The bulls of Nvidia (NVDA) remain steadfast in their support for the stock as the chipmaker approaches its earnings announcement scheduled for next week.
While DeepSeek’s recent insights have sparked some debate regarding AI expenditures, many analysts continue to express optimism about Nvidia. However, they do note potential concerns regarding a first quarter forecast that could fall short of high market expectations.
“Short-term dynamics are changing (from January to July) … whereas long-term (starting October) our analysis indicates an incredibly promising outlook,” noted Loop Capital analyst Ananda Baruah in a recent client memo.
Baruah has maintained a price target of $175 for Nvidia, which implies a potential upside of 25% from its current levels.
He added, “From a broader perspective, we believe that the Street’s projections for the next two to three years remain insufficient, as our engagements with both clients and Nvidia’s build ecosystem suggest … Nvidia’s GPU shipments could reach 10 to 12 million, as hyperscalers aim to increase their non-CPU compute share to over 50% in the coming years (up from around 10% currently). It’s important to remember … for Nvidia, the narrative revolves around accelerated computing combined with Gen AI, presenting them with two $1.0 trillion market opportunities on their horizon, both of which are just beginning.”
In premarket trading, Nvidia stocks are up by less than 1%, currently priced at $140.
At close: February 18 at 4:00:00 PM EST
Despite a 23% rebound in Nvidia shares from early February lows, opinions on the company’s fundamentals have grown more mixed.
Podcast: How Bill Gates views Nvidia
Evercore analyst Mark Lipacis stated in a recent note that there are three key factors contributing to the cautious sentiment: 1) DeepSeek’s assessment of a reduced overall demand for AI, 2) a shift in AI compute cycles from Nvidia GPUs to ASICs [custom chips], and 3) delays regarding Blackwell chips.
In late January, China-based DeepSeek shocked markets with the introduction of RI, its AI model which reportedly delivers performance akin to ChatGPT at a significantly lower cost. RI’s base model was developed for approximately $5.6 million, contrasting sharply with the hundreds of millions spent by American firms like OpenAI and Anthropic.
This revelation quickly raised concerns that US corporations might be overspending on AI infrastructure, which prominently includes Nvidia chips.
“The prevailing belief throughout last year was that only a select few companies could train exceptional models,” remarked Snowflake (SNOW) CEO Sridhar Ramaswamy during an interview on Yahoo Finance’s Opening Bid podcast. “But what DeepSeek has accomplished recently is upending that notion by demonstrating that they can train a model for just $6 million.”
