On President Trump’s first full day in office – January 21 – he announced a partnership between ChatGPT founder OpenAI, database company Oracle, and Japan’s SoftBank called Stargate. Stargate’s purpose is to jointly deploy $500 billion to build data centers in Texas and develop the electrical power needed to run those data centers.
The first Stargate data center is already under construction in Abilene, Texas. Last year, Microsoft started building a massive AI “supercomputer” that will use tens of thousands of Nvidia chips. The processing power will be dedicated to OpenAI to run its AI models.
The announcement of the project on OpenAI’s website says:
For sure, the breakdown of responsibilities has more detail than what’s included in that statement. Still, it sounds as if SoftBank will act as the bank for Stargate – providing the funds for the data center buildout.
That’s great for OpenAI. And for Oracle too, as it’s the company that will provide the data on which OpenAI will train its AI models. Maybe Oracle just gets to hand SoftBank a bill for its data services? That’d be pretty sweet…
There’s just one problem – SoftBank. SoftBank is an investment company, roughly similar to Berkshire Hathaway, in that they both buy stakes in other companies or even buy out entire companies.
Except that SoftBank founder and Chairman Masayoshi Son is no Warren Buffett. In fact they’re pretty much polar opposites.
SoftBank Losses
Warren Buffett likes value, he likes businesses that have an easy path to compound growth over long periods of time. Buffett doesn’t make “splashy” investments…
Masayoshi Son seems to thrive on splashy investments. His investment style has been called Monopoly Investing – where he backs money-losing businesses with so much money that the companies can continue losing money and taking market share until the companies are so dominant that they can raise prices.
Uber (UBER) is a great example of monopoly investing. Uber posted massive losses for years to undercut traditional taxi costs and drive that business into a ditch. Once it had a dominant position, prices went up.
In 2017, Masa Son (for short) had SoftBank buy a 15% stake in Uber for $7 billion, valuing Uber at $48 billion. The same year he put another $1.25 billion into Uber at a valuation of $70 billion. SoftBank’s dollar cost for its Uber stake was $34.50.
Yes, he paid top dollar at the time…
You could look at Uber’s current $142 billion valuation and $67 share price and think “yeah he did pretty good.”
Except he didn’t. Massive losses for SoftBank forced SoftBank to sell its entire Uber stake for $44 a share in 2021 and 2022, netting the company $5.6 billion. Not terrible, except for the fact of those massive losses…
During the second quarter of 2022, SoftBank lost $21.6 billion! In three months! And it lost $19 billion the previous quarter.
Amazingly 2024 was its first profitable year since Covid – Masa Son turned a $4.6 billion profit last year – no doubt helped by its acquisition of chip-architecture company ARM Holdings in 2016. SoftBank paid $32 billion for ARM, which went public in September 2023 at a $54 billion valuation and is currently worth $169 billion.
Too bad SoftBank didn’t hold onto the 5% stake it bought in Nvidia in 2017. Masa Son sold the Nvidia stock in 2019 for $4 billion. That would’ve been worth $160 billion today! Doh!
The poster child for SoftBank’s bad investment decision was WeWork. If you don’t know, WeWork was a real estate company that somehow managed to convince investors that it was actually a tech company. Remember when “shared space” was a thing? That was WeWork…
The idea was that WeWork signed long-term leases and then offered upstart businesses short term leases on office space. Somehow, Masa Son pumped $16 billion into WeWork including $1 billion to WeWork’s founder to walk away after the company’ proposed IPO was scrapped and the shares were mostly worthless.
It hasn’t been all bad news for SoftBank. Its investment in Alibaba 23 years ago has grown 57% a year and is now worth $73 billion.
Late to the Party
It’s hard to say exactly what OpenAI was worth when Microsoft invested its first $1 billion in 2019.
Another $10 billion in 2023 gave Microsoft a 49% share of OpenAI profits until an undisclosed cap is hit…
Early in 2024, OpenAI was worth $80 billion. Today it’s about to raise more money at a $157 billion valuation. Pretty obvious that Microsoft’s investment was about as savvy as it gets…
But you know who’s about to buy a $25 billion stake at a very high $157 billion valuation? SoftBank.
I’m not going to tell you that SoftBank will lose money on an investment in OpenAI. It probably won’t. But it sure waited a long time, letting OpenAI get a lot more expensive, before making a move.
This is a pretty consistent pattern for Softbank. It tends to be late to the party.
So what does that mean for Stargate? Could it be that SoftBank’s move to get into the AI data center trade via Stargate means it is late to the party once again?
It wouldn’t be the first time a Masa Son investment came when the market was close to a top.
Cheers,
Briton Ryle
Chief Investment Strategist
Outsider Club
X/Twitter: https://twitter.com/BritonRyle
You Might Also Like:
ASTS Rallies
https://www.outsiderclub.com/space-race/
More on Stargate
https://www.outsiderclub.com/half-a-trillion-for-ai/
Foghat? Really…?
https://www.outsiderclub.com/foghat-really/
