Piper Sandler analyst Harsh V. Kumar reiterated an Overweight rating on $NVIDIA (NVDA.US)$ with a $1,050 price target, his top large-cap pick.
During his recent AI Discovery Bus Tour, Kumar engaged directly with Nvidia's management team, which allowed him to delve deeper into the company's current operations and future prospects.
Despite being in the market for nearly two years, the demand for Nvidia's Hopper GPU remains robust, with the supply needing to catch up, as per the analyst.
He said the supply-demand imbalance will likely continue throughout the year, potentially offering a slight margin boost until the introduction of the Blackwell GPU.
As per Kumar, the launch of the Blackwell GPU later this year will likely mirror the supply and demand challenges faced by Hopper.
Customers are reluctant to switch their orders from Hopper to Blackwell, fearing even longer wait times due to anticipated supply constraints.
Nvidia's management addressed concerns about the power requirements for upcoming data centers, who assured that their accelerated computing solutions are significantly more power-efficient than traditional computing, Kumar said.
Plans are in place to optimize power needs for future data centers, including leveraging "trapped power" in specific regions and considering new data centers in countries that can meet these significant power demands.
Pricing discussions centered on the Total Cost of Ownership (TCO) benefits provided by Nvidia to its customers, indicating no change in pricing philosophy with the introduction of the Blackwell architecture, he said.
While the company aims to maintain competitive pricing, gross margins for the Blackwell architecture will likely be slightly below the corporate average as it ramps up.
Revenue opportunities from data center products sold to sovereign entities are currently being realized by Nvidia, with expectations for this demand to significantly contribute to the company's revenue in the near term, as per Kumar.
The move towards liquid-cooled systems, especially with the GB200, is another area of innovation, with major Cloud Service Providers committing to adopting the GB200, primarily for inference applications.
In terms of valuation, Nvidia's shares are trading at a premium compared to its peers, reflecting Nvidia's continued leadership and innovation in the GPU market.
Kumar projected first-quarter revenue and EPS of $24.01 billion and $5.41.
After hosting an industry expert on the data center and AI supply chain, Mizuho analyst Vijay Rakesh noted Nvidia's leading AI GPU with a strong software moat. He also said Nvidia is leading AI accelerators, especially for training.
Nvidia continues to be the AI leader, with continued tightness on H100 and H200. Blackwell B100 is set to launch in 2024 and ramp up in 2025, as per the analyst.
However, he also flagged some potential challenges with new high-power Nvidia systems with power-to-rack and cooling solutions, a major data center challenge.
Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.
派珀·桑德勒分析师哈什·库马尔重申了增持评级 $英伟达 (NVDA.US)$ 目标股价为1,050美元,这是他的首选大盘股。
在最近的人工智能探索巴士之旅中,库马尔直接与英伟达的管理团队进行了接触,这使他能够更深入地研究公司当前的运营和未来前景。
分析师称,尽管上市了将近两年,但对Nvidia的Hopper GPU的需求仍然强劲,供应需要迎头赶上。
他说,供需失衡可能会持续一整年,在推出Blackwell GPU之前,可能会略微提高利润率。
根据库马尔的说法,今年晚些时候布莱克韦尔GPU的推出可能会反映出Hopper面临的供需挑战。
由于预期的供应限制,客户不愿将订单从Hopper转移到Blackwell,他们担心等待时间会更长。
库马尔说,Nvidia的管理层解决了对即将到来的数据中心电力需求的担忧,他们保证他们的加速计算解决方案比传统计算更节能。
已经制定了优化未来数据中心的电力需求的计划,包括利用特定地区的 “滞留电力”,以及考虑在能够满足这些重大电力需求的国家设立新的数据中心。
他说,定价讨论集中在Nvidia为其客户提供的总拥有成本(TCO)优势上,这表明随着Blackwell架构的引入,定价理念没有改变。
尽管该公司的目标是保持有竞争力的定价,但随着价格的提高,Blackwell架构的毛利率可能会略低于公司平均水平。
根据库马尔的说法,Nvidia目前正在实现向主权实体出售数据中心产品的收入机会,预计这一需求将在短期内为公司的收入做出重大贡献。
向液冷系统的转变,尤其是 GB200,是另一个创新领域,主要的云服务提供商承诺采用 GB200,主要用于推理应用。
在估值方面,英伟达的股票交易价格与同行相比处于溢价,这反映了Nvidia在GPU市场的持续领导地位和创新。
库马尔预计第一季度收入和每股收益为240.1亿美元和5.41美元。
瑞穗分析师维杰·拉克什在接待了一位关于数据中心和人工智能供应链的行业专家后,指出了英伟达领先的人工智能GPU具有强大的软件护城河。他还表示,Nvidia是领先的人工智能加速器,尤其是在训练方面。
英伟达仍然是人工智能的领导者,H100和H200一直处于紧张状态。根据分析师的说法,布莱克韦尔B100定于2024年推出,并于2025年上市。
但是,他还指出了采用机架供电和冷却解决方案的新型大功率 Nvidia 系统面临的一些潜在挑战,这是数据中心面临的主要挑战。
免责声明:此内容部分是在人工智能工具的帮助下制作的,并由Benzinga的编辑审阅和发布。