Jensen Huang did not come to the Goldman Sachs Communacopia + Technology conference to hedge. The Nvidia founder and CEO delivered a characteristically unambiguous message to investors and analysts on Thursday: the company that has become synonymous with the artificial intelligence boom expects its revenue to grow by roughly 70 percent in the next fiscal year, a figure that would push annual sales toward $680 billion. It is a stunning number by any measure, and one that raises as many questions about competitive threats, market saturation, and the sustainability of AI spending as it does about Nvidia’s continued dominance.
Huang’s appearance came just weeks after Nvidia reported yet another record-breaking quarter, and he used the platform to reaffirm guidance first issued during that earnings call. The core claim is simple: the company sees a clear path to 70 percent year-over-year revenue growth in fiscal 2028. But the reasoning behind that confidence reveals a great deal about how Nvidia views its own role in the AI ecosystem, how it defines its products, and how it intends to fend off a growing list of challengers that includes the world’s largest cloud providers and a new generation of well-funded chip startups.
Seventy Percent Growth: The Numbers Behind Nvidia’s Boldest Forecast Yet
Analysts currently expect Nvidia to close its current fiscal year with approximately $400 billion in revenue. A 70 percent increase would bring that figure to roughly $680 billion in the following year. To put that in context, $680 billion would be more than the combined annual revenue of every major semiconductor company on the planet just a few years ago. It would also represent one of the largest single-year revenue jumps ever achieved by a technology company at this scale.
Huang did not offer this guidance as a hopeful target. He presented it as a certainty. “I think we could grow 70 percent year over year. We’re confident about that,” he said on Thursday, repeating a claim he first made during Nvidia’s earnings call the previous month. The confidence stems from a vantage point that Huang argues is unique in the industry: Nvidia claims to have visibility into virtually every major AI infrastructure project under development anywhere in the world.
How Nvidia Claims to See the Future of AI Infrastructure
Huang described a monitoring operation that tracks data center construction at a granular level. “We’re tracking every single gigawatt of land, power, shell around the world. Literally everything on the planet,” he said, using the industry term “shell” to refer to the physical structure of a data center before it is equipped with computing hardware. This global surveillance of AI build-out is possible, he argued, because Nvidia sits at the center of an unusually dense network of partners who report back on their plans and purchases.
“How many neoclouds are reporting back to us? How many OEMs are reporting back to us? How many clouds are reporting back to us? How many AI-native companies are reporting back to us? We’re working with everybody, and so we kind of know where everything is,” Huang said. The implication is that Nvidia’s visibility across the AI supply chain gives it a forecasting advantage that rivals cannot easily replicate. The company sees orders, construction timelines, and capacity plans months or even years before they translate into publicly reported spending.
This intelligence network is not merely passive. Nvidia runs what Huang described as every major AI model in existence, including models from Anthropic, OpenAI, Google, and numerous open-weight projects. “We are a foundational platform of the AI ecosystem, foundational platform of the AI industry,” he said. That platform position means that when a new model is developed or a new AI application gains traction, Nvidia typically knows about it first because the company is providing the infrastructure to train and run it.
What Is a GPU Now? Nvidia Redefines Its Core Product
A significant portion of Huang’s presentation was devoted to correcting what he sees as a persistent misunderstanding about what Nvidia actually sells. The company invented the GPU, and for many years those chips were sold primarily to gamers at prices measured in hundreds of dollars. That perception, Huang argued, is dangerously outdated.
“Most people think Nvidia builds a chip. I mean, you need airplanes to ship what we build,” he said. He then described what Nvidia now considers a single “GPU”: a fully integrated system that combines 36 Grace CPUs with 72 Blackwell GPUs, connected via NVLink, containing 2 million individual parts, consuming 250,000 kilowatts of power, and priced at $8.5 million. “That’s one GPU, and we ship thousands of them,” Huang said.
The redefinition is strategic. By framing its product as an entire computing system rather than a chip, Nvidia raises the competitive bar. Competitors like Cerebras, which went public recently, and startups such as Etched, which has reached a $5 billion valuation on the strength of its AI chip designs, are not simply trying to build a better processor. They are trying to replicate an entire integrated hardware-software ecosystem that Nvidia has spent years optimizing. The system that Huang described is not something a customer can assemble from off-the-shelf components. It requires Nvidia’s proprietary NVLink interconnect, its CUDA software platform, and the deep integration work that turns individual chips into a functional supercomputer.
Huang also revealed that orders for this specific system, the GB200 NVL72, are growing at 27 percent month over month. That pace of growth, if sustained, would compound into enormous demand within a single fiscal year and helps explain the 70 percent revenue forecast.
The Competitive Landscape: Why Nvidia Is Not Worried About Challengers
The question that hangs over every bullish Nvidia statement is whether the company’s dominance can survive the determined efforts of its largest customers and a wave of well-funded startups. Amazon, Microsoft, and Google are all developing their own AI chips. Anthropic and OpenAI are building custom silicon. Cerebras, now a public company, claims architectural advantages for certain workloads. Etched has attracted significant investment and customer interest. The list of companies trying to displace Nvidia is long and growing.
Huang acknowledged the competitive pressure but dismissed the notion that it poses a near-term threat to Nvidia’s trajectory. His argument rests on two pillars. First, Nvidia’s integration into the AI ecosystem is so deep that no single competitor can replicate the full stack. Second, the sheer scale of demand means there is room for multiple suppliers, but Nvidia will capture the largest share because it delivers the most comprehensive solution.
He also addressed the so-called “circular deals” criticism, in which Nvidia invests in companies that then use the capital to purchase Nvidia products. This practice contributed to the downfall of previous generation infrastructure suppliers like Lucent Technologies during the dot-com boom, when the appearance of real demand was later revealed to be a cycle of vendor-funded purchases. Huang acknowledged the comparison with characteristic bluntness.
“Well, it’s not circular because we put a little bit of money in, and a lot of money comes back,” he said, adding with a joke, “I look at the spreadsheet, we put in $1 and $100 comes back in. Is that circular? If that is, let’s do more of that.”
Beyond the humor, Huang insisted that Nvidia’s investment process is disciplined. Before the company writes a check to any startup, it verifies that the recipient has real contracts generating genuine revenue from third-party customers. He claimed to have seen $100 billion worth of such contracts, adding, “I’m not taking any risks. … I need a sure thing.”
How Nvidia’s Partner Network Creates an Intelligence Advantage
The density of Nvidia’s partnerships is itself a competitive moat. Every neocloud provider, every OEM, every major cloud platform, and every AI-native company that reports its plans to Nvidia is simultaneously providing the company with data it can use to forecast demand, allocate supply, and identify emerging trends before they become visible to the broader market. A startup building a new AI model must choose its hardware platform early. If it chooses Nvidia, Nvidia knows about it. If it chooses a competitor, Nvidia still knows about it, because the company tracks the broader ecosystem.
This information advantage compounds over time. Nvidia can adjust its production plans, allocate scarce manufacturing capacity at TSMC, and prioritize the highest-value opportunities. Competitors are flying blind by comparison, reacting to demand signals that Nvidia already saw months ago. It is a structural advantage that no amount of chip design talent can easily overcome.
The Risks: Disruption, Efficiency, and the Inevitable Maturation of AI
No technology company has ever remained dominant forever. The golden rule of the tech industry is that all big things get disrupted eventually, and Nvidia is not immune to that principle. Huang himself acknowledged that much of the current AI boom is fueled by AI-native startups that raise enormous sums of venture capital and spend most of that cash on Nvidia’s infrastructure. As those startups mature, they will face pressure to become more efficient in how they use compute resources and tokens, which could reduce the rate at which they need to buy new hardware.
There is also the question of whether the hyperscalers developing their own chips will eventually succeed in reducing their dependence on Nvidia. Google’s TPU, Amazon’s Trainium and Inferentia, and Microsoft’s Maia chips are all real products with real deployments. None of them yet match Nvidia’s breadth of software support and ecosystem compatibility, but they do not need to match it across the board. They only need to be good enough for the specific workloads that matter most to their parent companies, and they enjoy the enormous advantage of being designed in-house for in-house use, with no margin required and no need to support a broad external customer base.
Custom silicon efforts take years to bear fruit, however, and Nvidia is not standing still. The Blackwell architecture represents a generational leap in performance, and future architectures are already in development. Huang has shown that he understands the threat of commoditization and is determined to stay ahead by continuously redefining what Nvidia delivers. If a GPU is now an $8.5 million system, the next generation may be something even more ambitious.
The AI industry is also likely to undergo a shift from training to inference as deployed models become more widespread. Inference workloads are less computationally intensive than training, and they are more amenable to optimization and specialization. Competitors that focus on inference efficiency, as some of the startups are doing, could find a growing market even if they cannot compete in training. Nvidia’s response to this shift will be critical. The company has already introduced inference-optimized products, but the economics of inference are fundamentally different from training, and margins may be harder to sustain.
What Nvidia’s Forecast Means for the AI Industry and the Broader Market
A $680 billion revenue run rate for Nvidia next year would have profound implications for the entire technology sector. It would mean that AI infrastructure spending is not slowing down but accelerating, and that the companies building that infrastructure are willing to pay enormous sums for Nvidia’s integrated systems. The figure would also reshape the competitive dynamics of the semiconductor industry, as Nvidia alone would account for a share of global chip revenue that has no modern precedent.
For investors, the 70 percent growth guidance provides a clear benchmark. Nvidia’s market capitalization has already been volatile as the market debates whether the AI boom is real, sustainable, or overhyped. Huang’s message is unambiguous: the boom is real, it is sustainable, and Nvidia is best positioned to capture its value. The question for skeptics is whether the company’s unparalleled visibility into global AI infrastructure gives it a genuinely superior forecasting ability or whether it creates an illusion of certainty that could be shattered by a shift in the technological or competitive landscape.
For the startups and cloud providers that make up Nvidia’s customer base, the guidance is a double-edged sword. It confirms that Nvidia expects them to keep spending, and spending heavily. But it also puts pressure on them to demonstrate that their own revenue growth can justify the infrastructure costs. If AI-native companies cannot generate revenue at a pace that matches their spending on Nvidia hardware, the circularity critique that Huang dismissed so casually will become harder to ignore.
Huang’s confidence is grounded in data that no other company has, but data about the present and near future is not the same as certainty about the long term. The AI industry is still young, its use cases are still being discovered, and its economic structure is still being built. Nvidia has navigated previous technology cycles with remarkable skill, and the company’s current position is stronger than any it has held before. But the history of technology is a history of surprise, and the only sure prediction is that the future will not look exactly like the present. For now, Nvidia sees another year of extraordinary growth ahead, and it is telling the world to get ready.
