A recent tech merger and acquisition news story has been dominating the headlines, but the most interesting part is the silence of a certain individual.

On April 21st, when SpaceX announced the $60 billion acquisition of Cursor, some reactions in Silicon Valley were quite telling.
This deal should have been the biggest tech M&A of the year, but one person chose to stay silent — Elon Musk’s old rival.

Why stay silent?
Three years ago, this person invested $8 million in seed funding for Anysphere, the parent company of Cursor, under the name OpenAI — the company's first institutional investment.
Later, he was rejected when trying to acquire Cursor, and instead turned around and spent $3 billion to buy Windsurf, a competitor of Cursor.
Now, looking at Musk’s price, it’s twenty times what he once wanted for the asset.
Six days later, he will face Musk in Oakland court — a lawsuit about betrayal and trust.

On the surface, this looks like a simple acquisition, but digging deeper reveals five hidden truths behind the deal.

First layer: Musk isn’t as dominant as the outside world thinks.
xAI claims to have the world’s largest AI computing cluster, Colossus, with 100,000 NVIDIA H100 GPUs, planning to expand to 200,000.
This number is often cited as proof that Musk has won the compute war.
But there’s an awkward detail — a memo from inside xAI has leaked.
The memo was written by Michael Nicolls, who was transferred from SpaceX’s Starlink engineering team.
He bluntly states that xAI is “significantly behind” in AI competition.
What’s the key data?
xAI’s model floating-point utilization rate is only 11%, while industry average is 35-45%.
In other words, over 60% of those 100,000 H100 GPUs are idle.
The world’s largest compute cluster, operating at less than one-third of industry average efficiency.

The timing of this memo leak is also quite ironic.
All eleven co-founders of xAI have left, and Musk himself admits that “xAI was built poorly the first time, and is now being rebuilt from scratch.”
The most ambitious project, Macrohard, has stalled due to the departure of its core leadership.
Nicolls was brought in to transplant SpaceX’s ultra-efficient engineering culture and revive this massive machine.

Second layer:
This acquisition isn’t about a dominant player expanding territory, but a rebuilding company using money to buy time.
Forty days before the announcement, Cursor’s two product engineers, Ginsberg and Andrew Milich, both joined xAI, reporting directly to Musk, tasked with rebuilding Grok’s programming capabilities from scratch.
These two were key figures in Cursor’s journey from zero to $2 billion in annual revenue.
If you map out the timeline, Ginsberg’s return, Milich’s joining, and their clear assignment to “rebuild Grok’s programming capabilities” — the core of this deal was already completed forty days before the official announcement.
The contract is just the endpoint, not the starting point.

Third layer:
That $1 billion isn’t severance pay, but a prepayment for compute leasing.
The deal structure is clever: SpaceX can choose later this year to either pay $60 billion to complete the acquisition or pay $1 billion as a “partnership fee.”
This $1 billion is widely called “severance,” but that label completely misses the point.

What predicament is Cursor facing?
Its core technology is based on Anthropic’s Claude model.
But Anthropic has launched Claude Code, directly competing with Cursor.
More painfully, internal analysis shows that Claude Code costs $200 per month per user, but each heavy user consumes compute resources worth up to $5,000.
Last year, that number was $2,000 — a 150% increase in one year.
This means Anthropic is losing $4,800 per month per user, subsidized by venture capital, using a price war to corner the market.

In this situation, any company relying on Anthropic’s models can’t cover API costs with subscription fees.
Cursor must train its own models, which requires sourcing compute power independent of competitors.
xAI’s Colossus cluster is currently the largest independent compute source.
That $1 billion “partnership fee” is actually a prepayment for leasing that compute capacity.

Fourth layer:
A bigger strategic game.
The relationship between OpenAI and Cursor spans a full three years.
In October 2023, OpenAI led an $8 million seed round for Anysphere.
Two years later, the situation had completely reversed.
In November last year, Cursor completed a $2.3 billion Series D funding round, with a valuation of $29.3 billion.
OpenAI seriously considered acquiring Cursor but was rejected.
Instead, it bought Windsurf for just $3 billion — a tenth of Cursor’s valuation.

On April 21st, when SpaceX announced the $60 billion deal, only six days remained before Musk and his rival faced off in Oakland federal court.
The lawsuit stemmed from the rival’s accusation that Musk betrayed OpenAI’s nonprofit mission.
While there’s no evidence that April 21st was deliberately chosen, the timing is undeniably dramatic:
Six days before the court showdown, Musk publicly pegged a $60 billion price on a company that the rival had once invested in, wanted to acquire, but was rejected.
The $8 million seed investment from three years ago is now worth $60 billion.
The rival understands what that number means.

Fifth layer:
All of this is paving the way for a larger IPO.
On April 1st, SpaceX secretly filed for an IPO with the SEC, targeting a valuation of $1.75 trillion.
What’s the core narrative supporting this figure?
Musk plans to deploy up to one million data center satellites in space, using solar energy and natural space cooling to replace ground-based data centers’ power and water cooling needs, providing cheaper infrastructure for AI computing.
Musk has repeatedly claimed “space will be the cheapest place to run AI.”

But according to media reports like Reuters, the S-1 filing with regulators admits that this plan involves “unverified technology,” and its commercial viability is uncertain.
The official assessment of this biggest AI gamble is just a question mark.

This question mark explains the real logic behind acquiring Cursor.
The feasibility of the space compute plan depends on unverified physical and engineering assumptions.
But Cursor generates 150 million lines of code daily for global companies, representing real, tangible demand.
If the space plan’s future is uncertain, the most critical thing is to find a sufficiently large commercial load for that idle Colossus cluster.
Cursor’s $2 billion annual revenue makes it one of the fastest-growing SaaS products worldwide.
Integrating it into the SpaceX ecosystem isn’t just about monetizing compute — it’s also about proving to the capital markets that “AI infrastructure is already in place.”
In IPO roadshows, this is more convincing than any predictions about space data centers.

Using a software asset to support an uncertain hardware vision — that’s the essence of this $60 billion deal.