Nvidia posted quarterly revenue of $96.2 billion (€82.4bn) on Wednesday, comfortably beating the $92.2 billion (€79bn) Wall Street had expected, as chief executive Jensen Huang declared that artificial intelligence had reached “its inflection point” and guided next quarter revenue to $108 billion (€92.5bn) — again, above forecasts.
“AI has reached its inflection point,” Huang said in a statement. “It’s doing useful work. Its tokens are productive and profitable. Now, compute is revenue, and demand is accelerating.”
Shares fell 1.8% in after-hours trading shortly after the release. The stock had already ended the regular session 1.6% lower.
Days on which Nvidia reports its quarterly results have an outsized importance for the world’s most valuable company.
At a market capitalisation north of $5 trillion (€4.3tn) it is the most valuable company in the world, worth more than the GDP of Japan, the fourth largest economy in the world.
Yet there is a strange rite that plays out each quarter: Nvidia beats expectations, and the stock falls anyway. It happened after the first-quarter beat this year, when shares slid close to 5% in the days that followed.
The pattern has left a dynamic where the company needs to significantly outperform an already-bullish consensus, or provide a stronger-than-expected outlook, to move the share price higher.
The concentration problem
The main concern around Nvidia is not whether it can grow; it is more of a question of who it is growing for.
The firm still draws the bulk of its revenue from Amazon, Google and Microsoft, each of which is now designing its own chips to reduce reliance on the company. The new disclosure will show precisely how large that dependence remains.
Chief Executive Jensen Huang’s commentary on demand into 2027 therefore carries more weight than any single figure. The question has grown sharper since July, when markets fell on doubts about whether vast AI investments will ever generate proportionate returns.
Nvidia’s answer has been to help finance the buildout itself. This month alone it assembled a $500bn (€428bn) capital pool with six Wall Street asset managers for data centre projects, and separately committed up to $105bn (€90bn) to back an OpenAI data centre in Pike County, Ohio, with an initial capacity of 4.25 gigawatts and an option for a further 3.75.
The China wildcard
Export policy to China has been one of the most volatile threads in Nvidia’s recent history and is likely to feature heavily on the earnings call.
Washington barred sales of the China-specific H20 chip in April 2025, reversed course, and Nvidia has since received approval to ship the more capable H200 chip to vetted Chinese customers.
Reports have surfaced of large allocations to ByteDance and Tencent, though Beijing has been encouraging domestic firms to limit their purchases and prioritise homegrown alternatives.
Vera Rubin and the road ahead
Nvidia’s current growth is being driven by its Blackwell chips, the generation of processors powering most AI data centres today. Their successor, known as Vera Rubin, is expected to begin shipping in the second half of the year.
Nvidia has a tradition of naming its chip architectures after scientists and past generations include Ampere, named after physicist André-Marie Ampère; Hopper, named after computer scientist Grace Hopper; and the current Blackwell chip named after mathematician David Blackwell.
Vera Rubin continues that pattern. She was an American astronomer whose observations of how galaxies rotate provided some of the strongest early evidence for the existence of dark matter, the invisible mass thought to make up much of the universe. She died in 2016 and is widely seen as someone who was overlooked for a Nobel Prize during her lifetime.
The company has pointed to an order backlog it says is worth around $1 trillion (€857bn) across 2026 and 2027, though that figure comes from company commentary rather than independently verified financial disclosure.
A week stacked with catalysts
Wednesday’s data offered no relief on inflation.
The personal consumption expenditures index, the Federal Reserve’s preferred gauge, rose 0.2% in July against expectations of 0.1%, leaving the annual rate at 3.7% rather than easing to the 3.6% forecast. Core prices held at 3.3% over the year, above the Fed’s 2% target for a 65th consecutive month.
The Federal Open Market Committee held rates at 3.50% to 3.75% in July, with three regional Fed presidents dissenting in favour of a quarter-point increase.
Markets currently put the probability of a September hike at around 40%.
Attention now shifts to Jackson Hole, where Fed Chair Kevin Warsh delivers his keynote on Friday morning, his first since taking office in May, 19 days before the next rate decision. The ECB’s Isabel Schnabel joins a panel the same afternoon.
US stock markets drifted through a quiet session on Wednesday after data showed inflation last month was a touch higher than economists had expected.
The S&P 500 edged down less than 0.1% and remains near the all-time high it set earlier this month. The Dow Jones Industrial Average dipped 0.2% and the Nasdaq composite slipped 0.1%.
Treasury yields ticked higher after the inflation update, which has traders still largely betting the Federal Reserve will raise the federal funds rate before the end of this year. Oil prices fell after another volatile session.