Nvidia's results show the AI trade is alive and well

Nvidia’s has delivered with another beat and raise in today’s quarterly earnings.

The company has crushed it on all the street’s key sensitivities.

Gross margins – which measure a company’s pricing power – have increased from 73% up to 75%. This suggests that Nvidia, like Apple, is becoming so powerful it can squeeze suppliers. More importantly however it also suggests that the much-vaunted threat of Google’s rival microchips isn’t materialising. When moats are breached the first signal tends to be shrinking margins.

Data centre revenue – key for the street as it proxies AI demand – is up an absurd 75% from a year ago. This result confirms complaints from the ‘hyperscalers’ – Microsoft, Google, Amazon, Meta – that a problem is they cannot get enough of Nvidia’s chips.

It also chimes with what’s happening to Nvidia’s old microchips on the secondary market. There is so much demand for Nvidia’s chips that even its old ones are becoming more expensive to rent. As JPMorgan notes these lease rates were the best indicator that a beat and raise was coming (table below).

Importantly, while hyperscalers still represent slightly over 50% of data centre revenue, the "rest of the market" is now the primary growth engine Nvidia said. 

Despite all this the market has nudged only slightly higher in after hours.

Revenue guidance confirmed

Nvidia has confirmed guidance. Nvidia – and the street – now has greater visibility into demand into 2027 thanks to the capex commitments of the hyperscalers. While the market has been spooked by high capex spend and its impact on hyperscalers’ free cash flow margins, for Nvidia these capex commitments translate directly into revenue growth. Much of the spend goes to them.

Jensen Huang stressed in the earnings call repeatedly that agentic AI – where AI stops acting like a glorified autocomplete and starts acting like a Jarvis from Iron Man – has hit an inflection point.

Huang stressed that the inflection point has come at the same as hyperscalers’ revenues are growing. This creates forced buying of Nvidia’s chips. They can’t take the 2 – 3 years to build out rival architectures to Nvidia’s. They have to move now or they get disrupted.

Crucially, Nvidia assumed China revenue is zero in its guidance. This means if it gets any revenue from China at all its pure upside. And China is highly likely to participate somewhat as it doesn’t want to be frozen out of CUDA.

Perhaps most importantly of all, Nvidia confirmed that Vera Rubin – its next generation of chip – is getting the cost of energy and costs of AI queries (“tokens”) down significantly, perhaps as much as 90%. This matters because it creates more room for Nvidia’s own customers to succeed and use AI to grow their own margins. (They’re the ones that pay those energy costs.) And this keeps demand growth healthy longer term.

Looking forward – optionality

Nvidia has confirmed that it will continue to make VC style investments in Anthropic, OpenAI, Coreweave and many others. These days “just about every startup in the world is on Nvidia’s ecosystem”, Huang noted on the call in what is probably an overstatement. Nevertheless, nearly every new AI model published on Hugging Face uses CUDA.

Some have raised the concern that this creates “circular financing risk” and justifies a compressed valuation. My own view is the opposite: that this creates unrivalled optionality and justifies a greatly expanded valuation. It means Nvidia’s shareholders participate in VC style AI activity without the 2% and 20% fee structure.

Most promisingly, Nvidia is also strengthening and broadening its ecosystem, becoming an all-encompassing AI picks and shovels trade. It is embedding its AI software, CUDA, deeper into chip design with its investments in chip designing company Synopsys. It is creating a more diverse set of chips to cater to broader needs (e.g. CPUs) and thanks to buying Groq it can increasingly do the Broadcom thing where it builds chips based on others’ requirements.

The next big thing – probably AI robotics and unmanned vehicles – hinges heavily on Nvidia’s infrastructure. Leading companies like Anduril are building on Nvidia. And means the next Nvidia will be Nvidia. 

About ETF Shares

ETF Shares is a low-cost index ETF issuer, based at the Macquarie University Incubator. We specialise in US-focused ETFs, such as the ETFS Magnificent 7+ ETF (ASX: HUGE), ETFS US Technology ETF (ASX: WWWW) and ETFS US Quality ETF (ASX: BEST).