Nvidia CFO Colette Kress has become interim chief executive following Jensen Huang’s November 12 resignation, according to CryptoPotato. The transition puts her in charge of Nvidia data center strategy ahead of the company’s November 17 quarterly results.
The company’s infrastructure business spans AI chips, enterprise software, server partners and systems deployed by cloud operators. Its leadership change comes as operators commit billions of dollars to expansion while identifying power and capacity as constraints on growth.
Nvidia’s Leadership Transition Comes Before Quarterly Results
What changed at the top of Nvidia?
Huang stepped down as CEO and a board member, according to CryptoPotato, with Kress taking over on an interim basis. She has spent more than a decade at Nvidia, and directors are considering internal and external candidates for the permanent job.
Read More: Jensen Huang Exits Nvidia After 33 Years; Shares Fall Over 10% Premarket
The succession places an experienced finance executive at the center of a business built around Huang’s unusually centralized leadership. AI-chip demand, the CUDA ecosystem and relationships across the technology industry remain important strengths during the transition.
How did the transition unfold?
According to BeInCrypto, directors began an investigation on November 8 into whether a foreign intelligence operation had targeted Huang. His resignation followed on November 12, and Elon Musk publicly supported a petition for his reinstatement on November 13.
The board engaged Wachtell Lipton, with former acting US Attorney Joon H. Kim leading the investigation. The November 17 earnings date now puts a scheduled company disclosure close behind the leadership upheaval.
Who Controls Nvidia’s Next CEO Appointment?
What authority does the board have?
Under the bylaws described in the governance coverage, a majority of directors in office controls executive appointments. Following Huang’s departure, that threshold is six of the 10 remaining directors. Shareholders have no direct vote on the CEO.
Musk has no Nvidia board seat. His public backing gives the reinstatement campaign a prominent supporter, while the appointment decision remains with directors handling the investigation and permanent succession process.
How much influence does Huang retain?
Huang controls about 3.58% of voting power, based on Nvidia’s 2026 proxy. BlackRock and Vanguard each hold more than 7%. An internal post urging Huang’s return drew more than 1,600 comments, illustrating the employee response to his departure.
Kress’s interim appointment brings institutional knowledge to the immediate transition. The board’s consideration of candidates both inside and outside Nvidia leaves the longer-term leadership choice open as the company approaches its quarterly announcement.
Nvidia Data Center Products Extend Beyond AI Chips
What did Nvidia introduce for enterprises in 2021?
At Computex 2021, Nvidia introduced the cloud-hosted Base Command platform and AI-optimized x86 servers through its certified systems program. TechRadar’s coverage of the launch named NetApp as a Base Command collaborator and identified Asus, Dell Technologies, Gigabyte, Hewlett Packard Enterprise and Lenovo among the server vendors.
The announcement connected hardware to enterprise software: the systems were certified for AI Enterprise on VMware vSphere, Omniverse Enterprise and Red Hat OpenShift. Support for Cloudera data engineering and machine-learning tools was also described.
How did Base Command connect teams to larger systems?
Base Command gave enterprise teams cloud access to Nvidia DGX SuperPOD computing, using NetApp data management tools. Its design supported multiple researchers, data scientists and teams working on AI projects simultaneously.
The early-access subscription cost $90,000 per month in the 2021 announcement. That historical offering shows how Nvidia paired large computing systems with a development platform and partner software, extending its role beyond supplying individual processors.
Power and Cooling Shape Nvidia Infrastructure Expansion
Where are operators encountering constraints?
DataCenterFeed’s industry coverage identifies power and capacity as growth constraints cited by Equinix and Digital Realty. Equinix booked $533 million in annualized gross bookings; Digital Realty recorded $855 million in bookings at 100% share and a $2.5 billion backlog.
Oracle’s infrastructure offering shows another route for Nvidia systems to reach customers. Oracle made Vera Rubin NVL72 racks and Vera CPUs generally available through OCI and outlined a multi-gigawatt Supercluster plan.
How are server suppliers adapting their hardware?
Supermicro’s factory-integrated 48U rack contains 256 Nvidia Vera CPUs, 22,528 Arm cores and up to 400 TB of memory. It uses direct liquid cooling and is designed to run agentic-AI sandboxes alongside GPU systems, extending liquid-cooling requirements to CPU racks.
Crusoe’s sixth-generation Spark module takes a modular approach, packing up to 3.5 MW of compute into a factory-built unit. The company describes closed-loop liquid cooling that uses no water and compatibility with Nvidia GB300 NVL72 and Vera Rubin NVL72 racks.
AI Infrastructure Spending Reaches Tens of Billions
What scale has CoreWeave reached?
CoreWeave posted $3.66 billion in third-quarter revenue and a $132.5 billion backlog, alongside more than 1.7 GW of active power. It raised its 2026 capital expenditure range to $37–39 billion, illustrating the scale of construction and equipment commitments surrounding AI computing.
The company says it will avoid issuing parent-level unsecured debt through 2027 at current yields. Instead, it intends to use contract-backed loans and customer prepayments, tying its financing approach more closely to contracted business.
What investment plans has Meta outlined?
Meta guided to $135–145 billion in 2026 capital expenditure. It also filed for about $35 billion in underwritten common and mandatory convertible preferred offerings and outlined a planned $60 billion at-the-market program. Nvidia says it will invest $3 billion in the deal.
Meta says the first phase of its Prometheus supercluster in New Albany, Ohio, is complete and trained its Muse models. The site is headed toward roughly 1 million GPUs. The company also says software and batteries help it smooth power swings of 30–40%, making power management part of its infrastructure expansion.
DGX Spark Brings Nvidia Computing to Smaller Sites
What does the compact Nvidia system offer?
A May 2026 account from Trak.in describes DGX Spark as a compact Grace Blackwell system, originally introduced as Project DIGITS. Nvidia says it delivers up to 1 petaflop of AI performance, with 128GB of unified memory and support for models with up to 200 billion parameters locally.
The desktop-sized system targets developers, researchers and enterprises. It gives the Nvidia infrastructure story a smaller-scale dimension alongside cloud deployments and large campuses, with local model work among the uses described for the machine.
What is Span proposing for residential computing?
Startup Span is developing compact XFRA server nodes for residential areas. The company claims thousands of these nodes could collectively provide computing power comparable to a traditional 100-megawatt AI data center.
The proposal centers on using locations that already have electrical service. Nvidia executive Marc Spieler described using existing powered locations as making sense when large power allocations are difficult to obtain. DGX Spark and XFRA are separate products discussed within this distributed-computing approach.
Governance and Financing Risks Accompany the Buildout
Why does the succession matter for infrastructure execution?
Huang’s central role in Nvidia makes the appointment process significant for a business spanning chip suppliers, cloud companies and data center operators. The investigation adds a governance issue to that transition, while large infrastructure investments face scrutiny over whether they will generate sufficient returns.
CoreWeave’s stated financing approach highlights the importance of borrowing costs and customer commitments to continued expansion. For distributed computing, Span’s capacity comparison remains a company claim; the proposal depends on assembling thousands of smaller nodes around existing residential electrical infrastructure.
What does the separate Singapore server case allege?
In a separate legal matter, The Defiant says Singapore filed 41 charges involving Shi Bingzhe and two companies. The money-laundering allegations concern purchases of 917 servers in a probe involving shipments to Malaysia.
The case adds legal scrutiny around Nvidia-linked server purchases to the industry backdrop. The allegations concern the named defendants and server transactions, separate from the reported board investigation surrounding Huang.
November Results and 2027 Manufacturing Plans Come Next
Which events are on the immediate calendar?
Nvidia’s November 17 quarterly results are the next scheduled opportunity for company disclosures under Kress’s interim leadership. The appointment search continues alongside that reporting calendar, with directors considering both internal and external candidates.
The House Foreign Affairs Committee has asked Huang to appear on December 1 and warned of a subpoena. That request creates a separate Washington milestone as directors handle the leadership process and investigation.
What infrastructure production is planned for 2027?
Crusoe plans to begin production of its sixth-generation Spark modules at its Brighton, Colorado, plant in 2027. Their support for Nvidia GB300 NVL72 and Vera Rubin NVL72 racks places the manufacturing plan within the wider supplier ecosystem serving AI campuses.
The factory-built modules combine computing capacity and cooling in a single unit. They are distinct from Nvidia’s DGX Spark desktop system: Crusoe’s product addresses infrastructure deployment, while the compact Nvidia machine is designed for local AI development and research.
FAQ
What specifications does Nvidia DGX Spark offer?
Nvidia’s stated specifications include 128GB of unified memory and up to 1 petaflop of AI performance. The Grace Blackwell system supports local models with up to 200 billion parameters and was originally introduced as Project DIGITS.
Which server manufacturers appeared in Nvidia’s 2021 certified systems announcement?
The named vendors included Asus, Dell Technologies, Gigabyte, Hewlett Packard Enterprise and Lenovo. Their AI-optimized x86 servers formed part of Nvidia’s certified systems program announced at Computex 2021.
What was Base Command’s monthly early-access price in 2021?
The 2021 early-access offering was priced at $90,000 per month. It provided cloud-hosted access to DGX SuperPOD computing with NetApp data management tools for collaborative AI development.
Who can appoint Nvidia’s chief executive under its described bylaws?
Executive appointments require a majority of the directors in office, described after Huang’s departure as six of 10 remaining directors. A shareholder petition does not provide a direct vote on the CEO position.
When are Nvidia’s next quarterly results scheduled?
The next quarterly announcement is scheduled for November 17, 2026. Kress enters that event as interim CEO following her appointment from the CFO role.
When has the House committee asked Huang to appear?
The House Foreign Affairs Committee requested an appearance on December 1. It has also warned of a subpoena, making this a requested appearance rather than confirmed testimony.
Where will Crusoe manufacture its planned Spark modules?
Crusoe plans to start production at its Brighton, Colorado, plant in 2027. Its sixth-generation modules are designed for Nvidia GB300 NVL72 and Vera Rubin NVL72 racks and provide up to 3.5 MW of compute per unit.

