic pressure. That public rupture changes the politics of the next decision. It doesn't settle Tehran's terms or make a tanker voyage dependable; cheaper oil can reflect a probability change before any shipping condition improves. (Source)
Meta buys Conductor and closes its independent cloud service. Conductor says cloud workspaces become read-only November 13 and shut down November 20; local software will receive models and critical fixes until early 2027. Meta gains a team building the interface for managing coding agents. Existing customers lose an independent hosted option before the replacement arrives, showing how acquisitions can concentrate distribution even when rival models remain available. (Source)
Toyota's better profit forecast comes with a volume qualification. Toyota forecasts ¥3.57 trillion in annual net income. Its first-half consolidated vehicle sales fell 0.5%, while revenue rose 9.5%; the company credits exchange rates for the automotive segment's profit increase. Toyota can earn more without selling more cars. That favors its current earnings outlook, but a stronger yen could remove part of the benefit before any improvement in competitive position is tested. (Source)

The Thread

Companies are changing who carries the risk between a customer promise and its fulfillment. CVS's separation will require outside shareholders to assess pharmacy economics on their own. A cloud developer finances the interval before a tenant starts paying, while an established cloud operator can choose how much additional capacity to commission. That creates a possible bias in the next investment cycle: the businesses that control customers can demand expansion from suppliers while limiting their own exposure. Pliant's payment controls and Cloudflare's agent policies point toward the same incentive in software. An interface can become easier to replace while the company authorizing the transaction keeps its position. The question for the next round of deals is whether suppliers get paid enough to carry those obligations, or merely receive a larger forecast of future business.

Predictions

I predict: Conductor will end cloud write access by November 13 while keeping its local application available, rather than extend the hosted service unchanged. The announced migration separates Meta's acquisition from continuity of the independent product. (Confidence: high; Check by: 2026-11-14)

November 5, 2026 · 3:35 AM ET