Nvidia stock has another eye-catching Wall Street number attached to it: $345. BNP Paribas has raised its Nvidia stock price target from $285 to $345 while keeping an Outperform rating, arguing that the chip giant remains its preferred semiconductor play as artificial intelligence spending shifts into a new and potentially more demanding phase. With Nvidia shares closing at $233.95 on October 2, the target implied roughly 47% upside from that level.
The call lands at an extraordinary moment. Nvidia is no longer merely riding enthusiasm for generative AI. Its latest reported quarter produced $96.2 billion of revenue, more than double the year-earlier figure, while Data Center revenue reached $89.0 billion, up 117%. Meanwhile, BNP Paribas estimates that spending across the broader AI infrastructure ecosystem is approaching $1 trillion annually.
Yet the most interesting part of BNP Paribas‘ argument isn’t simply that AI spending will continue. It is the idea that the next stage of AI could make Nvidia’s competitive position more difficult to attack.
That distinction could determine whether a $345 Nvidia stock price target eventually looks aggressive—or surprisingly conservative.
The $345 Price Target Is Really a Bet
BNP Paribas‘ new target represents a substantial increase from its previous $285 estimate. The brokerage continues to rate Nvidia Outperform and expects the company to retain at least three-quarters of the AI compute market in dollar terms despite competition from AMD and custom accelerators developed by large technology companies.
Related articles
That is an important assumption because the bear case around Nvidia has gradually changed.
Few investors seriously question whether AI infrastructure demand is large. The tougher question is whether Nvidia can keep capturing an enormous portion of the economics as hyperscalers develop their own silicon and competing chipmakers attack the accelerator market.
BNP Paribas‘ answer is essentially that investors may be looking at Nvidia too narrowly.
The company sells GPUs, but its competitive position increasingly includes CPUs, networking, NVLink interconnect technology and, critically, the CUDA software ecosystem. As AI infrastructure becomes more complex, coordinating those pieces can become almost as important as maximizing the theoretical performance of an individual accelerator.
And that is where the next stage of AI becomes particularly interesting.
Agentic AI Could Change the Economics
AI infrastructure was initially dominated by one giant challenge: training increasingly large models.
BNP Paribas estimates that inference could account for roughly 60% of compute demand in 2026, while agentic AI is changing the balance between CPUs, accelerators and other parts of the server architecture. Instead of simply generating a response, AI agents may perform multi-step reasoning, call software tools, retrieve information and coordinate multiple tasks.
Those workloads create a different infrastructure problem.
Performance increasingly depends on moving enormous quantities of information between processors, memory and networking components efficiently. BNP Paribas believes Nvidia’s integrated approach—including CUDA and NVLink—could therefore become more valuable as workloads grow more complicated.
Nvidia’s Vera Rubin platform is central to that argument. Rather than functioning as a standalone GPU story, Rubin integrates GPUs, CPUs, networking and other components into a rack-scale architecture built for increasingly sophisticated AI workloads.
Nvidia said earlier in 2026 that Rubin could reduce inference token costs by as much as 10 times compared with Blackwell, and major cloud providers including AWS, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure were expected to deploy Vera Rubin-based instances.
That gives investors a clue about what BNP Paribas believes happens next: Nvidia may not merely defend its accelerator business. It could capture more value from every AI system surrounding those accelerators.
Nvidia’s Revenue Numbers Are Making the Bull Case Harder to Dismiss
Normally, a stock carrying enormous expectations eventually runs into the law of large numbers.
Nvidia has spent much of the AI boom appearing to ignore it.
For its fiscal second quarter ended July 26, 2026, Nvidia reported revenue of $96.22 billion, up 106% year over year and 18% sequentially. Data Center revenue reached $89.0 billion, rising 117% from the previous year. GAAP operating income climbed 124% to $63.73 billion, while diluted GAAP earnings per share rose 128% to $2.46.
Even more striking was profitability.
GAAP gross margin reached 75%, compared with 72.4% a year earlier. That matters because one of the central risks surrounding Nvidia is that competition and increasingly sophisticated hardware will eventually squeeze its extraordinary economics.
BNP Paribas isn’t forecasting that collapse. The brokerage expects Nvidia to maintain gross margins above 70% despite intensifying competition. Its $345 valuation uses 15 times estimated calendar-2028 earnings plus cash.
For bulls, that combination—triple-digit Data Center growth alongside margins above 70%—is the crucial evidence that Nvidia’s moat is not disappearing yet.
But another bottleneck is emerging, and Nvidia cannot solve it simply by designing a faster GPU.
The Biggest AI Constraint May No Longer Be Chips
AI companies want more computing power. The uncomfortable question is whether the physical world can deliver it quickly enough.
BNP Paribas says AI infrastructure spending is approaching $1 trillion per year, spread across chips, servers, data centers, electrical grids, cooling systems and digital networks. Chips and servers capture roughly half of that investment, but electricity and infrastructure constraints are becoming increasingly important.
Morgan Stanley has estimated that U.S. data-center developers could face a 34% net power shortfall through 2028, equivalent to about 32 gigawatts even after accounting for measures such as behind-the-meter generation and fuel cells.
That sounds dangerous for Nvidia. If a data center cannot secure electricity, installing another rack of AI processors becomes considerably harder.
Yet Morgan Stanley said Nvidia and Broadcom appear relatively insulated from the worsening power crunch and that the constraints do not currently put their 2027 forecasts at risk. Secondary suppliers—including memory, optics, analog and power-management components—could face greater inventory disruption if projects are delayed.
In other words, the AI boom may be moving from a shortage of compute to a shortage of infrastructure capable of supporting compute.
For Nvidia investors, that is both a warning and a sign of just how enormous the buildout has become.
Nvidia Stock Is Already Pricing In a Lot of Good News
There is one problem with an exceptional company: investors usually know it is exceptional.
Nvidia shares were trading around $237 on the morning of October 5 after closing at $233.95 in the previous session, putting the stock close to its 52-week high. The company had a market capitalization of roughly $5.6 trillion at that point.
That changes the risk equation.
At this scale, Nvidia doesn’t simply need AI to remain important. It needs hyperscaler capital spending, inference growth, networking demand and adoption of new architectures such as Rubin to translate into earnings powerful enough to justify one of the world’s largest equity valuations.
Competition cannot be ignored either. AMD continues to pursue the accelerator opportunity, while hyperscalers have strong financial incentives to design custom silicon for particular workloads.
The BNP Paribas thesis, however, is that greater specialization does not automatically destroy Nvidia’s advantage. More complex AI systems may actually increase the importance of software, networking and system-level integration—the areas where Nvidia has spent years building its ecosystem.
That is why the $345 target deserves more attention than the number alone suggests.
What Could Turn $345 From a Bold Call Into a Realistic One?
The Nvidia stock price target ultimately rests on a simple but enormous wager: AI infrastructure spending keeps scaling, and Nvidia continues capturing an outsized share of it.
Right now, the operating numbers support that thesis. Revenue has crossed $96 billion in a single quarter. Data Center revenue has more than doubled year over year. Gross margins remain around 75%. BNP Paribas expects Nvidia to retain at least 75% of the AI compute market in dollar terms, and the firm sees the company’s expanding hardware, software and networking portfolio as a defense against increasingly aggressive competitors.
The next tests will be more difficult.
Investors should watch Rubin adoption, inference economics, gross margins, hyperscaler capital expenditures and signs that power shortages are delaying data-center deployments. Any meaningful slowdown in those areas could expose how much optimism is embedded in Nvidia’s valuation.
But if agentic AI produces another surge in compute requirements, the story changes again. Nvidia would no longer be selling the dominant chip into the AI boom—it would increasingly be selling the architecture surrounding that boom.
That is the hidden message inside BNP Paribas‘ $345 call. The bank isn’t simply betting that Nvidia sells more GPUs. It is betting that the AI data center becomes more complicated—and that Nvidia makes money from more of the pieces every time it does.
For the stock, the next battle may therefore have little to do with whether AI demand survives. The far bigger question is how much of a trillion-dollar infrastructure race Nvidia can ultimately own.
Disclaimer
This article is for informational purposes only and does not constitute financial or investment advice. Readers should conduct their own research and, where appropriate, consult a qualified financial advisor before making investment decisions. This article was researched and drafted with the support of AI, then reviewed, fact-checked, and edited by the editorial team before publication.
