RDDT's 677% Earnings Pop: The Data Play Everyone's Sleeping On (And the Squeeze That Might Not Be)

By Raj Patel | Risk & Reward

Here's the setup that's got retail buzzing: Reddit (RDDT) just posted 677% EPS growth—beating 2027 analyst expectations in 2026. The stock sits at under $200 despite what bulls call a "transformational" AI data licensing thesis. Meanwhile, SoundHound (SOUN) has zero shares available to short, a 58% borrow rate, and earnings this week. Two very different risk profiles. Two very different questions.

Let me walk you through what I'm seeing in the data, because the risk-reward math here matters.

RDDT: The AI Data Thesis

The bull case is clean: Reddit generates the exact data frontier AI models need—human, unfiltered, threaded debate with upvotes filtering signal from noise. You can't synthesize this without model degradation. The company has $1.4B cash, minimal debt, and only 203M shares outstanding. Forward P/E of 19 on 200%+ growth.

Here's the catch: The market is pricing this as a mid-tier ad-tech company, not an AI infrastructure play. That's either a massive inefficiency or a warning sign. The AI data thesis is real—I've seen the licensing deals—but the legal battle with Anthropic and Perplexity could go either way. If Reddit wins or settles favorably, it sets precedent that reprices training data across the industry. If they lose? The thesis unravels.

SOUN: The Classic Squeeze Setup

This is pure speculative flow. Up 20% Friday, climbing overnight, zero short inventory, 58% borrow cost. Twilio's earnings showed voice AI demand is real, and Cramer said he "can't be in it"—which, historically, is a contrarian buy signal. WSB has 700 contracts at $10.5 calls expiring Friday.

But here's what the squeeze crowd forgets: The borrow rate tells you shorts are already paying dearly to stay in. They're not capitulating yet. A bad earnings print and this thing reverses hard. You're not early. You're chasing.

What Retail Is Getting Right (And Wrong)

The RDDT thesis has genuine intellectual merit. The data layer is the next bottleneck in AI, and Reddit's user-generated content is uniquely valuable. But I'm seeing the same pattern we had with crypto in 2021—retail is so excited about the narrative that they're ignoring valuation discipline. A 47 P/E with 677% EPS growth is justified if that growth persists. But what if Q2 is just 150%? The multiple compresses fast.

On SOUN, retail is correctly identifying the squeeze mechanics but mis-sizing the risk. One trader posted a $735,000 position in weekly calls. That's not a trade—that's a lottery ticket with expiration. The borrow rate alone tells you this could go either way on earnings.

The GameStop-eBay chatter? Ignore it. A $12B market cap company offering $56B (half in stock) for eBay is financial theater. It's not actionable.

The Math

RDDT (Position Size: 2-3% if bullish):
- Upside: 40-60% if AI data thesis gains mainstream acceptance and legal wins materialize
- Downside: 25-35% if growth normalizes or legal challenges mount
- Risk-reward: ~1.5:1 (acceptable for a speculative growth position)

SOUN (Position Size: 1% max—this is gambling, not investing):
- Upside: 100%+ on a squeeze and blowout earnings
- Downside: 60-80% on a miss (volatility works both ways)
- Risk-reward: Maybe 1.5:1 if you believe the thesis, but the timing risk is extreme

Methodology Note: Analysis based on approximately 280 high-engagement posts and 3,500+ comments from Reddit's investing communities over the past 24 hours. I'm overweighting the RDDT AI data thesis because it aligns with my prior conviction that data scarcity is the next AI bottleneck—but I need to check my bias here. The legal uncertainty is real, and I may be too dismissive of the ad-tech bear case. Confidence: 52%.