Nvidia executives told analysts in a series of July meetings in California that the company expects its revenue growth to remain on an upward track, supported by demand for artificial intelligence infrastructure that is spreading beyond the largest AI labs and hyperscale cloud providers.
The message from Chief Executive Jensen Huang and Chief Financial Officer Colette Kress was that Nvidia is no longer operating simply as a supplier of graphics processing units. Management described the company as a full-stack AI platform provider, with revenue opportunities extending across chips, server racks, networking systems, central processing units, software frameworks, enterprise AI tools and government-backed AI infrastructure.
Morgan Stanley kept an “overweight” rating on Nvidia and maintained a $288 price target after the meetings. The bank projected Nvidia’s revenue could range from $215.9 billion to $783.9 billion between 2026 and 2029, depending on the pace of AI computing adoption, data-center spending, software growth and demand from newer markets such as sovereign AI and enterprise AI.
The meetings reinforced Wall Street’s view that Nvidia remains central to the global AI buildout. The company’s sales are already expanding at a scale rarely seen in the semiconductor industry, with quarterly revenue moving toward the $100 billion level as large technology companies, AI model developers, cloud platforms and government-linked buyers race to secure computing capacity.
At the same time, executives acknowledged that the next stage of growth will be shaped not only by demand, but also by limits in memory supply, advanced packaging, power infrastructure and data-center construction. Those constraints are expected to remain important for several years.
Nvidia shifts from chip supplier to AI infrastructure company
Nvidia’s management said the company’s transition from selling individual GPUs to providing integrated AI infrastructure has opened larger and more durable revenue channels.
Instead of focusing only on chip shipments, Nvidia is now selling systems that combine computing, networking, software and rack-scale infrastructure. That broader approach gives the company more influence over how AI data centers are designed and deployed.
The strategy also reduces the role of smaller rivals that compete only in one portion of the hardware stack. By supplying complete platforms, Nvidia can shape the server rack, the interconnect layer, the networking architecture and the software environment that customers use to train and run AI models.
This shift is important because AI data centers are becoming more complex. Large language models and other advanced AI systems require thousands of chips to work together efficiently. The performance of the entire system now matters as much as the performance of any single processor.
Customers are increasingly evaluating total token-generation costs, deployment speed, reliability and power efficiency rather than simply comparing chips on headline unit prices. Nvidia told analysts that this broader calculation supports demand for its complete systems, even when some customers also use custom AI chips.
Demand is spreading beyond top AI labs
Nvidia’s current demand mix remains heavily tied to the biggest buyers of AI infrastructure, but executives said the customer base is becoming more diverse.
According to management, roughly 20% of current demand comes from AI laboratories. Cloud providers account for about 50%. The remaining demand comes from newer groups, including industrial companies, Neocloud operators, enterprise users and sovereign AI programs supported by governments.
That mix matters because it suggests Nvidia’s growth is not limited to a small number of frontier AI model developers. The company is increasingly targeting businesses that want to run AI inside their own operations, as well as governments that want domestic AI infrastructure for security, language, research and public-sector uses.
Sovereign AI has become a major theme for Nvidia. Countries are seeking local computing capacity to reduce dependence on foreign cloud infrastructure and to support AI systems trained on local languages and data. Nvidia has positioned its hardware and software as a foundation for those national AI systems.
Enterprise demand is another major focus. Many companies want AI models that can operate within their own data environments, rather than sending sensitive information to external platforms. Nvidia says this shift creates an opening for its enterprise AI stack, which connects hardware, networking, software tools and model frameworks.
Asics and GPUs may coexist
One of the more notable points from the meetings involved a major client that had previously relied heavily on application-specific integrated circuits, known as ASICs.
According to the analyst report, that customer has nearly doubled its use of Nvidia computing power to close to 50%. The shift suggests that custom AI chips and Nvidia GPUs may coexist rather than fully replace one another.
ASICs can be efficient for specific workloads, particularly when a company has enough scale to design and deploy its own chips. But Nvidia’s GPUs remain attractive because of their flexibility, software ecosystem, availability across multiple use cases and ability to support fast-changing AI models.
For customers, the issue is no longer only the price of a chip. They are comparing the total cost of generating AI outputs, the time needed to deploy systems, the availability of software support and the ability to scale. Nvidia’s management indicated that this wider assessment continues to favor its integrated systems in many deployments.
That argument is central to Nvidia’s defense against custom chip competition. Large technology companies are developing internal AI chips to lower costs and secure supply. But Nvidia’s position remains strong because its platform is widely used, supported by mature software and tied into high-speed networking technology.
Vera Rubin Ultra remains on schedule
Nvidia also addressed reports that its Vera Rubin Ultra system could be delayed until 2028. Management said the product remains scheduled to ship in 2027.
The company plans to revise some rack designs, but executives said the core technology roadmap remains intact. Nvidia will maintain its 800-volt architecture and optical interconnect technology, both of which are intended to support higher-density AI systems.
The Vera Rubin platform is expected to be a major step in Nvidia’s next-generation AI infrastructure. As AI models grow larger and more computationally intensive, customers need systems that can move data quickly between chips while managing heat, power and reliability.
Rack-scale design is becoming more important because AI performance depends on how well thousands of components work together. Power delivery, cooling, memory bandwidth and networking can all become bottlenecks. Nvidia’s focus on full-system design gives it more control over these elements.
Management’s confirmation of the 2027 schedule was intended to reassure the market that Nvidia’s product roadmap remains on pace. For a company valued largely on future AI infrastructure growth, timing is critical.
Memory and power remain major bottlenecks
While Nvidia sees demand continuing to expand, executives said supply constraints will limit how quickly AI infrastructure can grow.
The company pointed to memory and power supply as key issues. High-bandwidth memory is essential for AI accelerators, and advanced AI servers require large amounts of power. Data-center sites also need electrical grid connections, cooling systems and physical space.
Nvidia described these limitations as supply issues rather than a sign of weaker end-market demand. In other words, customers want more computing capacity than the industry can currently deliver.
The company is exploring ways to improve delivery efficiency, including adjustments in DRAM allocation, network balancing and possible SRAM-based system designs. These efforts are aimed at reducing bottlenecks and improving the performance of large AI systems.
The constraints extend across the semiconductor supply chain. Taiwan Semiconductor Manufacturing Co. is working to increase advanced packaging capacity, with output targeted to reach 140,000 wafers per month by late 2026. Advanced packaging is necessary for many AI chips because it allows processors and memory to be connected in ways that improve speed and efficiency.
The capacity expansion shows how much strain AI demand is placing on the chip industry. Even with aggressive factory spending, supply delays in high-performance computing components are expected to persist.
Power demand is reshaping the data-center market
Electricity availability is becoming one of the most important factors in AI infrastructure growth.
Worldwide data-center power use is projected to reach 565 terawatt-hours in 2026, a 26% increase from the prior year. That level of demand would place added pressure on electrical grids that are already dealing with industrial electrification, renewable energy integration and rising consumption from digital services.
Industry analyst Wang said dedicated computing operations could consume nearly 68 terawatt-hours in the United States alone this year. The estimate highlights how quickly AI and high-performance computing facilities are becoming major electricity users.
For many data-center operators, the challenge is not only buying chips. It is securing access to power, land, cooling and grid connections. In some regions, new data-center projects face long waiting periods because utilities cannot provide enough capacity quickly.
This has changed the competitive dynamics around existing computing sites. Facilities that already have power agreements, substations and cooling infrastructure have become more valuable. Cloud providers and AI infrastructure companies are seeking locations that can support immediate deployments, rather than waiting years for new grid capacity.
Shared digital mining sites have become part of this discussion because many were built in locations with access to large amounts of power. Some operators of those facilities have received buyout interest from cloud providers that want to repurpose capacity for AI workloads.
Impact on digital asset infrastructure
The competition for power and computing hardware is also affecting digital asset markets, particularly operations tied to mining and decentralized computing.
Mining sites that once focused mainly on proof-of-work digital assets are increasingly being evaluated as potential AI data-center locations. The reason is simple: they often already have power access, land and cooling equipment. Those assets can be difficult and slow to recreate.
This does not mean every mining site can easily become an AI facility. AI workloads often require different networking, redundancy, security and hardware layouts. Still, the value of power access has risen sharply, and that has created new strategic options for operators with strong infrastructure.
The trend may reduce the availability of computing resources for some distributed ledger activities, especially if more facilities shift toward AI workloads. It could also increase interest in decentralized compute networks that offer shared processing capacity for smaller software developers.
Traditional server racks face long order backlogs, and smaller companies may struggle to obtain high-end hardware directly. Shared physical networks could benefit if they provide usable computing capacity at a time when centralized AI infrastructure remains constrained.
For traders in digital asset markets, the key issue is whether power scarcity and chip shortages redirect capital and infrastructure away from mining and toward AI. That shift could influence hosting costs, mining economics and demand for tokens linked to decentralized computing.
Enterprise AI and open-source models
Nvidia also emphasized its work on open-source enterprise models, including Nemotron.
The company said these tools allow businesses to train and deploy AI systems within their own controlled data environments. That approach is designed for companies that want AI capabilities but cannot expose proprietary or regulated data to outside systems.
Nvidia described Nemotron and related software projects as part of a broader enterprise AI stack. The goal is to connect hardware, networking and software frameworks so businesses can build AI applications more easily.
This strategy could help Nvidia expand beyond the largest cloud platforms. Enterprise AI adoption is still in an early phase, and many companies are testing how to use AI in customer service, software development, manufacturing, drug discovery, finance and logistics.
If those deployments move from pilot programs to broader production use, demand for AI infrastructure could widen further. Nvidia’s management appears to be positioning the company to capture that spending through both hardware and software.
Capital returns and valuation scenarios
Morgan Stanley said Nvidia intends to return at least half of its cash flow through share buybacks and dividends. The strategy is aimed at broadening the company’s appeal to long-term value-focused funds, not only growth-oriented technology buyers.
The bank’s base case projects earnings per share of $4.61 in 2026, $8.96 in 2027, $13.08 in 2028 and $17.63 in 2029. Gross margins are expected to range from 71.3% to 74.4% during that period.
Under an optimistic scenario, Morgan Stanley said Nvidia’s price target could rise to $330 per share if data-center and software revenue continue to grow strongly. In a downside case, the stock could fall toward $160 if data-center demand slows, ASIC substitution increases or policy restrictions limit exports.
Export controls remain a meaningful risk. Nvidia has faced restrictions on some chip sales to China, and future policy decisions could affect revenue from international markets. The company has developed modified products for certain regions, but regulatory uncertainty remains part of the outlook.
Quantitative indicators in the report showed active institutional ownership near 50.9%, suggesting many growth-focused funds already have large positions in the stock. That could limit the pool of new buyers in the near term, although strong earnings growth and cash returns may continue to support demand for shares.
Morgan Stanley concluded that Nvidia remains one of the most attractive risk-reward profiles in the semiconductor sector, provided the company continues to execute on its diversified AI platform roadmap.
The broader market takeaway
Nvidia’s July meetings underscored two connected themes: AI demand remains strong, but the infrastructure needed to support it is becoming harder to secure.
The company is benefiting from a rare combination of software advantage, hardware leadership, networking depth and customer urgency. Its move from GPU supplier to full AI platform provider has strengthened its position across the data-center market.
But the same growth that supports Nvidia’s outlook is also stretching the global supply chain. Memory, packaging, power and grid access are emerging as critical constraints. Those bottlenecks are likely to shape competition among cloud providers, enterprises, governments, mining site operators and decentralized compute networks.
For traders, the central question is whether Nvidia can keep converting AI demand into delivered systems while maintaining high margins and staying ahead of custom chip competition. For the wider technology sector, the bigger issue may be whether the world can build enough power and data-center infrastructure to keep pace with AI’s rapid expansion.
Explore how AI is transforming trading with Toobit’s AI copy trading guide to align with Nvidia’s infrastructure revolution.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
