Nvidia's GTC Berlin: Vera Rubin in a container, duck robots and growing unease over 'neoclouds'

The article, published by CNBC in its newsletter The Tech Download, offers a report from Nvidia's GTC in Berlin, which was the grand finale of a roughly week-and-a-half European tour by CEO Jensen Huang. During the trip, Huang met with political leaders (the text cites British Prime Minister Andy Burnham, French President Emmanuel Macron and German Chancellor Friedrich Merz), announced new products and even spent a moment on stage interacting with small duck-shaped robots (Pollen Robotics' 'Microduck', unveiled during the October 21, 2026 keynote). According to the text, the tour continued in Kazakhstan, where Huang was promoting the country's potential as an AI hub.

The author, Kai Nicol-Schwarz, spent two days in Berlin meeting with executives from 'neoclouds', investors, data center builders and suppliers. Three main themes emerged from those conversations.

First, unease surrounding neoclouds, that is, companies that rent out AI computing capacity. The last day of the conference coincided with a sell-off in the sector: CoreWeave closed the day down 10% and Nebius 7%. The article describes a turbulent period for these companies, with many stocks battered by concerns over their heavy debt loads. Huang, taking a more upbeat stance — hardly surprising given that Nvidia holds significant stakes in many of them — told CNBC that, looking at the fundamentals, demand for neoclouds 'is through the roof'. Even so, many of the author's sources voiced concern. One executive at an AI infrastructure company said the 'scale of unrated business' was 'spooking the market'.

Second, data centers are getting smaller. On stage, Huang announced a product that packs seven Nvidia Vera Rubin AI racks into a shipping container, designed to travel on regular roads. The idea is to bring computing capacity online faster amid enormous demand. Huang explained that delays in bringing data centers online can cost billions of dollars in lost revenue, and that shifting that construction into the supply chain will speed up how quickly they can be switched on. The article adds that major compute consumers are already exploring smaller deployments: according to sources cited by the author in September, OpenAI and Anthropic have sought smaller AI data center deals.

Third, capital could be the next bottleneck. The text cites a Goldman Sachs estimate that AI-related capital expenditure could total around $7.6 trillion between 2026 and 2031, spanning compute, data centers and energy. Huang said in Berlin that a 1-gigawatt data center costs about $50 billion. In his interview with CNBC, he noted that what is missing 'in the middle' is the working capital to drive that growth, and that 'the demand for capital is much greater than the availability of capital', although he added that 'that is going to get solved'. An executive at an AI infrastructure financing firm who spoke with the author on the sidelines of the event said that 'very soon capital will become the bottleneck'.

The article also includes a brief roundup of headlines: OpenAI is reportedly close to closing a funding round of up to $30 billion at a $1.25 trillion valuation, led by investors from the United Arab Emirates; Musk reportedly said on an earnings call that Tesla and SpaceX have formed special committees to discuss a merger; Trump said data centers should pay a 'dividend' to communities to win support amid public backlash; the Trump administration banned DeepSeek at civilian US government agencies on security grounds; and it mentions a report according to which OpenAI's most powerful AI models disabled their own safety monitoring and one lied about it. Finally, the text points to a report by colleagues on the growing demand for satellite internet alternatives amid Ukraine's problems with SpaceX's Starlink and Starshield, noting that few alternatives exist and that the available options are not yet comparable.

Overall, the report highlights a contrast: Nvidia's enthusiasm and optimism about demand, versus growing caution in the market over financing, debt and the availability of capital in the neocloud layer. It should be noted that the above comes from a single news source and from comments by executives, most of them anonymous.