Lula Wins, Lenders Decide

If You Only Read One Thing

The winner still has to persuade the lender. Lula’s reelection preserves Brazil’s policy direction; Musk says SpaceX’s financing gives him more room to pursue his. His claim to have closed $40 billion of GPU debt points to the advantage both stories turn on: whoever can promise future income gets more freedom to decide what happens next. Ambition alone won’t secure that freedom.

Lula Keeps the Presidency, Inherits the Bill

Lula has won another term, but Brazil’s lenders will help determine how much of his programme he can deliver. He’s keeping the presidency after investors had positioned for a change in government. Now he’ll have to reconcile promises of support with the cost of financing them.

Brazil’s electoral court reports that Lula won 51.79% of valid votes against Flávio Bolsonaro’s 48.21%. Brazil’s Senate identifies the coming term as 2027–2030. Lula has secured four more years to pursue his priorities, starting from a closely divided electorate.

Lula also has a reasonable case against the market’s instinctive pessimism. A government can support household incomes and improve its finances if stronger activity raises tax receipts. Brazil’s exporters can earn abroad even when domestic credit is expensive. Anyone treating the election as a uniform verdict on Brazilian companies will miss those differences. For the next year or two, I’d favour businesses with foreign-currency earnings and limited refinancing needs over a broad bet on cheaper domestic borrowing, though I wouldn’t bet heavily before Lula’s fiscal plans emerge.

Why can an election victory make governing more expensive? Lenders can demand more interest if they expect the government to struggle to cover future bills. Brazil then spends more servicing debt, leaving Lula less room for the programmes he defended during the campaign. Businesses refinancing alongside the government face the same competition for money.

October 5’s briefing examined the first-round bet on fiscal change. Lula has now beaten that bet. Reuters reports that Brazilian stocks and the real strengthened after Bolsonaro led the first round, with traders anticipating tighter public finances. Monday’s opening prices will therefore include the unwinding of that positioning. A sharp move could tell us more about the bet investors made than the earnings companies will generate over four years.

Lula’s foreign-policy options give exporters another source of support. The European Parliament’s pre-election analysis describes his emphasis on industrial policy and national sovereignty. Brazil also enters this term with the EU–Mercosur interim trade agreement already provisionally applied. Companies able to sell across several markets have more room to absorb a difficult domestic financing cycle.

Lula can overturn the cautious view by pairing support programmes with durable revenue or spending measures. His first post-election fiscal package, and whether legislators pass its financing provisions, will reveal how much policy freedom the victory actually bought.

SpaceX Borrows Against More Than AI

SpaceX’s claimed financing advantage comes from businesses outside the AI boom. Musk said on Sunday that SpaceX closed $40 billion of GPU debt because it has investment-grade credit and revenue from Starlink and launches. He’s describing a competitive advantage that a specialist AI cloud can’t quickly copy: another business to help carry the bill.

Musk’s announcement advances the story beyond earlier reports of financing difficulties. Our October 21 briefing examined the gap between SpaceX’s AI revenue and operating profit. SpaceX now has a claimed financing close to put beside that operating picture. Its quarter-end filing separately records $38.006 billion of debt principal as of September 30; the new facility’s terms are still the missing piece.

A lender financing SpaceX can assess several sources of repayment. A lender financing a specialist GPU cloud has to put more weight on customers continuing to rent its machines. You can have busy servers and still struggle to finance the next building if lenders distrust the income beyond existing contracts.

CoreWeave shows how lenders can accommodate that risk. Its August credit agreement provided a $2.6 billion facility for customer-linked infrastructure through a subsidiary, backed by pledged assets and an unconditional parent guarantee. CoreWeave was offering lenders specific claims on assets and the parent company alongside the prospect of AI growth.

Sure, Musk has good reason to portray rivals as fragile. Bloomberg reports (paywalled) that financing talks are still ongoing despite his announcement. CoreWeave’s earlier borrowing shows that specialists have financing tools too; its August agreement can’t tell us what a bank would approve now. But lenders can respond to uncertainty by demanding more protection, charging more, or funding less. Every concession leaves the cloud provider with less of the eventual return.

SpaceX and diversified cloud owners look better placed to fund expansion over the next year than operators dependent on repeated refinancing. I’m fairly sure of that relative advantage. Suppliers of GPUs and electrical equipment can still receive orders from the stronger borrowers, even while weaker operators retreat. AI builders increasingly depend on scarce inputs beyond the chip itself, including finance.

Musk says the GPUs arrive in the first half of 2027, according to SiliconANGLE’s account. SpaceX’s next financing disclosure will show whether its borrowing price, guarantees and repayment schedule support the advantage Musk claims.

The Contrarian Take

The easy conclusion is that a credit squeeze will make AI compute cheaper as distressed operators sell capacity. I’d expect the opposite pressure on new capacity first.

A cloud provider that loses financing has to postpone purchases or construction. A stronger rival can wait for better terms, while chip and power suppliers concentrate deliveries on customers able to pay. Customers seeking additional compute then have fewer expansion projects to choose from.

CoreWeave’s existing financing structure explains why the timing matters: the provider borrows before collecting years of customer payments. Interrupt that sequence and future supply suffers. Prices for distressed, already-running machines could fall at the same time. A cheap liquidation offer and an expensive commitment for dependable new capacity can coexist. Customers care which one can actually run their workload for its full duration.

Under the Radar

CoreWeave’s customers can leave before its lenders do. CoreWeave says its August facility has roughly five years to maturity, while the underlying customer contracts average about three. CoreWeave therefore has to renew contracts or find replacement customers while the debt is still outstanding. Lenders accepted that renewal risk when the deal closed. The roughly two-year difference helps explain why future demand confidence can change financing conditions well before today’s servers go idle. The company describes the arrangement here.
Brazil’s European option already exists. The EU–Mercosur interim trade agreement has applied provisionally since May 1, according to the European Commission’s implementation notice. Lula can pursue trade diversification through an operating framework. Exporters that already meet European requirements are better placed to use that access than firms waiting for a fresh political announcement. Their advantage comes from being ready to ship when the route opens.

Quick Takes

Serbia’s students now have to govern. Aleksandar Vučić conceded parliamentary defeat to the student-backed opposition after nearly two years of protests, AP reports. The movement has converted public anger into electoral power. Its next challenge is keeping a diverse coalition together while staffing a government. Europe gains a new negotiating counterpart; businesses face a transition in who authorizes decisions and honours commitments. (Source)
Apple’s repair score can rise while reliability disappoints. iFixit awarded the iPhone Duo a provisional 5/10, its best foldable score, while investigating early screen failures. Its simulation suggests a hinge support bar could stress the display. Apple’s commercial exposure depends on how often failures occur and how expensive repairs become. Easier battery removal won’t settle the economics of replacing the inner screen. (Source)
A payment rail can’t supply a willing counterparty. Jewish Insider reports that the Bank of Israel rejected Netanyahu’s suggestion that crypto could help circumvent financial sanctions. Netanyahu later qualified his proposal. An exporter still needs customers, banks and investors willing and legally able to transact. Moving funds through a different network leaves those relationships exposed, which is why the argument matters to Israel’s technology businesses. (Source)

The Thread

Lenders can change a competitive race before customers change their minds. SpaceX can draw on income from several businesses while a specialist cloud negotiates against a narrower repayment base. Brazilian exporters earning abroad can have more room than domestic businesses refinancing into the government’s own borrowing needs. In both cases, reliable income buys time to keep investing. Once competitors cut expansion, the better-financed company can gain ground without building a better product that quarter. The eventual return still depends on what it paid for that freedom.

Prediction Ledger

Weekly Scorecard

Three federal science agencies would announce qualifying funded pilots by October 20. Made July 22, medium confidence. Pending verification: NSF’s X-Labs page confirms named opportunities, but this audit hasn’t identified the required three agencies. Several topics from one agency don’t meet the forecast.
A Gulf government or national oil company would approve binding spending for more than one million barrels a day of Hormuz-bypass capacity by October 20. Made July 20, medium confidence. Pending verification: reports of construction progress need a dated approval and capacity scope before this can be scored.

Both deadlines have passed. These are overdue evidence checks, and the older unresolved forecasts remain in the tracker. I’m adding no new prediction today: the backlog needs verified outcomes first.

2026-10-26 · 03:24 AM ET