Nvidia expands control in AI infrastructure, and this time: from the cloud to financing
Nvidia continues to expand beyond chip sales, and recent events point to the building of a complete system around AI infrastructure, including chips, cloud, data centers, computing services, and even financing mechanisms.
Earlier today, news reached us that may shed light on Nvidia's plans:
IBM and Together AI signed a multi-year agreement worth $240 million, with operations expected to begin in Q1 2027.
Under the agreement, IBM will establish a large dedicated computing cluster in its cloud for AI data analysis; the infrastructure will be based on NVIDIA HGX B300 systems and NVIDIA Spectrum-X Ethernet network.
In fact, this is the first large cluster based on HGX B300 and Spectrum-X to be established in IBM Cloud.
According to Nvidia, the infrastructure is expected to provide 30 times higher AI throughput compared to previous generations.
Together AI will use the infrastructure to provide business customers with data analysis services for open-source AI models, with the goal of significantly increasing processing capability while reducing processing cost per unit of information.
Morgan Stanley reiterated its "overweight" recommendation for Nvidia stock, with a price target of $288.
Analyst Joseph Moore views Nvidia's move to raise external capital for AI infrastructure positively.
Yesterday, Nvidia announced the possibility of raising over $500 billion through six institutional partners, for potential investments in AI factories and building a financing platform for computing.
The main point for Morgan Stanley is that the financing relies heavily on third-party capital, and not solely on Nvidia's balance sheet.
According to him, this reduces concerns about circular transactions, where companies in the sector finance each other and purchase services or equipment from each other, and may also encourage additional investments from governments and cloud providers.
Wells Fargo also reiterated its "overweight" recommendation for Nvidia, with a price target of $315.
Analyst Aaron Rakers also refers to the partnership worth over $500 billion with leading global investment firms.
The important implication for Wells Fargo is that Nvidia is starting to build a much broader role for itself than that of a chip manufacturer.
The company is trying to build a model where it is involved in the planning, financing, and establishment of "AI factories," computing centers built specifically for AI system workloads.
Nvidia also emphasizes the co-design capability, meaning the adaptation of chips, network, computing, and the entire infrastructure for the AI factory.
This creates a system where customers gain access to large amounts of computing power, and Nvidia can be involved in a larger part of the value chain.
QumulusAI signed a computing services agreement with global trading firm DRW, under which it will provide DRW with a dedicated NVIDIA Blackwell B300 cluster.
The initial agreement is for one year, with an option to renew three more times, meaning a possible period of up to four years.
The infrastructure will be located in QumulusAI's data centers in the US.
DRW engages in quantitative trading, primarily in commodities and derivatives, and is considered a particularly demanding consumer of computing resources.
The company also operates activities related to the computing resources market, including the Compute Exchange, as well as indices and futures contracts tracking GPU prices under the Silicon Data brand.
Since early August, QumulusAI has signed agreements totaling over $246 million.
When considering all this news, it's easy to see the big picture:
The three events connect to the same trend:
IBM is using Nvidia's infrastructure to establish a large-scale AI cloud.
Together AI is increasing its use of computing for open-source models.
DRW consumes advanced computing power for research and quantitative trading.
QumulusAI is turning GPUs into a service that can be leased to customers for years.
Nvidia is simultaneously advancing into the financing sector, with a platform that can raise over $500 billion for AI infrastructure.
In other words, demand for Nvidia no longer comes only from companies buying chips, but from a broader ecosystem of cloud, data centers, computing services, AI models, quantitative trading, and infrastructure financing.
And this is precisely the interesting part: Nvidia is trying to transform the demand for AI computing from a one-time hardware sale into a complete system that generates continuous use of its infrastructure for years.
Jensen Huang has made a tremendous move here; meanwhile, analysts are backing him, and on the ground, it is certainly noticeable that the compass of demand continues to point north.