Nvidia Q2 earnings preview: another record quarter in the making?

Nvidia has spent two years teaching the market to expect the extraordinary, which is now the source of its biggest problem. The company reports second-quarter fiscal 2027 results after the US close on 26 August. Nobody doubts it leads in AI accelerators, and nobody doubts the hyperscalers are still spending. The question is whether the numbers can keep pace with a share price that already assumes they will.

What the market expects

Consensus sits at around $2.08 per share on roughly $92 billion of revenue. Nvidia's own guidance was $91 billion, give or take 2%, so a small beat is more or less baked in and would barely move the needle on its own. What the market actually wants is a strong outlook for the following quarter, because that is where the argument about digestion and the durability of AI spending will be settled.

The first quarter set the tone. Revenue came in at $81.6 billion, up 85% on the year, with data centre alone contributing $75.2 billion and non-GAAP gross margin holding at 75%. Management then guided higher for the second quarter and, tellingly, did so without assuming any data centre compute revenue from China. That last point is worth holding onto, because it means the coming numbers are cleaner than the China headlines make them look.

Bank of America is looking for more than consensus, modelling $94 billion to $95 billion of revenue and a guide toward $107 billion to $108 billion next quarter, well ahead of the $104 billion or so the Street currently carries.

Data centre is still the whole game

Data centre is not some early-stage growth story running off a low base. It is already the largest part of the business by a distance, and it is still compounding at a rate that ought to be impossible at this size. Plenty of firms grow fast when revenue is small. Adding tens of billions a quarter while keeping margins in the mid-70s is a different thing entirely, and it is really the only reason to own the stock.

Two things are driving it. The Blackwell ramp is still working its way through the big cloud providers, and underneath that sits the wider build-out across sovereign AI, enterprise, AI clouds and inference workloads. Jensen Huang has put the Blackwell and Vera Rubin opportunity at north of $1 trillion through to the end of 2027. That is the sort of number to treat with caution, but it tells you how management sees the shape of the cycle.

Vera Rubin is now starting to ship, alongside new Vera CPU ramps, and spot rental prices for GPU capacity have pushed to record highs. That is a genuinely useful signal. Customers scrambling for compute at ever higher prices is not the behaviour of a market about to roll over.

Margins are the part that matters most

Revenue stories are cheap these days. What separates Nvidia is that it is scaling faster than almost anyone and still throwing off exceptional margins, and that combination is what earns it the premium.

On the day, the margin line may tell you more than the revenue line. A big top-line number paired with slipping margins would immediately raise the question of whether competition, mix or system costs are starting to bite. Hold gross margin near 75% and the valuation is a far easier case to argue.

The obvious threat is memory. It now makes up 40% to 50% of the cost of building Nvidia's systems, against 15% to 20% not long ago, and prices are rising across the industry. BofA still sees margins settling only a little lower over time, around 73% to 74%, helped by long-term supply deals including the relationship with SK Hynix. The hit on the coming Vera Rubin racks looks manageable. Heavier pod-level systems are more expensive to build, but should stay a small slice of the mix.

The China overhang

China is still the biggest short-term argument, and it is not going anywhere. Export controls have left Nvidia unable to serve the Chinese data centre compute market without sign-off from both Washington and Beijing, and there is now the added complication of US authorities reportedly looking into whether Chinese buyers are getting hold of chips via other Asian countries.

The saving grace, again, is that guidance already assumes nothing from China data centre compute. If the quarter is strong on that basis, the bull case stops depending on any kind of thaw and rests instead on the demand that is already visible. Investors will still push for detail on licences and the pace at which domestic Chinese rivals are catching up. It just shouldn't be the thing that makes or breaks the print.

What to watch on 26 August

Beyond the numbers themselves, the things worth watching:

• Revenue near or above the $91 billion guide, with a strong outlook for the next quarter
• Data centre still carrying the business, with no hint of hyperscaler digestion
• Gross margin holding around the guided 75%
• Blackwell and Vera Rubin demand confirmed as broad, not concentrated in a handful of buyers
• Any steer on the OpenAI financing deal and when cash might be turned toward buybacks

The equity stakes in customers are likely to come up too. Critics reckon Nvidia's holdings in the likes of OpenAI and Anthropic flatter demand. The rebuttal is that what has been committed so far is a modest fraction of the free cash flow Nvidia expects to generate, which leaves plenty of room to keep buying back stock.

Nvidia stock price – technical analysis

It has been a choppy ride for Nvidia this year, though it still hit a new record high in May. Recent weakness found support twice around $190, leaving this as the line in the sand for those looking for a more substantial correction.

Historically, August is one of the four strongest months of the year for Nvidia, aligning with its earnings report for Q2. The month is positive 82% of the time since Nvidia’s IPO, providing a tailwind for bulls.

Recent price action saw resistance around $225, but a close above here opens the way to the record highs around $240.

Valuation is the real risk

​None of this makes the stock cheap, and it would be dishonest to suggest otherwise. A lot of future growth is already in the price, and any sign that AI spending is cooling could bring a sharp pull-back. That is the honest weakness in owning it here.

​But valuation only means anything set against the quality of what you are buying. Revenue is rising at a pace you simply cannot find elsewhere, margins are close to 75%, and free cash flow leaves the rest of the sector well behind. Nvidia generated $48.6 billion of it in a single quarter a year ago and still handed roughly $20 billion back to shareholders. The premium holds because this is the main platform supplier for the entire AI build-out, not just another chipmaker, and that stays true until the data says demand is fading or margins are cracking. The quarter does not need to be flawless. It needs Blackwell demand to look broad, margins to hold, and the next guide to point higher.