Article

Nvidia's next test is the power grid, not the latest earnings print

by TechDefused Newsroom
The image captures a street view of Wall Street, showcasing the iconic Wall Street sign beneath American flags. The backdrop features a large advertisement displaying 'DATA + AI' and logos of associated brands. — Credit: Photo by Robb Miller / Unsplash cPhoto by Robb Miller / Unsplash
Photo by Robb Miller / Unsplash

Nvidia reports second-quarter results on Wednesday 26 August after the US market closes, in what has become the single most consequential earnings event on the global equity calendar.

The chip designer guided to revenue of around $91 billion for the quarter, roughly double the same period a year earlier, and consensus sits close to that figure.

Estimates diverge more than usual, with one widely cited compilation putting the average at about $91.8 billion within a range of $90.3 billion to $96.7 billion, and other previews pitching the consensus band higher at $93 billion to $95 billion.

Earnings per share are forecast at about $2.08.

The stock has risen roughly 17.7% this year to around $219, a respectable return that nonetheless leaves it lagging the sort of moves investors once took for granted from it.

The spending is not the question

Demand visibility is no longer really in doubt, which is what makes this quarter awkward to trade.

Amazon, Alphabet, Meta and Microsoft have collectively guided to about $725 billion of capital expenditure in 2026, up around 77% on last year's record $410 billion, with the bulk directed at data centres, graphics processors and custom silicon.

Alphabet, Amazon and Meta all raised those forecasts again at second-quarter results last month, and the market's reaction was instructive.

Alphabet shares fell 7% on the news, dragging the other three down with it, as investors registered discomfort at capital intensity that has already turned the company cash flow negative for the first time.

Part of the increase is not additional capacity at all, with Microsoft attributing an estimated $25 billion of its $190 billion calendar-year budget to higher memory chip and component prices.

That matters for Nvidia because it flags a cost squeeze running through the supply chain that capital alone does not fix.

Where the bottleneck moved

The constraint on the build-out has shifted from chip supply towards electricity, grid connections and construction timelines, none of which respond to a bigger cheque.

Microsoft has told investors it expects to remain capacity constrained through at least 2026 despite the spending.

Analysts have also begun questioning the sector's heavy reliance on natural gas generation to bridge the gap, a decision that looks efficient now and may look expensive later if power forecasts hold.

The positioning going in

Sentiment around the trade has fractured, which is itself unusual.

Michael Burry, the investor known for shorting the mid-2000s housing market, has increased bets against Oracle, Micron and Nebius, while Mark Cuban has warned publicly that the vendor financing arrangements underpinning AI demand could destabilise the market.

On the other side, Cathie Wood's ARK Invest bought about $40 million of Nvidia, Tesla and SpaceX exposure during last month's sell-off, and Dan Ives of Wedbush argues the sector has entered a monetisation phase rather than a speculative one.

The bulls need Wednesday's guidance to confirm that the Blackwell ramp is converting hyperscaler budgets into shipped revenue on schedule.

The bears need only a hint that it is not.

Flagging one issue with the source material: the transcript is a headline dump with no Nvidia specifics beyond the fact of the preview, so the figures above are drawn from published estimates and company guidance. Happy to tighten to 300 words or add a Blackwell and Vera Rubin section if you want more product detail.

by TechDefused Newsroom