Interesting how the AI infrastructure play evolved over the past couple months. So CoreWeave and SoundHound both reported earnings back in late February, and there was definitely some movement worth dissecting.

CorWeave's situation was pretty compelling from a stock prediction standpoint. The company had this insane $56 billion revenue backlog sitting there at the end of Q3—up 271% year-over-year. That's the kind of number that gets people's attention. They were guiding for $5B in 2025 revenue, with analysts expecting it to potentially double to $12B in 2026. The kicker was Nvidia's $2B investment in them last month, which basically signaled confidence in their ability to scale. They're building out AI data center capacity with Nvidia's new Vera Rubin chips, which supposedly cut inference costs by 10x. That's not trivial.

What made this an interesting AI stock prediction was whether they could actually convert that backlog into revenue faster than expected. The demand for GPU capacity from Meta, Microsoft, OpenAI and others was insatiable. So the question going into their earnings was really about execution.

Then there's SoundHound AI, which had gotten absolutely hammered—down over 65% from its October peak. But here's the thing about beaten-down stocks: sometimes the market overreacts. SoundHound's voice AI technology was seeing genuine adoption across automotive, restaurants, healthcare, retail, banking. Their AI Amelia agentic platform was gaining traction in a market expected to hit $47.5 billion by 2034. They had a $1.2 billion backlog at the end of 2024, up 75% from the prior year.

The valuation had compressed to under 20 times sales versus 27 times at year-end, which made it look interesting on a stock prediction basis. The real question was whether their quarterly results would convince investors the company was still on a growth trajectory despite the stock's decline.

Both of these AI companies had very different narratives—one was about explosive infrastructure demand, the other about a turnaround play in voice AI. Watching how their earnings played out was useful for understanding where the AI market was actually allocating capital versus where sentiment was pushing it.