Wall Street analysts are breathing a bit of a sigh of relief over Nvidia after comments from Big Tech heavyweights on earnings calls Wednesday evening reassured investors that artificial intelligence spending remains on track. Chinese startup DeepSeek’s new AI reasoning model , R1, fueled a massive sell-off in technology stocks on Monday and alarmed investors primarily because it rivals OpenAI’s o1 model and was made at a supposedly fraction of the cost that U.S. hypercalers are spending on AI. DeepSeek also built its model using Nvidia chips with reduced capability, leading to panic that major tech companies will spend less on Nvidia products and drag down the AI darling’s stock — threatening the AI-fueled gains of the broader market. Wall Street’s mood on AI has picked up after quarterly earnings calls from Meta Platforms and Microsoft reiterated both companies’ huge AI spending plans, however. Analysts, including from Goldman Sachs and Bank of America, reiterated bullish sentiment on Nvidia in the last 24 hours. In its quarterly report, Meta maintained its 2025 capex guidance of between $60 billion and $65 billion — driven by the company’s increased investments in its generative AI efforts and its core businesses — which is consistent with its original guidance given in January 2024. Meta’s chief financial officer said during the earnings call that Meta is “planning to significantly ramp up deployment of GPUs in 2025.” “Despite concerns of cloud capex slowdown from DeepSeek’s notable AI efficiency pushes, we highlight strong AI capex commentary from META and MSFT on Wednesday evening,” wrote Bank of America analyst Vivek Arya in a Thursday note. “Importantly, both companies continue to see their capex increase through CY25/26.” “Meanwhile, the mix of spend continues to skew towards more servers (CPUs, GPUs, ASICs, etc.), with both companies looking to further expand their GPU and custom silicon adoption/purchases,” Arya added. Goldman analyst Toshiya Hari said in a Thursday note that “while Meta’s stated intent to use more of its in-house silicon for their ‘unique workloads’ will undoubtedly feed the merchant vs. custom debate, we believe the reiteration of near-term capex budgets by Microsoft and Meta as well as their long-term thinking on scaling laws will be perceived positively by the market, particularly given the stock’s recent underperformance following DeepSeek’s R1 announcement.” NVDA 1D mountain Nvidia, 1 day Microsoft’s CEO also said on a call with analysts that he expects to see “exponentially” more AI demand as the technology becomes more efficient. The company has committed $80 billion for AI-enabled data centers in its current fiscal quarter, and suggested its third and fourth fiscal quarter capex spending will remain similar to its second fiscal quarter. Nvidia shares lost 17% during the sell-off , and shares traded lower on Thursday on a Bloomberg report that officials under President Donald Trump’s administration are discussing tighter restrictions on the sale of Nvidia chips to China. Shares of several other megacap technology stocks have also seen steep declines this week. Nvidia shares were higher briefly to start Thursday’s session, but were under pressure again as the trading day continued. Hari, who backs his buy rating on Nvidia, said that the launch of DeepSeek’s reasoning models is a “positive development for Meta and the industry, particularly as Meta hopes to integrate what DeepSeek has achieved into its own AI systems.” “The industry has historically benefited from an open source approach and the recent models from DeepSeek, if anything, reinforces their [Meta’s] view,” the analyst added. Hari noted that Microsoft remained positive on its AI plans even after DeepSeek’s success, as he Microsoft management cited that the “reduction in compute cost should make the technology more ubiquitous and, in turn, drive higher adoption in the cloud but also on the edge.” Bank of America’s Arya noted that Meta’s and Microsoft’s AI capex spending remains on track with strong ramps ahead in GPU and ASIC computer processors. “We highlight the race for AGI (artificial general intelligence) and the top-of-the-line frontier models (i.e. OpenAI/Meta models) continue regardless of the derivative or ‘distilled’ models from the likes of DeepSeek, and AI compute/networking remain important enablers of the AI golden age,” Arya wrote in a Thursday note to clients. Jeffries agrees with analyst Blayne Curtis noting, “We believe the news is highly misunderstood and the trend towards ‘reasoning’ based models and more inference implies compute demand should actually accelerate.”
