Nvidia reports results for the first fiscal quarter of 2027 on May 20, entering as the dominant AI infrastructure vendor with a giant share of datacenter revenue, net cash of over $50 billion, and above-average margins.
Still, Goldman Sachs warns that "the bar for the stock's performance is relatively high" as the market is already pricing in a very ambitious future growth scenario.
What exactly is Goldman Sachs changing in its estimates
Goldman Sachs $GSanalyst James Schneider raised his earnings per share estimates for Nvidia $NVDA by an average of 12% and maintains a "Buy" recommendation with a $250 price target. This target implies roughly 20% potential upside from current levels, while the consensus of 42 analysts estimates an average target price of around $274.
For the first quarter, Goldman expects revenue of $80.05 billion, about $2 billion above the consensus of about $78.3 billion, and adjusted earnings per share of $1.86 - about 7% above the market estimate of $1.74. Schneider also raised estimates for calendar years 2026 and 2027 so that they are now about 14% and 34% above the Wall Street average, respectively, showing that Goldman believes in a longer duration AI investment cycle.
For the second quarter, Goldman is modeling revenue of about $87.68 billion, about 3% above consensus, anticipating continued growth in demand for the Blackwell and Grace Blackwell platforms. The bank is basing this on comments from TSMC and SK hynix, among others, about strong demand and aggressive capex plans from large US cloud players.
Datacenters as the key and the "trillion-dollar" outlook
Nvidia's $NVDA today stands virtually entirely on datacenters. In the last fiscal year, datacenter revenue accounted for over 90% of the company's total revenue, marking a leapfrog transformation from the company's erstwhile "GPU for games" to a backbone infrastructure for generative AI. Internal presentations and market commentary speak of a "trillion-dollar" outlook for cumulative datacenter revenue in the coming years, built on continued investments by hyperscalers, large model houses and sovereign AI programs of the states.
Importantly, this long-term outlook doesn't yet include the full spectrum of new products - notably the Ruby Ultra generation, the new Vera CPU racks, and specialized inferencing configurations that should expand the addressable market toward a broader range of workloads. This suggests that Nvidia can still grow not only "in depth" (more computing power per customer) but also "in breadth" (new system types and new customers).
Margins: superior but sensitive to cost and mix
Goldman Sachs estimates that Nvidia should maintain a gross margin of around 75% in 2026, an extremely high level for the semiconductor business. In the short term, however, margins are being squeezed by more expensive HBM memories and the new cost structure of platforms that integrate multiple components - such as Blackwell and the future Ruby.
Meanwhile, the company is operating from a position of strength: it enters the results with a net cash position of around $51.1 billion, a full-year gross margin of around 71.1% and a 12-month EBIT margin of around 60.4%. In addition, Blackwell generation analyses show an approximate tenfold improvement in cost efficiency per unit of power versus the previous generation, leaving Nvidia room to absorb higher nominal component costs through a better power/watt/dollar ratio.
New opportunities: CPU seagulls and the emergence of AI agents
In addition to traditional GPU systems, Nvidia plans to start shipping pure CPU-based racks in the second half of 2026 to target specific workload types - from AI agents to traditional cloud workloads where GPUs don't make economic sense. This opens up another revenue tier for the company and moves it closer to a full-fledged system platform where customers can cover most computing needs within a single ecosystem.
Goldman will be particularly watchful during the earnings conference call for comments on the adoption of AI agents and the impact on CPU-based systems. This is because in addition to hyperscalers, customers outside of this segment are also gaining traction - for example, OpenAI, Anthropic and sovereign state AI programs that buy large blocks of infrastructure and may create new demand and contracting models over time.
But the competition is not sleeping: custom ASIC solutions from large cloud companies, in-house chips, or alternative AI accelerator vendors are topics that investors will be watching closely in Q&A. The key question is whether Nvidia will maintain its technology lead and software lock-in (CUDA, libraries, ecosystem), or whether the first more visible signs of competition will start to appear in the numbers.
Awards
In his note, Schneider stresses that he expects a "beat-and-raise" quarter, but also reminds that the bar for the stock to significantly outperform the market after earnings is very high. Since the March low, the title has added roughly 28% and is currently trading around $210-215 per share. Options markets are pricing in an implied move of at least around 10% in either direction, reflecting a combination of high expectations and uncertainty around the next phase of the AI cycle.
At a P/E of around 43-44 (TTM), Nvidia is trading lower than in some past "euphoric" periods - the 10-year average P/E is around 50-55, the five-year even higher. From one perspective, this makes the stock look cheaper than in the past; from another, it's still a very high multiple that requires the company to continue to deliver high revenue and earnings growth and maintain its dominant position in AI infrastructure.
On the recommendation side, there is extraordinary consensus: of the 42 analysts following the title, 40 recommend buying the stock, one holds and one sells, creating a very strong "Strong Buy" consensus. The consensus target price of around $274 implies around 30% potential, but the market has already priced in much of the AI story - hence Goldman's talk of a "high bar" for further upside revaluation.