Softbank's New Playbook

The AI Assembler

It is a little strange when the world's most famous venture gambler starts acting like a blue-chip portfolio manager. And yet, here is Masayoshi Son, the man who lost more money than anyone in history during the dot-com bust and then poured billions into WeWork, now selling off his high-risk Vision Fund assets to buy stakes in the most obvious winners of the AI boom: Nvidia, TSMC, and Oracle.

On its face, this is a straightforward move to gain exposure to the AI hardware stack. But for Son, this is not just a portfolio rebalance. It is the raw material for a wildly ambitious, multi-trillion-dollar plan to assemble and control the entire AI supply chain. While rivals are focused on building a better model or a faster chip, Son is playing a different game. He isn't trying to build the engine, the fuel, or the factory. He's trying to own the entire railroad.

One way to think about it is that Son is collecting the key pieces of the AI puzzle, rather than trying to build them from scratch:

• The Brains (AI Models): Son is not building his own frontier model. Instead, SoftBank is set to invest up to $32 billion in OpenAI, aiming to become a key financial partner to the world's leading AI lab.
• The Muscle (GPUs): Son is not designing a rival to Nvidia's chips. Instead, he is building a $3 billion stake in Nvidia, a move that one analyst noted could "buy more influence and access to Nvidia's most sought-after chips." It is a multi-billion-dollar attempt to "skip the line."
• The Factories (Fabs): Son is not building his own foundries. Instead, he's buying into TSMC while courting it to participate in his planned $1 trillion AI hub in Arizona.
• The Blueprint (Chip Design): This is the one piece Son already owns. SoftBank’s roughly 90% stake in Arm Holdings serves as the strategic core of this entire vision.

This strategy is a pivot born of necessity and, perhaps, remorse? SoftBank famously sold a 4.9% stake in Nvidia in 2019 that would be worth over $200 billion today (Jensen Huang even joked about it publicly). As one portfolio manager told Bloomberg, Son’s new vision is about controlling the entire ecosystem:

I think he sees himself as the natural provider of AI semiconductor technology. What Son really wants to do is capture the upstream and the downstream of everything.

The sheer scale of this ambition is also its greatest challenge. The $500 billion Stargate data center project and the $1 trillion Arizona manufacturing hub require a level of financial engineering that is unprecedented. Wall Street remains deeply skeptical, with SoftBank's stock still trading at a roughly 40% discount to the value of its assets. And the entire plan hinges on a close and productive relationship with the Trump administration, even as his proposed $6.5 billion acquisition of chip firm Ampere faces a probe by the Federal Trade Commission.

Ultimately, Son is betting that capital, vision, and his controlling stake in Arm are enough to weave these disparate, powerful entities into a cohesive empire. It's a gamble on becoming not just a player in the AI race, but the person who owns the entire stadium.

The Intel Squeeze

For a company that once defined the very rhythm of technological progress with Moore's Law, Intel has been having a rough go of it. The latest indignity arrived this week in the form of a credit rating downgrade from Fitch, which pushed the chipmaker's debt to just two notches above "junk" status. It is a striking development for a former titan.

The downgrade itself isn't the whole story. The real story is in the reasoning, which paints a picture of a company trapped between relentless competitors and a paradigm shift. Here's Reuters:

"Credit metrics remain weak and will require both stronger end markets and successful product ramps, along with net debt reduction over the next 12-14 months" for Intel to recover its recent ratings, Fitch analysts wrote on Monday.

Fitch added that while Intel holds a better market position than other similarly rated peers, its financial structure is relatively weaker and it faces "higher execution risk."

One way to think about this is that Intel is getting squeezed from two fronts. From the bottom, its traditional cash-cow businesses in PCs and servers are under a relentless assault. Longtime rival AMD continues to chip away at its market share, while a new wave of Arm-based processors from companies like Qualcomm is attacking its position in the emerging AI PC market. This is the slow, grinding erosion of the business that is supposed to be funding everything else.

From the top, the single biggest technological boom in a generation—the buildout of AI infrastructure—is happening almost entirely without Intel. The new world of AI data centers is being built on Nvidia's specialized GPUs, not Intel's general-purpose CPUs. While Nvidia's revenues are exploding, Intel's management has had to concede that the company is not "participating in the cloud-based AI data center market in a meaningful way."

Intel's grand plan to escape this squeeze has been its audacious bet to become a world-class chip foundry, competing directly with TSMC. The only problem with this long-term gamble is that the plan itself now seems to be up for debate. Recent reports suggest that Intel's new CEO is already considering a major pivot, potentially writing off billions invested in the company's near-term "18A" manufacturing process to instead focus on a future "14A" technology that might be more competitive.

This is the sort of expensive strategizing that a company with a rock-solid balance sheet can afford. Intel no longer has that luxury. And this is where the credit downgrade becomes so significant. It is Wall Street's way of saying it's hard to fund a multi-billion-dollar, decade-long turnaround effort when your core business is struggling and you can't seem to decide which turnaround you're attempting. Intel is stuck, with its future looking not just expensive, but increasingly uncertain.

The Scoreboard

• Semiconductor: TSMC cracks down on potential trade secret breach, initiates legal action (Reuters)
• Software: Palantir tops $1 billion in revenue for the first time, boosts guidance (CNBC)
• Self-driving: Baidu plans to expand its robotaxis to Europe with Lyft deal (CNBC)