The Rally Has Reached the Part Where the Roadmap Matters

My instinct throughout the move was that the rally was real, but the final burst was not entirely about fundamentals. It was increasingly being carried by traders trying to recover exposure after cutting it at precisely the wrong time, short sellers being forced back through the door and an options market that kept adding fuel as prices rose.

Takeaways 

• The rally to record highs was built on genuine earnings strength, but the final acceleration owed plenty to record tech capitulation, short covering and an options market feeding the chase.

• Hormuz relief pulled oil, yields and the dollar lower, yet the proposed shipping arrangement is only the opening move. The harder negotiation still runs through Washington.

• SpaceX and AMD showed that beating estimates is no longer enough when investors are focused on future capital spending, margins, valuation and incoming share supply.

• China’s low-cost AI advance and the renewed threat from long-end yields leave the market with two underpriced questions: who captures the economics of rising token usage, and how much valuation can survive another bond-market flare-up?

The Roadmap Matters

The S&P 500’s sprint to fresh records was always going to end in a session like this. After roughly $3.7 trillion was added to the index’s market value, the rally finally began to churn as the upside chase ran into profit-taking, softer labour data and a market that had already spent much of its near-term ammunition.

My instinct throughout the move was that the rally was real, but the final burst was not entirely about fundamentals. It was increasingly being carried by traders trying to recover exposure after cutting it at precisely the wrong time, short sellers being forced back through the door and an options market that kept adding fuel as prices rose.

Goldman’s own flow data now gives that view some backbone. Between July 24 and July 29, its prime brokerage book recorded the largest sale of global technology in the history of the dataset, while semiconductor positioning flipped net negative for the year. Gross leverage also suffered its second-largest monthly decline on record. Days later, the Nasdaq 100 had rallied roughly 9% in four sessions and the semiconductor index had jumped more than 15%.

That was not a sudden discovery that technology earnings were better than investors had imagined. It was what happens when positioning is cleared out so aggressively that even a modest improvement in the news forces everyone to rebuild at once.

The options market amplified the move. Goldman’s derivatives desk described clients rushing back into upside exposure, with record call demand leaving dealers increasingly short gamma as the market rose. In practical terms, investors who had already sold were buying calls to catch up, while dealers hedging those calls were forced to buy more stock into the advance.

That is how a rally becomes a chase. It also explains why the move could reach a record without every piece of the fundamental roadmap falling neatly into place.

The first accelerant came from the Strait of Hormuz. Markets entered the week hoping that a geopolitical reprieve might develop into the beginnings of a peace arrangement, and the relief trade strengthened when Iran said it had reached an agreement with Oman on a proposed shipping route through the waterway. Oil fell to new cycle lows, Treasury yields eased, and the dollar drifted lower as investors priced a faster return of energy flows.

But as we argued in Oil Market Chatter: Iran’s Hormuz Gamble Could Backfire, Tehran’s attempt to use the world’s most important energy chokepoint as leverage was always a dangerous card to play. It may have created immediate pressure on Washington, but it also risked damaging Iran’s own economy, alienating regional neighbours and eventually pushing the United States beyond the point where negotiation remained politically acceptable.

The Oman proposal may provide a route away from that outcome, but the easy part was producing the headline. The difficult work now begins with determining the terms, securing US acceptance and deciding what Tehran is actually prepared to concede.

President Trump said talks were moving well and that the Strait would reopen soon, but he paired that optimism with the warning that Iran would be hit hard if it failed to cooperate. That is not the language of a settled agreement. It is a reminder that the process can still change direction very quickly if Trump concludes that Tehran is buying time, preserving too much leverage or asking Washington to accept terms he cannot sell domestically.

The market has already priced a decent portion of the relief. From here, it needs ships moving normally, insurance and freight costs falling and an arrangement that survives more than one news cycle. Until then, Hormuz remains a negotiation rather than a resolution.

The second part of the roadmap is the US data. ADP employment disappointed, while the services surveys delivered an uncomfortable mix of softer hiring and rising prices. That pulled Treasury yields and the dollar lower and reduced near-term rate-hike expectations, but it hardly provided the clean disinflationary signal equity bulls would prefer.

This leaves Friday’s payroll report carrying even more weight. A softer number could extend the relief through lower yields, but too much weakness would shift the conversation toward the durability of growth. A stronger print would bring the inflation and Federal Reserve questions straight back into the market.

Minneapolis Fed President Neel Kashkari argued that the central bank should begin raising rates gradually and said three increases before year-end were not impossible. The front end largely looked through those comments, but the broader warning should not be ignored. The equity market may be treating lower oil as an inflation release valve, yet the Fed is still staring at service-sector price pressure and an inflation rate that remains above target.

The market is not merely waiting for payrolls. It is waiting to discover whether the bond market will allow the equity rally to keep its valuation.

That risk was also clear in Goldman’s institutional commentary. Tony Pasquariello identified the bond market as the most immediate danger to equities, recalling the late-2023 rise in long yields that cut deeply into the Nasdaq. His broader expectation is for steeper curves and periodic flare-ups around global debt loads.

That may be the most important line in the entire discussion. The market has just staged a record-high melt-up powered partly by short covering, call buying and performance anxiety, while the long end remains capable of repricing the entire trade. Earnings can remain strong, and AI spending can continue, but neither grants equities immunity from a renewed rise in the discount rate.

SpaceX provided the clearest single-stock version of the same problem.

As we discussed before the inaugural earnings report, this was never simply a question of whether the company could beat quarterly estimates. The trade was built around a limited float, exceptionally heavy short interest and a share unlock large enough to overwhelm almost any ordinary earnings result.

That setup delivered almost exactly as expected. The stock squeezed higher into the event, the results were respectable and investors then turned back toward the roadmap and the approaching supply.

SpaceX reported stronger-than-expected Starlink subscriber growth, improving Starship launch cadence and robust demand for AI compute. Nvidia also benefited after the company indicated it would standardise its AI infrastructure on Nvidia hardware and lift compute capacity substantially through 2027.

Yet SpaceX shares still fell sharply because the market was no longer asking whether the quarter was good. It was asking how much capital the AI strategy would consume, when the orbital data-centre ambitions would generate an adequate return and who would absorb the first $101 billion of stock becoming eligible for sale.

Up to 911.5 million shares can enter the tradable pool in the first unlock, potentially lifting the float from roughly 639 million shares to as many as 1.55 billion. By December, the staged process could increase available shares to more than 5.3 billion.

Against that backdrop, a headline earnings beat is little more than one input. Early investors sitting on enormous gains may take money off the table, special-purpose vehicles may begin distributing stock, and short sellers will continue trying to exploit every wave of incoming supply.

The squeeze can still return. With roughly 35% of the existing float sold short, any positive catalyst can force a violent rebound. But between now and December, the stock is likely to trade as much on supply, positioning and capital intensity as it does on rockets, satellites or AI.

AMD offered a quieter version of the same message. Its quarter was strong, but expectations for server share gains and the MI-series accelerator ramp were already high. Investors looked beyond the beat toward forward margins, execution and the cost of competing in AI infrastructure.

That is the new hurdle for the wider AI trade. A company can beat the quarter and still fall if the roadmap demands more capital than the market expected.

This brings us back to the AI token blind spot we discussed yesterday. The market has spent months celebrating soaring token usage as proof that AI demand is becoming deeply embedded across the economy. That conclusion is probably right. The unanswered question is whether the companies producing the most expensive American models will capture enough of the economics to justify their valuations and spending plans.

China’s latest model releases make that question harder. Alibaba, Moonshot, DeepSeek, Z.ai and ByteDance are no longer producing occasional surprises. They are building a repeatable system for delivering models near the global frontier at dramatically lower cost.

The competitive threat is not necessarily that every user abandons the premium US models. Developers may continue to use those systems for architecture, planning and the hardest reasoning tasks, while handing execution to cheaper Chinese agents. That hybrid approach would increase total token consumption while steadily commoditizing the execution layer.

This is where the market risks confusing volume with value. More tokens do not automatically produce greater pricing power. If Chinese open-source models can complete a growing share of agentic workloads for pennies rather than dollars, US providers may be forced to reduce prices while their compute and capital requirements continue rising.

The AI boom can therefore keep expanding even as the economics become more difficult for parts of the stack. Cloud providers, chipmakers and infrastructure owners may still prosper from the sheer growth in compute demand. But the model makers themselves could face a harsher pricing environment than current valuations imply.

That is the blind zone. The market sees every additional token as confirmation of the boom, but it has not fully decided who earns the margin.

There is also an earnings-quality question beneath the headline excitement. Goldman noted that aggregate earnings growth of roughly 46% falls to around 28% once private-company effects are stripped out. The median stock still delivered a healthy 12% increase, so the earnings season has hardly been weak. But the gap shows how much of the headline strength is tied to assets and investment vehicles that are difficult to value in real time.

At the same time, AI capital-spending ambitions are running far ahead of previous estimates. When companies begin describing enormous infrastructure outlays as having near-immediate paybacks, investors should listen carefully—but they should also keep one hand on the calculator.

My trader’s read is that the bull market still deserves the benefit of the doubt. Earnings are growing, economic activity remains resilient, and the recent positioning washout removed a large amount of speculative excess. But the character of the opportunity has changed.

The market is no longer lifting from an underowned, deeply washed-out base. It is sitting near records after a huge rebound, with investors having chased calls, rebuilt technology exposure and priced a meaningful geopolitical reprieve.

From here, the roadmap has to do more of the work.

Hormuz must move from hopeful negotiation to durable reopening. Payrolls must cool inflation concerns without breaking the growth story. SpaceX must absorb an unprecedented supply event while persuading investors that its AI ambitions justify the capital required. US model developers must demonstrate that rising token usage still produces durable margins in a world where Chinese competitors are offering increasingly capable alternatives at a fraction of the price.

Above all, the bond market must remain cooperative.

The rally has not run out of reasons to rise. It has simply reached the point where positioning can no longer carry the entire argument. The next leg will need evidence that the road ahead is strong enough to support the valuation the market has already travelled.