Nvidia without Jensen: hold, don't fold

We do not expect a material disruption to NVIDIA’s reported results over the next 18 to 36 months. The company has already locked in substantial multi-year supply and capacity commitments, and key customer arrangements extend well into the decade. The product roadmap through the early Vera Rubin generation is largely set. Existing contracts and residual-capacity arrangements (including CoreWeave’s residual-capacity order through April 2032) provide visibility that is independent of any single individual’s presence.

NVIDIA disclosed commitments by fiscal period (as of 26 July 2026)

The named customer contracts run longer still. NVIDIA’s US$6.3 billion capacity agreement with CoreWeave runs to April 2032, and its roughly US$3.4 billion cloud agreement with IREN also runs to 2032. Support for the OpenAI campus in Ohio phases in from NVIDIA’s 2029 financial year.

So where is the risk? It sits in the next wave of deals, not this one. Huang was the person who could get a customer, a chipmaker and a lender to commit before any of them knew the others would follow. That matters for orders placed in 2028 and beyond. It matters far less for the revenue NVIDIA reports over the next two years.

Colette Kress

Interim CEO and CFO

If you have listened to an Nvidia earnings call in the past 13 years, you have heard Colette Kress. She joined as CFO in September 2013, when Nvidia was, in her words, the smallest company she had ever worked for. She has since run the finances of a business that now books more revenue in a single quarter than it once did in years.

Before Nvidia she spent eight years at Texas Instruments, 13 years at Microsoft, including four as CFO of its Server and Tools division, and three years as a senior finance executive at Cisco. She holds a finance degree from the University of Arizona and an MBA from Southern Methodist University.

Kress is not just the person who reports the numbers. The Associated Press identified her as a key architect of Nvidia’s financing deals with CoreWeave and Anthropic. She already sits at the centre of the customer-financing machine investors are most worried about. She keeps the CFO role while the board searches for a permanent CEO.

Ian Buck

Head of hyperscale and high-performance computing

If Kress is the continuity, Buck is the engineering soul of modern Nvidia. He joined in 2004 and helped build CUDA, the software layer that turned graphics chips into general-purpose computers. Twenty years on, CUDA is the reason a rival chip has to beat Nvidia not just on hardware but on the cost of moving every line of code.

Today Buck runs the business that sells to the cloud giants and AI labs at the heart of Nvidia’s US$89 billion-a-quarter data-centre business. Reuters has named him a credible internal candidate for the top job. Keeping him, and giving him the final say on the product roadmap, is the single most important retention decision the board will make.

Overseeing both is new chair Harvey Jones, a director since Nvidia’s founding year of 1993 and a former CEO of chip-design software company Synopsys. The board is running a search with an outside firm and will consider internal and external candidates.

We are not selling into the panic, and we are not adding yet either. Two things anchor our Hold.

First, the pipeline. The committed supply and customer contracts above give unusually good visibility for the next two years.

Second, the price. At around US$160, where the shares traded after the announcement, Nvidia sits on about 10 times calendar 2027 earnings, based on Bloomberg consensus EPS of US$15.91. That is a valuation you would normally pay for a mature industrial business, not the world’s leading AI chipmaker.

Illustrative only. EPS figures are analyst forecasts (consensus per Bloomberg; the US$13.00 downside case is our own stress test), not guarantees, and actual results may differ materially. Apple’s forward P/E fell from about 17 times to a trough near 9 times in 2013 after Steve Jobs’s departure, before recovering to about 15 times once execution was clear. Past performance is not a reliable indicator of future performance.

What would make us add? A clean result on 18 November, Buck staying with clear authority, and no adverse finding from the investigation. Tick two of those three and the risk-reward shifts firmly in our favour.

We need to be honest about this part. Under Huang, Nvidia was priced as the purest growth story in markets, and the share price behaved like it, rising more than 1,700% in the four years to October’s record high.

That chapter is likely over. Without its founder, we expect Nvidia to drop down the ranking as a “growth” company and be treated as a more mature business: a highly profitable, cash-generative market leader, but one valued on earnings rather than on vision. In practice, that means a lower ceiling on the multiple and less chance of the significant, sustained share price upside long-term holders have enjoyed.

That is not a reason to sell a business trading on about 10 times earnings. It is a reason to reset expectations, size the position accordingly and let the next six months show us who is really running the roadmap.