The AI Model Race
AI model race has shaped my business perspectives
In May 2023, Google, facing pressure from LLMs on all sides, made a dramatic statement: "We Have No Moat, And Neither Does OpenAI." They then went all-in on AI, showing their determination to jump into the LLM battlefield - then the AI arms race kicked off.
Fast forward to today, the LLM landscape has completely transformed. But Google's statement? It's looking more and more spot-on. The AI industry has attracted massive amounts of capital, talent, and companies in these past few years. In less than 3 years, we've compressed what would've taken at least a decade of development. This power play, happening right before an AI revolution, has totally changed how I think about business moats. It touches on so many deep concepts: static vs dynamic business moats, market efficiency vs inefficiency, and how competitive advantages can grow or disappear.
Rewind to late 2022 - GPT's sudden emergence shook the tech world, and the market started reevaluating AI's worth. Then in 2023, startups and VC money flooded in. Companies like Anthropic, Cohere, Mistral, and Character went full steam ahead with their own LLMs. Nvidia, basically selling the shovels in this gold rush, saw their business explode and their stock go crazy. Everyone thought computing power and R&D speed were the key barriers to entry, thanks to scaling laws. The AI model war was on.
Come 2024, tech giants started throwing astronomical amounts of money around. Meta and Google's trillion-parameter models showed what brute force could achieve, and Musk jumped in with xAI. Among the early players, only OpenAI and Anthropic, backed by Microsoft and Amazon, managed to stay strong. The rest? Either got bought out or had to pivot. That's when the big players started dominating, and having deep pockets became the new price of admission.
By late 2024, people started questioning scaling laws after a boring stretch. Then OpenAI dropped the O1 model, blowing everyone's minds and opening up new frontiers with RL and test-time compute scaling. The market went nuts again, and ChatGPT's user base skyrocketed. More massive funding rounds and aggressive investment plans followed. I heard some Google folks say they were at least six months behind. First-mover advantage, brand power, and talent suddenly seemed like the barriers to entry. OpenAI was made king again.
Hit 2025, and OpenAI and the MAGA crew proudly announced Project Stargate. Then BAM - Deepseek came out of nowhere like a bolt from the blue. Not only did it wipe nearly $600B off Nvidia's market cap, but it also shook everyone's faith in competitive barriers. Deepseek had none of the usual advantages - no capital, computing power, or first-mover advantage - but they somehow matched O1's capabilities before all these American players with deep pockets. That "6-month lead"? Gone in less than a quarter. Suddenly everyone's questioning OpenAI, with some predicting they'll fall behind this year.
So what makes a real moat? The answer's fuzzy again. The market's split - some still believe in pre-train scaling laws (like Musk's xAI, which quickly built a 200K GPU cluster and cranked out the SOTA model Grok3, scoring one for MAGA). Meanwhile, Microsoft's cutting data center orders, thinking the industry's overinvested. Some say data will be the new moat, making companies with lots of data and use cases like Google/X/Meta the real winners.
The story's now come full circle. Google, who once said "We Have No Moat," is now seen as having real competitive edge. Whether this is right and how long these barriers will last? Who knows. These past three years have been wild, with dramatic turning points: GPT's emergence, Nvidia's stock saga, the O1 model's splash, tech giants' massive spending, Deepseek's surprise - business legends that should've taken a decade, all crammed into AI's revolution.
The future's hard to predict, but looking back is pretty exciting. It's amazing how efficient the market can be - as soon as a tech approach or business rule proves successful, huge money pours in and flattens the returns curve within six months. But there's also this herd mentality - massive funds rush in without caring about ROI, swinging between belief and doubt, causing huge market volatility.
We've seen how powerful business moats can be - first-mover advantage, capital intensity, and computing power have all given companies huge leads at different times. But we've also watched these moats crumble and shift. Market narratives jump around with each industry event. Companies rise to glory one minute and fall the next - you watch them build their towers, then watch those towers collapse.
2025 - can't wait to see how this wild story continues!