Google Is Pumping The AI Bubble HARD

The circle is growing larger.

Photo by Pawel Czerwinski on Unsplash

By now, we all know that AI is just one giant circle jerk of a bubble. Big Tech companies like Nvidia, Microsoft, Google, OpenAI and Oracle have been passing the same few billion dollars amongst themselves for a while, creating a broken financial perpetual motion machine that does nothing but artificially boost their value. However, there was one glaring outlier to this arrangement: Elon Musk’s xAI, which kind of had its own thing going on. But thanks to Musk merging xAI with SpaceX and taking it public, xAI/SpaceX is now the centrepiece of this circular bubble, and no one is trying to capitalise on it more than Google. In fact, there is an argument that Google is trying to pump Musk so hard that it could cause the entire bubble to collapse. Let me explain.

Did you know Google is a shareholder in SpaceX? Back in 2015, they invested $900 million into the company and still hold a 6.11% stake (pre-IPO). If SpaceX reaches its utterly ludicrous IPO valuation of $1.75 trillion, then this stake would be worth $122 billion. In other words, if this IPO goes well, Google stands to make well over $100 billion in profit from a less-than $1 billion investment. For some context, Google made just under $110 billion in profit last year. So this one investment could literally double their profits this year, if all goes well.

But there is a problem — SpaceX is not exactly IPO-friendly.

Financially, the company is in a sticky situation, having lost $4.9 billion last year and losses are set to rise dramatically this year. The xAI side of the business is dead, with crushingly low infrastructure utilisation, functionally no market share, severe legal troubles, and horrifically large losses. The only profitable arm of the company was Starlink, which wasn’t actually all that profitable, and while it is growing, it shows signs it can’t sustain that growth in the long-term (read more here). All these problems should dramatically undermine SpaceX’s speculative value.

But SpaceX is asking to be valued at a level no company has ever seen. They made $18.7 billion in revenue last year, and are targeting a $1.75 trillion valuation, which equates to a revenue multiplier (how to value a business based on its annual revenue) of well over 90. For some context, a revenue multiplier of ten is considered extremely high. This has led some analysts to suggest that SpaceX is worth half its target IPO valuation, or just $780 billion. I, along with many others, believe even this figure is too high. Considering the immense amount of baggage SpaceX is carrying, a low-end revenue multiplier would be closer to six, and a high-end, extremely optimistic one would be 12. That would yield a more realistic valuation of $112 billion to $224 billion, or less than 13% of SpaceX’s targeted IPO value.

So, for Google to make their bank, they have to polish this turd, improve the perception of its internal workings, and push the hype to draw in as many suckers as possible.

Now, here is an interesting fact: Google also holds a rather large stake in xAI competitor Anthropic. Google owned 14% from a $3 billion investment but has since invested an additional $10 billion, specifically for compute power, which has likely raised its stake to the upper teens. Why does that matter?

Well, both Google and Anthropic have revealed that they would be interested in using Musk’s orbital AI data centres, making them so far the only companies to do so. The fact that anyone is interested is surprising enough, as the concept is utterly stupid. As I have covered before, orbital data centres will be exponentially more expensive than those on Earth, devastatingly unreliable, and almost impossible to deploy at scale. As it is, AI data centres are already a cash black hole, so no sane business would go anywhere near Musk’s moronic orbital ones. Therefore, this looks like Google and Anthropic using their clout to legitimise SpaceX’s orbital AI push, generate hype, and pump up that speculative value. It’s either that, or the business leaders at both companies are utterly braindead.

But Google and its partner, Anthropic, went one step further and have just signed deals to rent terrestrial AI computing power from SpaceX at extortionate prices.

Google has recently agreed to rent the equivalent of 110,000 Nvidia GPUs from SpaceX’s Colossus data centre for $920 million per month from October 2026 to June 2029. Back in May, Anthropic agreed to rent 222,000 Nvidia GPUs from SpaceX’s Colossus data centre for $1.25 billion a month for the next three years. Together, these agreements will generate $2.17 billion in revenue a month, or $26 billion a year, for SpaceX.

I cannot overemphasise the instability and overpricing of these deals.

For one, both agreements have a 90-day termination notice, meaning they can pull out with very little notice. In other words, this revenue is not guaranteed, and it allows both Anthropic and Google to rapidly jump ship once a cheaper option is available — whichthere will almost certainly be, because these agreements are shockingly overpriced!

SpaceX’s data centres use a combination of Nvidia H100, H200 and B200 chips. On the low end, a single H100 GPU rents for $3–4 per hour, and on the high end, a B200 GPU rents for $4–6 per hour. Let’s be generous and say that the going market rate for a GPU in SpaceX’s data centre is $5 per hour, and because there are 730.5 hours in a month, each chip’s going rate is $3,652.50 per month.

Google is paying $920 million per month for 110,000 of these chips. But 110,000 chips should cost $401.78 million per month, based on this $3,652.50-per-chip-per-month estimate. So, Google is paying double the market rate!

Anthropic is paying $1.25 billion a month for 220,000 of these chips. But 220,000 of these GPUs should cost $803,550 per month. So, Anthropic is paying 56% over the going rate.

No wonder they included these 90-day clauses in the deals. Neither Google nor Anthropic can stomach paying such excessive prices for compute. They will need to jump ship as soon as doing so benefits them.

So, why have Google and its underling paid so much?

Well, again, it legitimises SpaceX’s AI credentials, bringing the company further into the AI bubble and validating much of its speculative value. You could see this move as a flagrant PR stunt to pump speculation ahead of the IPO.

But this will also save SpaceX in the mid-term. Thanks to its AI and currently failing Starship project, SpaceX’s losses are mounting quickly. It posted a $4.27 billion loss for 2026 Q1, which is nearly as much as it lost in the entire year of 2025. Right now, it is losing about $1.4 billion a month, and if that trend continues, monthly losses could easily exceed $2 billion, if not $3 billion. This $2.17 billion a month from Anthropic and Google will almost certainly prevent SpaceX from posting devastating losses for its quarterly reports after its IPO, which could undermine and tank its gargantuan, lofty valuation.

Oh, and wouldn’t you know it, SpaceX has considerably watered down the insider lock-up rules (read more here), meaning Google can legally sell most of its SpaceX shares to the public market just a few quarters after the IPO. So, Google only needs to keep SpaceX looking financially buoyant or preserve its gargantuan speculative value for six to nine months, which will enable it to sell its shares at an eyewateringly high profit.

In other words, it looks like Google is using its sizeable coffers and its ‘desperate for cash’ underling Anthropic to prop up SpaceX via extortionately expensive and short-term compute deals to increase the value of its shares in SpaceX by legitimising its bad business, plugging its gaping financial holes, and pushing its laughably false speculative hype so it can offload its shares in SpaceX at a staggering profit.

That 90-day clause is the smoking gun. It enables them to make a profit and jump ship quickly.

Now, am I saying they will do this? No. Businesses like these do weird things. What I am saying is that this is what it looks like they are trying to do.

It’s also not like we don’t have precedent for such financial tomfoolery. Google’s profits recently jumped by $28 billion or 81% year on year! The media claimed this was due to their cloud computing doing rather well, but that isn’t entirely true. The vast majority of this $28 billion came from an increase in the value of their stake in Anthropic. What drove that increase? Well, they invested $10 billion into Anthropic for a tiny share, which dramatically raised the value of their previous stake and generated almost all of this $28 billion ‘profit’. Not to mention that this fake profit situation gets so much worse. Go watch Sasha Yashin’s amazing video to find out more. But my point here is that Google has no qualms about using questionable financial engineering to its benefit. So we have good reasons to be sceptical about their decisions with SpaceX.

So, how could all of this pop the bubble? If SpaceX’s IPO doesn’t fully play out — i.e.,if its value drops before these insiders, like Google, can make their billions — then this gravy train stops. This circular financing scheme only works if the money perpetually spins around and around, and more and more investors are willing to add fresh money into the perpetual motion machine. Google’s doubling down on SpaceX’s IPO creates a weakness in that system because if this plan doesn’t pay off, it will break this cycle, spook investors, and bring the perpetual motion machine to a screeching halt. The question is, was this enormous risk worth it? In time, we will see.

---

Thanks for reading! Everything expressed in this article is my opinion, and should not be taken as financial advice or accusations. Don’t forget to check out my YouTube channel for more from me, or Subscribe. Oh, and don’t forget to hit the share button below to get the word out!