For NVIDIA, 2026 is All About Holding Ground
By Sanjana Gupta ·
NVIDIA is heading into 2026 with a level of revenue momentum few companies can rival. Yet, the central question is whether speed alone is enough to sustain its lead. After several years as the clearest beneficiary of the AI boom, the company now faces scrutiny over hyperscaler behaviour, power constraints, circular-economy risks, and whether its dominance rests on a durable structure or fortunate timing.
The stakes could hardly be higher. NVIDIA has become the market’s primary proxy for AI investment, influencing both data centre capital allocation and equity sentiment. With its valuation stretched and alternatives beginning to emerge, 2026 is shaping up as a year in which NVIDIA’s narrative may shift from acceleration to examination.
By conventional measures, revenue growth remains exceptional. Michael Ashley Schulman, chief investment officer at Running Point Capital, points to the company’s latest results as evidence. “NVIDIA just printed $57 billion of quarterly revenue in fiscal Q3 2026, a significant jump of 22% from the prior quarter and 62% from the year before, with $51.2 billion from data centres.”
Management has stressed that demand is not the bottleneck. “Blackwell demand is not the problem; supply and deployment are,” Schulmgan adds.
At a market capitalisation approaching historic levels, investors are no longer valuing NVIDIA as a cyclical hardware winner, but as a long-term infrastructure franchise.
That long-term view introduces a new set of variables. Chief among them is the pace at which the world’s largest cloud providers can meaningfully reduce their reliance on NVIDIA silicon, and what that would mean for the company’s grip on the AI value chain.
NVIDIA no longer just sells chips; it now delivers full-stack solutions spanning racks, networking, systems, and software. As a result, the focus is moving beyond quarterly beats. By 2026, investors are likely to assess NVIDIA not only on the pace of its growth, but on its ability to sustain market dominance through 2027 and 2028.
“When evaluating a $4.5 trillion company, you have to look out further than 12 months,” Schulman says.
Hyperscaler Question Will Not Go Away
One of the most significant risks to NVIDIA’s medium-term outlook comes from its largest customers. Cloud providers want greater control over their infrastructure, and custom silicon remains a central ambition.
Holger Mueller, vice president and principal analyst at Constellation Research, frames the issue bluntly. “It will be driven by the ability of the cloud vendors to provide their own silicon,” he says. “Google is good with TPUs; we have to see how well AWS does, and Microsoft, the kingmaker for NVIDIA, failed with Gen2 of their AI chip and delayed it.”
For now, those delays work in NVIDIA’s favour. “If there is no news there, it is good news for NVIDIA,” Mueller says.
Over the longer term, however, hyperscaler efforts still pose pressure. Schulman describes the risk not as sudden displacement, but gradual erosion. “The hard to price risk is that customers with deep pockets unbundle the stack, use their own accelerators where they can, and buy fewer NVIDIA chips at the margin when supply normalises,” he says.
Brady Wang, associate director at Counterpoint Research, expects NVIDIA to remain ahead, but with tighter competition. “NVIDIA is likely to remain the leading AI accelerator vendor through 2026, despite rising competition from hyperscaler ASICs and AMD,” Wang says.
He further notes that NVIDIA’s edge now lies less in raw GPU performance and more in “system-level integration, software lock-in, and supply-chain control.”
The Software Moat
CUDA has long been presented as NVIDIA’s strongest moat, but analysts disagree on how durable it will be in the era of generative AI.
“For NVIDIA’s yellow brick road, adoption of CUDA and the broader software stack matter most to lock customers and pricing power into a developer ecosystem,” Schulman says. He points to NVIDIA’s expanding portfolio of inference software and microservices as critical to maintaining margins.
Mueller is more sceptical. “It has been overrated,” he says. “In the age of generative AI, where I can translate Cobol workloads to e.g. Java, I can also migrate out of CUDA easily.”
That divide highlights a deeper uncertainty. If AI workloads become more abstracted and portable, software lock-in weakens. If performance, tooling, and reliability still matter most at scale, NVIDIA’s ecosystem remains a barrier.
Counterpoint’s Wang believes the truth lies between those views. While ASICs may take share in inference, “NVIDIA’s dominance in large-scale training should largely persist,” he says. That suggests pressure without collapse.
Power & the AI Bubble Debate
While markets often fixate on interest rates and growth forecasts, Mueller highlights a more practical concern. “Power concerns are bigger than interest rates and growth,” he says. As data centres expand, access to electricity, cooling and grid upgrades is increasingly the factor that determines how quickly AI infrastructure can scale.
In this cycle, macro factors look different. Interest rates and GDP growth still matter, but they are secondary to physical limits. As Mueller put it, power availability now outweighs traditional macro drivers.
That concern is widely shared across the industry. Schulman warned of what he calls “data centre capex indigestion”, where customers pause spending not because they lack demand, but because they lack electricity, cooling, and a clear return on investment.
Investor sentiment reflects those frictions. NVIDIA remains the cleanest proxy for AI capital spending, but the narrative has grown noisier. “There is learning that NVIDIA may sit on a bubble,” Mueller says.
“There will be even more [volatility] this year, especially as investors see other ascending investment options,” Schulman says, noting that “AI won’t be the only investment story for 2026.” He points to mergers, IPOs, and other sectors competing for attention.
Supply chains introduce an additional layer of risk. Wang says NVIDIA’s shipments by 2026 are more likely to be constrained by “a systemic supply-chain bottleneck” than by process technology alone. Advanced packaging, high-bandwidth memory, substrates, and liquid cooling all create pressure points. NVIDIA’s scale offers some insulation, but costs are likely to rise.
What 2026 Really Tests
The central question for 2026 is not whether AI demand fades. Most analysts agree it remains strong, albeit more disciplined and driven by ROI considerations.
The real test is structural. Can NVIDIA defend margins as hyperscalers buy selectively? Can it manage power and supply limits while expanding systems sales? And can it avoid becoming a company defined by a single cycle?
Mueller offers a caution rooted in history. “The biggest risk is that NVIDIA will be back at where it was six years ago, growth was supposed to come from crypto and automotive,” he says. “Now there is robotics additionally as a growth driver.”
Robotics could offer a new leg of growth, though Mueller tempers expectations. He calls it “the backup for NVIDIA” but notes “it will not be as soon and as conform as a load like the LLMs have been.”
For now, NVIDIA still benefits from AI’s underpenetration. Schulman compares the moment to earlier platform shifts. “We continue to see AI weaving its way into every industry and facet of business,” he says.