TL;DR: The market is panicking because LLMs (Text) seem to be hitting a performance ceiling, causing AI stocks like Oracle to slide. Investors worry that if the magic stops, OpenAI becomes just another software vendor. But this view is myopic. The "Scaling Law"—the formula that turns capital into intelligence—hasn't failed; it just migrated. We are moving from Text to Video. And this is where the real monetary capture begins.
James here, CEO of Mercury Technology Solutions. Taipei - December 25, 2025
The stock market is a moody beast. Recently, we saw a sharp correction in the AI sector. The narrative? "LLMs have hit the ceiling."
If text models have indeed peaked, then the "AGI Premium" evaporates. OpenAI stops being a wizard and starts being a utility company, while ecosystem giants like Apple and Google regain the throne because they own the distribution.
But I argue that the panic is misplaced. The "Scaling Law" is far from over. It is simply shifting mediums.
1. The Predictability of Money: Why Investors Love Scaling Laws
To understand the market, you must understand the difference between R&D and Scaling.
- Pure R&D is Gambling: You fund 100 algorithms. 99 fail. 1 works. This is high risk.
- Scaling is Engineering: You know that if you double the compute and data, the model gets X% better. This is predictable.
Investors love predictability. The text frontier might be settling, but the Video Generation frontier is currently at the "GPT-4" stage—early, rough, but primed for exponential growth. This means we can pour money in and reliably get better performance out for at least two more generations.
2. The Monetization of Scale: The "100,000 Ads" Theory
Why is "Application" essential to catching monetary value? Because AI allows us to invert the economics of production.
The Advertising Use Case:
- Old World: A brand spends $10 million to produce one Super Bowl ad shown to 1 billion people. It’s a shotgun approach.
- AI World: The brand spends the same $10 million to generate 100,000 different ads, each tailored to a specific micro-segment of that 1 billion audience.
This is why Meta (Facebook) is unkillable. Despite burning billions on the failed Metaverse and the open-source Llama project (which had chaotic leadership and massive talent bleed), Meta’s stock survives. Why? Because their core ad engine is bulletproof.
When Video AI matures, Meta won't just target you; they will generate the video you see on the fly. Conversion rates will jump by an order of magnitude. That is how you capture value: Scale the output to match the diversity of the audience.
3. The Deflation of Content Costs
The monetary value of AI isn't just in making things "better"; it's in making expensive things "cheap" at scale.
The Gaming Market: A "Triple-A" game costs hundreds of millions because you need armies of artists designing 3D textures and NPCs. AI Video/3D generation reduces this cost to 1/10th. Suddenly, the "Blockbuster" experience becomes accessible to mid-sized studios.
The "Dopamine" Market (Short Dramas): Consider the booming "Short Drama" industry (e.g., "The CEO Fell for Me").
- Current Cost: Top actors charge 80k-100k RMB per day. A cheap production costs 10 million RMB.
- Future Cost: GPT-5 level Video AI generates the entire series—consistent characters, fluid motion, perfect lighting—without a single human actor.
The willingness to pay for this "trashy" entertainment is incredibly high. When you remove the human labor cost, the profit margin explodes.
Conclusion: The Tool vs. The Scale
The current stock dip is just a sanity check on CapEx (Capital Expenditure). Investors are asking, "We spent $10 billion; where is the return by 2027?"
The return isn't in a smarter chatbot. The return is in Industrialized Scale.
A tool (AI) is only useful if it solves a problem. But a tool only captures Monetary Value if it can solve that problem at a scale humans cannot match.
- Humans cannot make 100,000 unique ads.
- Humans cannot shoot a movie for $100.
- AI can.
The investors catching the falling knives right now aren't betting on chatbots. They are betting that the Scaling Law will do for Video what it did for Text. And when that happens, the revenue won't just add up; it will multiply.
Mercury Technology Solutions: Accelerate Digitality.
