Anthropic’s Profit Needs Capital

If You Only Read One Thing

A profitable quarter can make a company need more capital. Anthropic’s IPO filing shows why: Claude’s growth can pay for current operations while encouraging much bigger commitments to future capacity. Ovintiv’s purchase of Logan Energy tackles the same risk from the other end, buying oil assets it can operate more cheaply. Both companies have to make tomorrow’s capacity earn its cost.

Anthropic Has Earned a Harder Question

Anthropic’s business has crossed a more useful threshold than the headline loss suggests: Anthropic generated a quarterly operating profit. Anthropic now has to show that expanding the business can keep paying as much larger compute commitments arrive.

Anthropic’s October 26 S-1 reports $11.555 billion of June-quarter revenue and $974 million of operating income. Anthropic still lost $693 million from operations over the first half. But the company’s $80.748 billion first-half net loss includes $72.388 billion from revaluing convertible notes, derivatives and similar financing instruments. Anthropic’s rising valuation helped enlarge that accounting charge.

So is Anthropic already self-financing? It isn’t. Anthropic had $112.4 billion in cash and $19.3 billion in short-term investments at June 30, alongside approximately $265.6 billion of future fixed or minimum compute commitments spanning one to 20 years. Anthropic has to secure machines before knowing exactly how much customers will spend on them.

A profitable service can still consume cash when its owner builds ahead of demand.

Anthropic does have an unusually strong answer on customer expansion. Its September net revenue retention was approximately 900%: qualifying existing customers generated roughly nine times their year-earlier quarterly revenue. Anthropic excludes fixed-price Pro and Max subscriptions from that calculation. Enterprise customers are expanding inside the product, giving Anthropic a better starting point than a capacity builder still searching for buyers.

Anthropic’s best defense is therefore real usage growth. Its biggest vulnerability is that suppliers can demand commitments extending far beyond the quarter that demonstrated profitability. Anthropic also warns that it’s changing its cash-flow profile through infrastructure investment. Customers who adopt Claude more deeply can support that expansion; customers who shift workloads to cheaper competitors can leave Anthropic carrying the bill.

I’d give Anthropic more credit for its operating business after this filing. Over the next two years, Anthropic can retain more value than a simple model-commoditization story allows if customers keep expanding faster than the cost of serving them. But Anthropic’s prospective shareholders still need an offer price before judging the stock. Compute suppliers collect against contracted demand while shareholders carry the risk that later capacity earns less.

Anthropic’s next quarterly disclosure needs to answer one question: does operating profit survive the next increase in delivered compute?

Ovintiv Buys Wells It Can Run Cheaper

Ovintiv is making a narrower bet than a forecast that oil stays expensive. Ovintiv says it can extract more value from Logan Energy’s assets because it already operates nearby.

Ovintiv’s October 26 agreement values Logan at approximately US$890 million, including expected net debt. Ovintiv gains acreage in Alberta’s Montney, two oil batteries, one expected online by early 2027, and interests in gas-processing facilities. Ovintiv would also gain existing production of roughly 20,000 barrels of oil equivalent a day, including about 7,000 barrels of oil and condensate.

Ovintiv’s production mix matters because a buyer can overpay for the word “oil.” Gas and other liquids contribute much of the acquired output. Ovintiv is buying a connected operating system: wells, land and equipment that separates and processes what comes out of the ground. Ovintiv can then decide where another dollar of drilling earns the best return across its enlarged footprint.

Ovintiv projects approximately US$60 million in annual savings, including capital, production and overhead costs. Ovintiv also estimates that its drilling costs on the acquired assets can be more than US$2 million lower per well. Ovintiv expects to close in early 2027, subject to approvals. Shareholders can check management’s savings forecasts against the costs Ovintiv actually reports.

Ovintiv doesn’t need an entirely new customer market to capture savings from equipment and operating routines it already uses.

What makes this better than simply buying more reserves? Ovintiv can spread its existing capabilities across additional wells. If you own nearby processing capacity and know the formation, another parcel can be worth more to you than to a standalone operator. A distant bidder can’t easily reproduce Ovintiv’s local fit.

Ovintiv still faces the obvious objection: cheaper wells can lose money when commodity prices fall far enough. Ovintiv plans to fund the transaction with cash, its credit facility and commercial paper, so financing costs also enter the return. Logan shareholders would retain a separate development bet through the planned Logan Oil spinout described in Ovintiv’s announcement.

I’d bet more on Ovintiv’s ability to improve these assets than on a lasting rise in oil prices. Ovintiv shareholders are positioned to benefit over 2027–2029 if the company delivers lower well costs without letting acquisition borrowing consume the savings. Ovintiv’s first post-close drilling-cost disclosures will show whether the promised US$2 million advantage actually appears.

The Contrarian Take

Everyone says: An $80.7 billion loss shows that Anthropic’s current business burns cash on an almost unimaginable scale.

Here’s why that’s wrong: Anthropic’s financing revaluation accounts for $72.4 billion of that loss, while its June quarter produced an operating profit. Anthropic’s real financing problem lies ahead: it has promised suppliers money for capacity that must earn revenue in later periods. Investors who dismiss the business because of the headline loss miss its operating progress. Investors who celebrate the quarterly profit without reading the commitments miss the risk they’re being asked to finance.

Under the Radar

India’s trade talks have reached the wording stage. Commerce Minister Piyush Goyal says most US–India discussions are complete and negotiators are fine-tuning the agreement, according to ANI’s account of his Hague remarks. Exporters gain planning certainty only when the final text specifies access and implementation. A small unresolved product list can matter more to an individual manufacturer than a broad declaration of progress.
Cloudflare has put a number on a shared-service failure. Cloudflare’s new incident disclosure identifies 155 responses containing other customers’ data, confined to one hosted model during an August–September window. Cloud customers buy isolation alongside intelligence. Hosting companies therefore face a recurring cost of proving that efficiency improvements preserve customer boundaries; cheaper model access alone won’t settle enterprise procurement.

Quick Takes

Microsoft is buying flexibility before reporting the savings. Microsoft’s voluntary exit offer covers about 60,000 eligible employees, with decisions due in early December. Microsoft told GeekWire it expects a much smaller acceptance share than its earlier retirement program. Employees choose whether to leave, so the offer creates uncertainty over both the final payroll saving and which expertise departs. Microsoft’s eventual restructuring charge and subsequent payroll expense will be more informative than the eligibility count. (Source)
Broadcom is supporting the collateral behind its own sales. Banks and Blackstone announced US$63 billion of committed financing for equipment to be leased to Anthropic, with deliveries expected from 2027. Broadcom provides residual-value support equivalent to the senior financing portion. Broadcom gains a funded route to orders, while accepting exposure to what the equipment is worth later. Supplier financing can expand a market and keep part of its risk with the supplier. (Source)
EQT’s headline value describes the combined company. EQT’s all-stock Range Resources agreement creates a business with more than US$50 billion of enterprise value. Range holders receive 0.987 EQT shares per share and would own approximately 27% of the combination. EQT expects US$325 million in annual base synergies within two years. Shareholders are joining their exposure to gas prices and operating execution; the exchange ratio makes the division of future gains explicit. (Source)
OpenAI has made the mid-tier price fight harder. GPT-6.1 Terra launches at US$1 per million input tokens and US$5 per million output tokens. OpenAI’s cheaper computer-use offering puts pressure on application businesses charging for access to a capable model. Applications that own customer data, permissions or an indispensable workflow have a stronger defense than those selling a convenient chat interface. The technical performance comparison belongs in today’s AI briefing. (Source)

The Thread

A supplier can help finance a sale and still carry the risk that the equipment loses value. Broadcom’s collateral support adds something to yesterday’s story about who can borrow: committed orders can expose the seller as well as the buyer to weak future demand. Ovintiv has a more direct way to improve its return, cutting the cost of operating neighboring wells, while Anthropic needs customers to keep expanding into capacity it has committed to pay for. Ovintiv’s borrowing costs can consume its operating savings; Anthropic’s upfront compute spending can absorb cash even after its reported profit improves.

Predictions

New predictions:

I predict: By March 31, 2027, Microsoft will disclose a charge or expense estimate specifically associated with its October 2026 Voluntary Exit Offer in an earnings release or SEC filing. (Confidence: medium; Check by: 2027-03-31)

Generated October 27, 2026, 03:29 AM ET. Issue date: 2026-10-27.