NVDA NOTE (Continued)
Bubble Concerns and Analyst Rebuttal
Skepticism about a possible AI bubble has risen, especially after multibillion-dollar deals between NVIDIA, OpenAI, and other leaders. The analyst response is that most current spending is backed by profitable tech firms with clear infrastructure needs and long-term revenue projections, unlike the vendor-financed dot-com bubble of the late 1990s. Cantor Fitzgerald and others argue that rapidly scaling AI demand, major partnerships (e.g., OpenAI’s $100 billion purchase plans from NVIDIA), and international expansion (like chip exports to UAE) substantiate the rally and differentiate this from classic speculative bubbles.
Nvidia (NASDAQ:NVDA) shows no signs of slowing its efforts to dominate the processing power behind the multi-trillion-dollar buildout of artificial intelligence infrastructure, prompting Cantor to reiterate its "Top Pick" Overweight rating and increase its price target to a Wall Street high of $300 from $240.
The next highest target was set by Melius Research, which increased its target to $275 from $240 earlier this week. Goldman Sachs recently increased its price target to $210 from $200. Last month, Bank of America increased its target to $235 from $220, and Barclays boosted its target to $240 from $200. In August, Wells Fargo hiked its target to $220 from $185.
During early market action on Thursday, Nvidia shares had increased nearly 2%, reaching a new intraday record high. Shares have doubled in value in the last six months.
The new target follows a series of investor meetings Cantor analysts had with Nvidia's top executives this week, including CEO Jensen Huang and CFO Colette Kress. They discussed details of the new partnership with OpenAI (OPENAI).
"The vision here is to remove margin stacking by server ODM's and CSP's, enabling the cost delta between NVDA and ASICs to only an average of 15% - truly a win-win for both and placing potentially further pressure on the ASIC market," said Cantor analysts, led by C.J. Muse, in a Thursday investor note. "The company continues to work at Extreme Co-Design at an annual cadence, enabling the optimization of the entire AI Infrastructure (no one wants to buy chips only, they want to deploy AI at scale), helped by a full-stack solution including CUDA-X."
Cantor now sees Nvidia hitting an $8 earnings per share in calendar year 2026 versus the consensus of $6.26 and $11 in 2027 compared to the $7.37 consensus.
The analysts also pointed out that token demand has skyrocketed in the last few months, which is increasing profit per token.
"OpenAI and other platforms are now generating 50-70% gross margins, leading to every stood up GPU now sold out," Muse said. "Customers are searching desperately for compute wherever they can find it. More specifically, in the last 12-16 weeks we have seen an inflection - likely tied to time-based reasoning explosion coupled with multimodal inputs, led by video where context length is super long."
Cantor is also convinced the rapid, wide-scale adoption of generative AI proves this is not a bubble.
"In the last 12 months, recommenders have adopted Generative AI," Muse noted. "Search has moved to Generative AI. Social Media to Generative AI. User-Generated content, Ad recommendation, everything has moved from classical machine learning to Generative AI. NVDA sees $2 Trillion in spending here from traditional compute to Generative AI alone. We are simply not in a bubble, rather the market is beginning to see how 'quality' AI can deliver meaningfully positive ROI's."
Cantor believes the AI infrastructure market will range from $3T to $4T by 2030. |
