Goldman Sachs: Nvidia (NVDA) Reiterate buy
Reiterate buy on favorable risk/reward; Raise 12-month price target to $1,100
We reiterate our Buy recommendation on NVDA (also on the Conviction List) with an updated 12-month price target of $1,100 (previously $1,000) by increasing our non-GAAP EPS estimates (excluding SBC) for FY2025-27, on average, by 8% to reflect intra-quarterly data points indicative of continued robust demand for AI servers and improvement in supply. Despite NVDA's YTD outperformance (+86% vs. SOX/SPX +15%/+9%, respectively), we see positive revisions to EPS (note our updated non-GAAP EPS estimate for FY2026 [or CY2025] of $40.78 is 35% above the Street consensus) driving another rally in the stock, especially with NVDA trading at a 35x or just a 36% premium to our coverage universe vs. its median premium for the last 3 years of 160%. See our full note for more details, including an updated version of our Bull/Bear framework.
Updated Thoughts on Nvidia's Key Business Segments
AI Remains Key Driver: In line with the past several quarters, we expect Nvidia's FY1Q (April) results and FY2Q (July) outlook for its Data Center segment are driven by continued strength in demand for Computing and Networking related to artificial intelligence. Reflecting universally positive feedback from the AI ecosystem (i.e. cloud service providers or hyperscalers, ODM, other component suppliers) and several new product launches (e.g. H200 in FY2Q and Spectrum-X and Blackwell in FY2H) , we now forecast 10%/17%/5% sequential growth in Nvidia's Data Center business in FY2Q/3Q/4Q (previously 10%/15%/0% quarter-over-quarter), respectively. Notable intra-quarter data points supporting the idea that AI spending will likely continue beyond 2024 include; 1) TSMC reiterated its constructive short- and long-term outlook for HPC/AI. Specifically, in the near term, management expects server AI processor revenue (including GPUs, accelerators, and CPUs used for Training and Inference) to grow >2x year-over-year, and represent a low-single-digit percentage (%) of the total of revenue in 2024. In the medium to long term, management continues to expect HPC/AI revenue to grow at a CAGR of ~50% and represent >20% of total revenue by 2028 (Exhibit 2 and our benchmark note), 2) Tier 1 hypermarketers in the US stated/implied that AI-related capital investments would likely increase in 2025 from an already elevated base in 2024 (more details below), 3 ) Some hypermarketers and enterprise software companies cited early signs of AI monetization, 4) AMD positively revised its Data Center GPU revenue outlook for 2024 to $4.0bn from $3.5bn previously, 5) Supermicro reported strong revenue growth, although still limited by supply, and record backorders during its FY3Q earnings call driven by continued strength in demand for AI servers (Exhibit 3), and 6) our colleague, Bruce Lu, based on his conversations in the industry, updated his TSMC annual CoWoS capacity assumptions for 2024/25/26 to 319k/600k/720k from 304k/441k/522k previously (Exhibit 8) as highlighted here. From a competitive standpoint, while AMD has made considerable progress on its Data Center GPU roadmap and several hyperscalers have also announced new custom compute programs or upgrades to existing programs, in short, we believe Nvidia will remain the standard. of the industry for the foreseeable future given a) its competitive advantage that encompasses its hardware and software capabilities, as well as the installed base and ecosystem it has built over several decades, and b) the pace at which it is and will innovate in the coming years (remember that Nvidia will introduce new products/platforms annually going forward).
Supportive Capex Feedback from US-based Hypermarketers: Forward-looking commentary on Gen AI-related capex was consistently positive across all major US-based hypermarketers. Specifically, 1) Alphabet highlighted strong traction for its Generation AI-related services and guided capex for the remaining quarters in 2024 to be at or above what it spent in 1Q24 (~$12bn) driven by investments in technical infrastructure; 2) Microsoft highlighted that AI contributed 7 percentage points to Azure growth in FY3Q (March), compared to 6 percentage points in FY2Q (December) and 3 percentage points in FY1Q (September) with demand for AI at short term in Azure greater than its current capacity. On capex, Microsoft guided for a material sequential increase in the June quarter, and talked about FY25 being higher than last year as the company looks to meet further growth in demand for cloud and AI products; 3) Meta increased its 2024 capex guidance to $35-40bn (vs. $30-37bn previously) and shared its expectation that 2025 capex will increase year-over-year driven by investments in support of its AI R&D efforts ; 4) Amazon expects a significant year-over-year increase in capex in 2024 driven by increased infrastructure capex to support growth in AWS, including Generation AI. In summary, we increased our bottom-up global cloud capex forecast for CY2024/25 by 16%/22% and now forecast year-over-year growth of 46%/11% (vs. 26%/5% previously) , respectively (Exhibit 6). While we expect the trajectory of Gen AI capex beyond 2024 to remain a source of intense debate among investors, the combination of a) constructive feedback on hypermarket capex, b) early signs of monetization of AI and c) several new product launches at Nvidia, increase our conviction in the short term.
Supply Comments: In line with the last several quarters, we expect investors to scrutinize management's comments on Nvidia and its partners' ability to meet demand, especially with the launch of Blackwell on the way. Despite comments from Nvidia management at GTC about the supply chain being better prepared for the launch of Blackwell compared to that of Hopper, however, we expect demand to outstrip supply, at least in the first few quarters. of the GB200 ramp-up. Recall that TSMC management, in the recent 1Q24 earnings call, stated that its CoWoS capacity remains tight despite plans to increase capacity by >2x year-over-year (i.e. 319k wafers in 2024 vs. 144k wafers in 2023, according to our analyst, Bruce Lu).
Mixed Signals in Gaming: Intra-quarter data points related to the Gaming industry have been mixed, with positive reports from a) Logitech (which reported an 8% YoY increase in Gaming revenue on a constant currency basis) and b) Kinsus (which spoke of much stronger-than-expected demand for Gaming GPU ABF substrates), offset by more cautious data points from c) insert board (AIB) makers in Taiwan, including Micro -Star International (MSI) and ASRock (which reported a 4% sequential revenue decline in 1Q24 vs. typical seasonality of -1% quarter-over-quarter) and d) AMD (which reported a 33% quarterly decline in revenue Gaming in 1Q24 and led to a similar sequential decline in 2Q24). For Nvidia specifically, we continue to expect the ongoing transition/upgrade to the Ada Lovelace architecture, along with the company's push with its Generation AI-related services (e.g. Nvidia ACE), to drive double-digit growth ( %) in Gaming segment revenues in FY2025 and 2026.
