Core PCE Holds at 3.3% | Nvidia Gets the Last Word | The Ten-Year Backs Up | The Print That Did Not Free the Fed

Stocks barely moved Wednesday because the market was waiting for Nvidia.

The S&P 500 was flat. The Nasdaq slipped slightly. The Dow lost 0.2%. The VIX fell 1.55% to 15.21.

The 10-year yield rose to 4.65%. Oil fell 0.5% to just under $82. Gold dropped 0.9%. The dollar was stronger on the day.

The Fed did not get a clean inflation answer.

Headline PCE rose 0.2% in July and 3.7% over the year. Core PCE rose 0.2% on the month and 3.3% over the year.

The core number matched expectations. It did not fall.

Personal income rose 0.4%. Spending rose 0.2%. The consumer still spent, but not fast enough to make the print feel clean.

The tape waited for the other answer after the bell.

Nvidia (NVDA) had to prove the AI buildout still works.

Core PCE did not break higher. That was the good news.

It also did not cool.

Polymarket had already priced the answer. Core PCE at 3.3% carried a 73.5% chance. The 3.2% leg sat at 21.5%. Anything at 3.1% or below barely traded.

So the print did not shock the market.

It confirmed the range.

That matters because the Fed's preferred gauge still starts with a three. The target is two. A September hold can survive that. A cut case cannot.

The Fed book showed the same line. September no-change sits at 66%. A 25 basis point hike is 33%. October no-change is 72%. December no-change is 60%, with a hike at 32%.

The Sticky Number

Inflation was not hot enough to force September. It was not cool enough to free the Fed.

The inside of PCE kept the same split.

Goods prices fell 0.1%. Gasoline and energy goods fell 2.7%. Furnishings and household equipment fell 0.9%.

That helped.

Services did the other thing. Services prices rose 0.3%. Financial services and insurance rose 1.2%. Housing rose 0.3%.

That is the Fed's problem.

Energy can fall for a month. Services have to slow for a regime.

The 10-year yield rose anyway. It finished at 4.65%. The front end did not get enough relief to reprice the policy path. The long end did not get enough relief to forget oil, deficits, or term premium.

The Service Line

Goods gave the market a discount. Services kept the Fed in the room.

Nvidia became the whole market's after-hours test.

Analysts expected earnings near $2.10 a share on revenue around $92 billion. That would be close to double last year. For the next quarter, the Street wanted guidance near $104 billion.

The product focus was Vera Rubin. Shipments have started to major customers including Microsoft and OpenAI. Jensen Huang has said Blackwell and Vera Rubin could drive $1 trillion in sales through 2027.

That is the upside.

The risk is how the money gets there.

Last quarter, about half of Nvidia's sales went to a small group of internet giants. Those buyers are still funding massive data-center spending. Memory costs are rising. Margins are expected near 75%. Nvidia has also helped arrange up to $500 billion of AI infrastructure financing.

The test is no longer demand alone.

The Customer Ledger

Nvidia must show growth is broad, margins are intact, and the buyer is not just being financed back into the order book.

Oil fell, but Hormuz did not clear.

WTI closed near $82. The market got some energy relief on the day. Then oil bounced off its lows after Iran's Revolutionary Guard said Tehran reached a Hormuz revenue-sharing deal with Oman.

That is not normal traffic.

Polymarket puts Strait of Hormuz traffic back to normal by September 30 at 10%. October 31 sits at 18%.

Those odds remain low because the physical route is still constrained.

A revenue-sharing deal can be a step toward a route. It can also be a toll by another name.

The Route Price

Oil can fall on the day. The Strait is still priced as broken.

Prediction markets kept raising money while the rulebook tightened.

Kalshi has sold about $1.12 billion of a planned $1.5 billion equity offering. The raise began in April and is now about three quarters sold.

That shows capital still wants the category.

Polymarket showed the other side. It withdrew several NFL player-participation market filings one day after they were certified. Those markets sit close to injury and availability data. That is where insiders matter.

Sports event contracts are already sensitive. Player-status contracts are more sensitive.

The same issue has followed war, elections, and mention markets. The deeper these platforms move into real events, the more the market needs surveillance.

The Integrity Cost

Prediction markets are raising like exchanges. They are also learning what exchange-grade oversight costs.

Wednesday did not resolve the week. It narrowed the question.

The consumer confidence split said households like the present and fear the next six months. PCE said inflation is still sticky. Nvidia had to say whether AI demand is still strong enough to carry the market.

That is the same structure in three places.

The present is fine.

The forward read is harder.

Income rose. Spending rose. Core inflation stayed at 3.3%. The Fed can wait, but it cannot declare victory. Nvidia can report a huge quarter, but it still has to defend customer depth, memory cost, and financing risk.

The Split Verdict

Every current reading held. Every forward reading tightened.

Wednesday answered the morning with a narrow tape.

PCE reached July. It did not reach August. Nvidia reports after the close. Warsh speaks at Jackson Hole on Friday.

What is priced: core PCE at 3.3%, a September Fed hold, Nvidia as the AI test, and Kalshi still raising capital.

What is not priced: services inflation staying firm, Hormuz revenue sharing turning into a toll, Nvidia's customer base proving too narrow, or sports markets facing harder limits after player-status filings.

The print did not free the Fed.

The call has to carry the tape.

Capital moves early. Coverage catches up. The gap between the two is worth watching.