SoftBank’s $40B OpenAI Gamble: Is the $300B Valuation Sustainable?
Superthread founder David analyzes the $300B OpenAI valuation, the rise of Google Gemini 3, and why the 'DeepSeek effect' is changing the cost of AI infrastructure.
The $300 Billion AI Bet
Welcome to the latest update on the world of AI, where I try to talk about these shifts in a language an ordinary person understands.
The big headline: SoftBank has invested $40 billion into OpenAI at a $300 billion valuation. SoftBank is led by Masa Son, who’s Japanese, but this is essentially a Saudi-backed fund. Whether this is a good idea remains to be seen. SoftBank has made massive wins, but it’s also had high-profile blunders, like the WeWork situation.
That said, when it comes to business-to-consumer AI, OpenAI is dominant. They have a massive user base willing to pay $20 a month, which means they’re making more money per user than platforms like YouTube or Facebook. Plus, the experience is currently ad-free, which users love.
But has SoftBank seen something we haven’t? I’m on the fence about whether a $40 billion injection at this price point is a ‘good’ idea, though in this world, even doubling your money is a massive return.
The Rise of Gemini and the DeepMind Factor
Recently, Google Gemini 3 has surpassed OpenAI in several key rankings. This isn’t surprising to me. Google’s AI division, DeepMind, has been playing this game for a long time.
Consider this: OpenAI’s former chief scientist, Ilya Sutskever, was originally an intern at DeepMind in London. That speaks to the depth of talent Google has. If anyone has a shot at achieving AGI (Artificial General Intelligence), it’s probably DeepMind.
The Talent Exodus and Seed Round Madness
We are seeing incredible movement in the talent market.
- Mira Murati (former OpenAI CTO) has raised at a $12B valuation for her new startup, Thinking Machines. Source - financialit.net
- Ilya Sutskever has raised funding for his venture, Safe Superintelligence (SSI), at a $32B valuation. Source - computing.co.uk
It’s wild to see people raising ‘seed rounds’ at these numbers. It raises a fundamental question: Are these foundational models a final product, or just the infrastructure?
The ‘DeepSeek Effect’: Cheaper, Faster, Open
The Chinese company DeepSeek has changed the math. They created a foundational model as capable as ChatGPT but on a fraction of the budget, offering it at a much lower cost to users.
This leads to the question of Europe’s place in the race. Does Europe stand a chance? We have Mistral in France, and the talent in Europe is 100% there. We have the engineering history and the expertise. However, the investment climate isn’t as aggressive as it is in America.
In Europe, the culture is more conservative. Because entrepreneurship isn’t as supported locally, our best talent often leaves for Silicon Valley.
XAI and the Twitter Data Play
Finally, there is Elon Musk’s XAI. In my opinion, his move is likely an attempt to recover the money he overpaid for Twitter (X). Twitter has a lot of content, though I wouldn’t call it high-quality. You want to train these models on high-signal data, not nonsense.
Source - datastudios.org%2C%20and%20xAI’s%20Grok%204.1)
Personally, I don’t know anyone using XAI yet; everyone is still on ChatGPT. This circles back to why that $300 billion valuation for OpenAI might actually make sense; they have the mindshare.
Build faster with the tool designed for the AI era.
High-performance teams don’t have time for slow, bloated software. Experience the speed of Superthread. Sign up for Superthread for free
