Chip designer Nvidia (NVDA:NASDAQ) has become the most-watched stock on the planet in recent years, having built a fast-growing business empire in one of the most exciting investment themes ever: AI (artificial intelligence).

The date for your diary: 26 February, that’s when Nvidia will report earnings for its fourth quarter and full year for fiscal year to end January 2025.

Soaring demand for its data centre GPUs (graphics processing units) have become the gold standard for developing AI, catapulting Nvidia’s revenue, profits and cash flows higher, and adding more than $3 trillion to its market cap since the start of 2023.

Since then, Nvidia has barely put a foot wrong, beating market expectations time and again, but the market’s confidence was given an almighty shake in recent weeks. Last month, Chinese hedge fund-backed start-up DeepSeek released R1, its open-source reasoning model that reportedly outperformed the best models from US companies, such as OpenAI.

R1’s self-reported training cost was less than $6 million, a fraction of the billions that Silicon Valley companies are spending to build their artificial intelligence models. That said, there remains healthy scepticism among analysts over those cost claims, and the Nvidia chips that power R1.  

One sceptic is Alexandr Wang, the CEO of Scale AI and the world’s youngest self-made billionaire, who claimed that DeepSeek and other Chinese labs have gotten their hands on more cutting-edge Nvidia H100 chips, which it can’t admit to because of US trade embargos on advance tech kit. Wang’s view has been backed by Elon Musk and Ted Mortonson, a managing director and tech strategist at Baird, who also doubts DeepSeek could have built its model with reduced-capacity H800 chips. 

‘Pegging R1’s price tag at $6 million is wildly misleading,’ said Gavin Baker, managing partner and CIO at Boston hedge fund Atreides Management.

Crucially for investors, Nvidia’s comments in response to R1’s launch indicate that it sees DeepSeek’s breakthrough as creating more demand for its GPUs. 

‘Inference requires significant numbers of Nvidia GPUs and high-performance networking,’ the company has said. ‘We now have three scaling laws: pre-training and post-training, which continue, and new test-time scaling.’

How all of this plays out will surely keep investors on the edge of their seats through 2025 and beyond, but analysts have so far not indicated any meaningful change to rampant growth forecasts. Consensus data from Koyfin for the upcoming results indicate Nvidia quarterly and full year EPS (earnings per share) of $0.84 and $2.95, on revenue of $38.08 billion and $129.19 billion.

Estimates for 2026 and 2027 put sales growth at above 50% and 20%, implying EPS of $4.44 and $5.60 over the next two years. This apparent analyst confidence that Nvidia can hang on to AI chip design leadership and keep the cash flows flowing thick and fast also seems to be returning for investors, given how quickly the stock has recovered from the sharp sell-off straight after R1’s release.

Nvidia shares lost 17% on 24 January, a $600 billion capital value loss, the biggest on record. The stock, at $138.85, is now less than $4 below the $142 close before the DeepSeek release, and less than $10 off the all-time $149.43 high set in early January 2025.

‹ Previous2025-02-20Next ›