After markets closed last Wednesday, Nvidia delivered the most impressive single quarter in tech history and the market barely blinked. That wasn’t investors being irrational — it was investors telling you something: one great earnings report doesn’t settle the debate about whether the AI spending cycle is real, sustainable, and actually flowing through to revenue. It just adds one data point.
This Wednesday, three more data points land at the same time. Salesforce, Snowflake, and Marvell Technology all report after the close — and together they cover the full stack of enterprise AI: the software layer, the data layer, and the chip layer. By Thursday morning we’ll know a lot more about whether corporate America’s AI investment is translating into actual business results, or whether the gap between AI hype and AI revenue is still very much open.
Here’s what to watch in each.
Marvell Technology (MRVL) — The Chip Layer
Consensus estimates: ~$2.4 billion revenue (+26% year-over-year), ~$0.75 adjusted EPS
Of the three companies reporting Wednesday, Marvell has the most specific and testable AI story. The company makes custom AI chips — application-specific integrated circuits, or ASICs — for hyperscale cloud customers, alongside optical interconnects that move data between AI accelerators at the speeds required for modern training and inference workloads.
The bull case is straightforward: as hyperscalers like Amazon, Google, and Microsoft build out their own custom silicon to reduce dependence on Nvidia, Marvell wins design wins. The optical interconnect business benefits from the same buildout — you can’t run large AI clusters without high-bandwidth, low-latency data movement between chips, and that’s Marvell’s lane.
Stifel raised its price target ahead of the print and explicitly flagged it expects a beat-and-raise — meaning Marvell tops the $2.4 billion revenue estimate and guides the next quarter above current consensus. That’s the setup heading in.
The one question: Did custom AI chip revenue accelerate or decelerate sequentially? Marvell’s data center segment has been growing fast, but the market wants to see that the design wins it announced over the past 12 months are converting into shipments — not just promises. Watch the data center revenue line and the forward guidance more than the headline beat.
Snowflake (SNOW) — The Data Layer
Consensus estimates: ~$1.32 billion revenue (+26.8% year-over-year), ~$0.32 adjusted EPS
Snowflake is the place where enterprise data lives, and AI workloads run on data. The thesis is that every company building AI applications on top of their own data — which is most of what enterprise AI actually looks like in practice — needs a place to store, query, and process that data at scale. Snowflake has positioned itself as that place.
The company has spent the last year building out AI-native features: Cortex AI for building AI applications directly within Snowflake’s platform, and a series of partnerships with AI companies that use Snowflake as their data backbone. The pitch to enterprise customers is that they don’t have to move their data to use AI — they can bring AI to their data.
Revenue growth of 26.8% year-over-year would be strong in any environment. The question is whether the AI-driven workload acceleration is showing up in the metric that matters most for Snowflake: product revenue growth and the net revenue retention rate, which measures how much existing customers are expanding their spend. A retention rate above 130% signals the upsell motion is working. Below 125% and the story gets more complicated.
The one question: Is AI actually driving more consumption on the platform — more queries, more data processed, more Snowflake Marketplace transactions — or are customers adding AI features without meaningfully increasing their overall spend? The NRR is the number that answers it.
Salesforce (CRM) — The Software Layer
Consensus estimates: ~$11.2 billion revenue, ~$2.30 adjusted EPS
Salesforce is the most complicated story of the three because the AI thesis here isn’t about chips or data infrastructure — it’s about whether AI can make the core software product meaningfully better, and whether customers will pay more for it.
The company’s big bet is Agentforce, its AI agent platform that lets businesses deploy autonomous AI agents across sales, service, marketing, and commerce workflows. The pitch is that instead of a human answering a customer service ticket or a sales rep manually updating a CRM record, an AI agent does it — faster, at scale, and at lower cost per transaction. Salesforce wants to be the company that makes enterprise AI agents mainstream.
The stock has been part of the broader software sector rebound — up roughly 20% from April lows — but the rebound has been driven more by multiple expansion than by earnings revision. Heading into Wednesday, the question isn’t just whether Salesforce beats $11.2 billion. It’s whether Agentforce is showing up in the numbers in a way that justifies the premium.
The one question: How many paid Agentforce seats are live, and what is the attach rate to existing customers? Salesforce will give some version of this disclosure on the earnings call. If Agentforce is mostly in pilot or the numbers are small, the AI premium in the stock is running ahead of the fundamentals. If adoption is accelerating, the 20% rally from April lows has room to extend.
The Bigger Picture
Each of these companies is answering a version of the same question in a different part of the stack: is enterprise AI spending real, and is it translating into revenue?
Nvidia’s $81.6 billion quarter last week confirmed that AI infrastructure spending is real at the hyperscale level. What Wednesday’s reports will tell us is whether that infrastructure investment is producing returns in the layers above it — whether the chips are being used to run software (Salesforce), process data (Snowflake), and fuel demand for more specialized silicon (Marvell). A clean sweep of beat-and-raises across all three would be a strong signal that the AI spending cycle is in full momentum. A miss in any one of them — especially a guidance cut — would raise questions about whether AI monetization is ahead of schedule or still years away from showing up in income statements.
The secondary story on Wednesday is the macroeconomic backdrop. Iran deal negotiations continue over the holiday weekend, with oil already falling 5% on hopes of a Strait of Hormuz reopening. If crude breaks below $90 heading into Wednesday’s session, risk appetite opens up and a strong earnings night could do real work for the NASDAQ. If the deal stalls and oil bounces back, macro pressure absorbs some of the good-news potential from the earnings.
Reports are expected after the close Wednesday, May 27. Earnings calls typically follow within an hour or two.