Nvidia reports after the market close on May 20. Morgan Stanley’s analyst Joseph Moore published his preview note on May 18. And the number at the center of his argument is one that the rest of Wall Street is, in his view, significantly underestimating.
That number is $1 trillion. And whether you believe it or not determines almost everything about how you should think about Nvidia’s stock right now.
Morgan Stanley raises Nvidia ahead of May 20 earnings
Morgan Stanley raised its price target on Nvidia to $285 from $260 on May 18, reiterating an overweight rating and designating it the firm’s top semiconductor pick, according to Morgan Stanley. The new target is based on 22 times Morgan Stanley’s calendar year 2027 EPS estimate of $12.99. The multiple was deliberately lowered from the prior 26 times to reflect the firm’s view that Nvidia’s already dominant market share and elevated gross margins leave limited room for further multiple expansion in the near term.
Moore raised his April quarter revenue estimate to $79.264 billion from $78.25 billion and adjusted EPS to $1.72 from $1.69. For the July quarter, he lifted revenue to $87.88 billion from $84.837 billion and EPS to $2.01 from $1.93. Looking further out, Morgan Stanley now models FY27 revenue of $380.591 billion with EPS of $8.61, and FY28 revenue of $587.45 billion with EPS of $13.11.
The Nvidia $1 trillion data center argument and why consensus is too low
The core of Moore’s bull case is a revenue trajectory that the rest of Wall Street is not yet pricing. Morgan Stanley estimates Nvidia will generate approximately $1.07 trillion in cumulative data center revenue across calendar years 2025 through 2027.
Within that, the firm projects $884 billion of data center revenue in calendar years 2026 and 2027 alone. The Street consensus for that same two-year period sits at just $785 billion, a gap of nearly $100 billion.
More Nvidia:
- Nvidia is losing an industry that saved it from bankruptcy
- Nvidia CEO makes surprising admission on OpenAI and Anthropic
- Goldman Sachs just found a reason to like Nvidia stock again
Moore’s math works as follows. Calendar year 2025 included approximately $25 billion of legacy Hopper compute and another roughly $30 billion in networking, leaving approximately $155 billion in Blackwell and Rubin data center revenue for that year. That implies $845 billion across 2026 and 2027 before factoring in newer product categories, including Groq rack-scale inference, standalone CPUs, and ICMS, which Moore described as “substantial upside” on top of the core figure.
Moore expects the May 20 earnings call to be a catalyst for consensus to move toward Morgan Stanley’s estimates as Nvidia reaffirms visibility into those numbers, including a discussion of the supply constraints it is managing. He identified three specific bottlenecks: powered shell availability, leading-edge wafer capacity, and DRAM.
How Nvidia’s supply positioning protects its margin despite rising costs
A significant section of the Morgan Stanley note addresses cost inflation. As of Nvidia’s most recent 10-K, the company held $95 billion in purchase commitments and $21 billion in inventory. Moore argued that if those commitments reflect 75% gross margin cost of goods sold, they represent sufficient supply to cover approximately $464 billion in revenue, effectively locking in the economics for much of what Nvidia intends to ship over the next 18 months before cost inflation becomes a meaningful headwind.
That pre-purchasing strategy, Moore argued, gives Nvidia a structural advantage over competitors developing custom AI chips or merchant ASIC alternatives. Those rivals will either face greater margin pressure or be forced to raise prices to offset input cost increases, while Nvidia’s supply chain commitments shield it from the same dynamic in the near term.
The one acknowledged margin headwind is the Rubin architecture ramp. Launching a new architecture always carries higher initial costs, and Morgan Stanley is now modeling gross margins declining to 72.7% by fiscal year 2028, down from above 75% currently. The firm views that as a manageable compression rather than a structural deterioration.
Maule/Getty Images
The three debates Morgan Stanley says Nvidia earnings will address
Moore identified three investor debates the May 20 call is expected to move: market share versus ASIC alternatives, gross margin trajectory, and Vera Rubin readiness.
On market share, Moore remained optimistic that Nvidia offers the best total cost of ownership across AI workloads, but conceded the debate is “difficult to disprove” while compute remains supply constrained. He does not expect the earnings call to fully resolve this question in either direction.
On Rubin readiness, the note was direct: Vera Rubin is on schedule, with some normal startup challenges that could push rack deliveries a few weeks beyond initial estimates. Morgan Stanley sees no disruptions and expects strength in both Rubin and Blackwell demand through the second half of 2026.
Moore also offered a framework for how to think about Nvidia’s valuation relative to these debates. He estimated Nvidia could exit 2027 generating approximately $16 per share in annualized earnings power. A $16 perpetuity discounted at 8% with 1% terminal growth produces a present value of approximately $229, close to where the stock recently traded at $225.
His interpretation: Investors are currently pricing in a significant probability that margin compression, market share loss, or a general slowdown in AI spending will prevent any meaningful long-term earnings growth. That is the bear case embedded in the current price.
Key figures from Morgan Stanley’s May 18 Nvidia note:
- Price target: $285, raised from $260; overweight, top semiconductor pick; analyst Joseph Moore; based on 22x CY27 EPS of $12.99
- Expected Q1 beat: Revenue approximately $3 billion above consensus; Q2 guide approximately $4 billion above consensus
- April quarter estimates: Revenue $79.264 billion, GM 75.0%, EPS $1.72; July quarter: revenue $87.88 billion, GM 75.1%, EPS $2.01
- FY27 estimates: Revenue $380.591 billion, EPS $8.61; FY28: revenue $587.45 billion, GM 72.7%, EPS $13.11
- Data center revenue projection: $1.07 trillion CY25-CY27; $884 billion CY26-CY27 versus Street consensus of $785 billion
- Supply positioning: $95 billion in purchase commitments, $21 billion in inventory as of the 10-K; covers approximately $464 billion in revenue at 75% gross margin
- Bull and bear cases: Bull case $330, bear case $160; present value of $16 perpetuity at 8% discount rate and 1% terminal growth equals approximately $229
Source: Morgan Stanley May 18 note
What investors should watch when Nvidia reports on May 20
The most important output from the May 20 report will not be whether Nvidia beats the April quarter estimates. Moore expects it to. The question is what management says about the July quarter guide and, more importantly, the visibility they provide into the 2026 and 2027 data center revenue trajectory that sits at the center of Morgan Stanley’s bull case.
If Nvidia’s management explicitly reaffirms that $1 trillion in cumulative data center revenue is achievable across 2025 through 2027, and if they provide detail on supply constraint management that validates Morgan Stanley’s supply chain framework, the gap between Street consensus at $785 billion and Morgan Stanley’s $884 billion estimate for calendar years 2026 and 2027 should begin to close. That revision process is what Moore describes as “a positive step toward a stock rerating.”
If instead management is cautious on forward visibility, the market share debate intensifies, or Rubin ramp concerns surface that go beyond the “normal startup issues” Morgan Stanley is already modeling, the $229 present value floor may prove difficult to hold.
Moore’s note is ultimately a statement that Nvidia’s current price reflects extreme pessimism about its long-term earnings power, and that the May 20 earnings call is the next opportunity for the company to begin correcting that misperception.
Related: Keybanc sets jaw-dropping Nvidia stock price target before earnings
