RD @rex_dlm

Compute economics will be central to how attractive this business becomes. Anthropic generated $16.3bn in revenue in H1 2026, with $8.9bn in compute and infrastructure expenses. That represented 55% of revenue, down from 201% a year earlier. Revenue is scaling against capacity already contracted, producing substantial operating leverage. The sensitivity is significant. At that revenue base, reducing compute expenses from 55% to 45% of revenue would save roughly $1.6bn over six months, holding everything else constant. The capital requirements ahead still deserve scrutiny. As of June 30, Anthropic disclosed $265.6bn in future fixed or minimum purchase commitments, primarily for cloud hosting and infrastructure, across contracts spanning 1–20 years. Some additional compute arrangements have payments that remain unquantified. Those obligations extend over many years, making their payment schedules and the revenue generated by that capacity essential to the analysis. The upside could be enormous if compute cost per successful task falls faster than the price customers pay. That would allow more revenue to become profit. Shareholder returns will also depend on the upfront capital required to keep expanding capacity.

Larry Culp’s turnaround at GE keeps coming to mind. Restoring quality, accountability and operational discipline takes time. Sustainable production growth follows from getting those foundations right. Ortberg’s willingness to accept a slower ramp strengthens my conviction that Boeing can achieve a similar turnaround. Resisting pressure for faster output is part of rebuilding the business.

@TheStockerMan My view: part of the $HOOD premium is crypto optionality. If the cycle turns, that is upside this comparison leaves out.

@matthew_sigel Net interest is the non-trading line worth watching here. $HOOD earned $415M, 29.1% of Q3 revenue. Schwab is 48.5%, IBKR 60.3%. The share is only the start: funding cost and credit quality decide how much of it is lasting profit.

A brokerage account can generate revenue without another trade. Robinhood (@RobinhoodApp, $HOOD) earned $415 million in net interest revenue in Q3 2026, accounting for 29.1% of total revenue. The equivalent share reached 48.5% at Charles Schwab and 60.3% at Interactive Brokers. Cash and credit are substantial businesses within the brokerage model. Customers leave cash in their accounts, borrow against securities, and use other financing products. These balances can keep generating income between transactions. Almost any company can earn interest on surplus cash. Building a large, consistently profitable interest business requires the right funding economics. Consider a hypothetical lender with $100 million in funding costing 3% annually. It lends that money at 7%. Annual interest income reaches $7 million. Funding costs consume $3 million. Operating expenses take another $1 million, while credit losses absorb $1 million. Pretax profit: $2 million. If funding costs rise from 3% to 5%, that profit disappears, holding everything else constant. An additional $2 million in credit losses produces the same result. This makes three capabilities particularly valuable. Low funding costs leave more income available to cover expenses and losses. Stable funding reduces the pressure to replace departing money at a higher price. Stability refers to the overall funding base, even as individual customers move money in and out. Disciplined underwriting helps preserve the spread after borrowers fail to repay. Customer relationships become financially valuable when they support these capabilities. Customers who maintain balances over time can provide a more predictable funding base. The benefit compounds when acquiring and retaining those balances remains inexpensive. For investors, the useful questions follow the money: where funding comes from, what it costs, how reliably it stays, and what risks the financed assets carry. Robinhood’s 29.1% revenue contribution establishes the scale of its interest business. Funding quality and credit discipline determine how much of that revenue can become repeatable profit.

Just starting and barely there annalist-fin.pages.dev Trying to build institutional-grade financial analysis using available public data. Another exciting features such as Deal Labs (simulate LBO, mergers, acquisitions) and Atlas (map relationships in term of credit, supply chains between companies) are on the roadmap.

@hthieblot Hi! It’s me.

Sigh….arsenal lose again :(

@Coinmaster100x And robinhood:0xdee52f2ab639b6942b0d0f0565400b93b7a0fbe5 will lead the way up

Down ~10% for the day Boeing $BA looks interesting. Market cap is below 150B. I still believe that Ortberg will succeed in the long run.

@ke0in @NotMattGeorge And it seems like market reacts positively to the announcement

@RobinhoodCrypto And more onchain activities please…

@AlphaProMaxxx Totally agree with you on robinhood:0x39dbed3a2bd333467115de45665cc57f813c4571. Robinhood chain activities has been decreased by 42% and that’s not good for $PONS at this time being. But crypto is a cycle and it can always comeback somedays.

@escship Congrats and way to go!

Just finished working and now it’s time to continue hustling. I cannot sit idly. I am driven by future ambition.

@anupamrjp Still prototype but worth to look at annalist-fin.pages.dev

@I71360 Yes bro!

I am building Annalist. Annalist is a verified record of company financials, built from official filings and nothing else. It publishes the statements, ratios and findings of each company, with every figure linked to the document it came from and each ratio shown with its formula and inputs. A new report produces a new snapshot with its own identifier, so a number cited today can be found again in the same form tomorrow. Credit is where the analysis starts: credit strength, liquidity and earnings quality, the questions that matter most for the debt and equity of a company. Returns on capital, capital allocation and segments come next, built with the same discipline. I build this because the numbers that investors and analysts depend on are either expensive or unverified. A single Bloomberg seat costs US$31,980 a year. Even the established databases are not exact: one study found that 6.5% to 7.7% of the amounts in widely used data sources disagree with the 10-K, enough to change the results of bankruptcy and earnings quality models. AI is fast, yet the best model scores 67.85% on the tasks of an entry-level financial analyst. And comparing companies remains slow work, since fiscal years end on different dates, filers restate earlier figures and the same line carries a different definition from one company to the next. It's still early, but the core already works. A value is checked against other official figures before it is published, and a value that the checks cannot prove is withheld, with the reason shown. Companies with different fiscal years line up by calendar quarter. AI agents can connect through a read-only MCP server and quote the filing behind each figure they use. My vision is a verified and comparable record of companies across the major markets of the world, listed and private, open at low cost to investors, analysts and the AI agents they work with. A credit analyst in Jakarta should be able to work from the same figures as a fund in London, and an AI agent should be able to cite a number as precisely as a careful analyst would. If you invest, analyse companies or build AI for finance, I would be glad to hear what you need from it.

@solyatic Hi I am new here, let’s connected. I am building an open, curated, and standardized financial analysis data that can be consumed by anyone including AI agents.