Behind OpenAI's $122B Funding: Stargate Project Stalls, Where's the AI Infrastructure Money Going?

Honestly, I nearly spilled my coffee when I saw OpenAI’s funding numbers. $122 billion, post-money valuation of $852 billion—this isn’t a unicorn anymore, this is Godzilla.

But here’s the interesting part: while I was still digesting this figure, another story broke—Stargate is hitting roadblocks.

On one side, a record-breaking funding round. On the other, an ambitious infrastructure project stalling. The contrast is fascinating.

Let’s put this money in perspective. $122 billion is roughly two and a half ByteDances, or about one Alibaba. Led by SoftBank with Microsoft following—this lineup basically tells the market: capital is still betting on the second half of the AI game.

But here’s the question: what does OpenAI need all this money for?

The official line is accelerating AGI research and expanding compute infrastructure. In plain English: buying GPUs, building data centers, hiring more people. But there’s a subtle point here—if Stargate were actually moving forward smoothly, OpenAI wouldn’t need to raise this much cash so urgently.

Stargate is OpenAI’s joint AI infrastructure project with Microsoft, SoftBank, and Oracle, planning to invest $500 billion in US AI data centers. It was hailed as crucial for America to maintain AI leadership.

But now we’re hearing about technical and regulatory challenges. What exactly? Nobody’s saying.

My personal take? This probably reflects a real dilemma in AI infrastructure: money is easy to raise, land is hard to find, and electricity is even harder.

Building a large-scale AI data center costs an astronomical amount just in electricity. I estimate a facility capable of training GPT-5-level models would burn through hundreds of millions in electricity alone per year. Not to mention that many parts of the US grid are already unstable—adding a few power-hungry monsters would definitely face pushback from local governments and residents.

Moreover, AI infrastructure has a longer payback period than expected. Unlike cloud computing in the mobile internet era where demand grew gradually, AI training demand comes in pulses—today I need to train a trillion-parameter model, tomorrow I might not need as much compute. This uncertainty makes infrastructure investments hard to justify.

So OpenAI’s $122 billion is, in some ways, preparation for Plan B. If Stargate works out, that’s icing on the cake. If it fails, at least they have reserves.

This is quite telling. On the surface it’s a huge win for OpenAI, but it might actually reveal a deeper anxiety in the AI industry: how much compute do we actually need? Can all this computing power really generate corresponding commercial returns?

Hold on, look at the data first.

From this angle, Anthropic’s Claude and DeepSeek’s R1 taking the efficiency-first route might actually be the more pragmatic choice. After all, if the marginal returns on model capability improvements are diminishing, then throwing more compute at the problem isn’t optimal.

I think the real significance of this funding round isn’t the money itself, but the signal it sends: the AI arms race isn’t over, but the battlefield is shifting. From who has the most GPUs to who can do more with fewer GPUs.

As for whether Stargate will succeed? I doubt it in the short term. Not a technical problem—a practical one. Money, land, electricity, talent. You need all four.

What’s your take on this?