SoftBank And White House Plan AI Industrial Parks

Summary

---

WSJ Tech News Briefing: SoftBank and White House Plan AI Industrial Parks

Date: December 5, 2025
Host: Patrick Coffey, The Wall Street Journal
Guests: Mark Vardabedian (WSJ Reporter), Elliot Brown (WSJ Reporter)

---

Overview

This episode of the WSJ Tech News Briefing dives into two major themes:

1. The challenges venture capitalists face evaluating new AI startups and their evolving revenue models.
2. SoftBank's partnership with the U.S. government to create massive "AI industrial parks," a proposal involving unprecedented levels of public-private collaboration and investment—with far-reaching implications for U.S. manufacturing, trade, and the global AI race.

---

1. Venture Capital and AI Startups: New Models, New Confusion

Evolving Revenue Models (00:49–01:59)

• Shift to Usage and Outcome-Based Pricing:
  Mark Vardabedian discusses how AI startups have abandoned traditional subscription models for usage and outcomes-based metrics.

  • Usage-Based: Revenue is tied to how much clients use the AI service, which can fluctuate rapidly.
  • Outcomes-Based: Startups charge customers based on tangible results—though these are often subjective and hard to standardize.

  "They're using usage and performance-based outcomes to measure revenue. That's totally different than the previous model and in a lot of cases confused and even frustrated venture investors..."
  — Mark Vardabedian, 01:22

• Experimentation is Rampant:
  Startups mix and switch between revenue models rapidly, sometimes even during fundraising.

  • Example: Prosper AI, providing voice agents for healthcare, pivoted through three pricing models in six months, all during a $5 million seed round.

  "Experimenting with one and then pivoting to another and then pivoting again. And this all happened as they were raising a $5 million seed round."
  — Mark Vardabedian, 02:53

Investor Reactions and Uncertainty (02:35–04:40)

• Moving Targets:
  The lack of consistency makes it tough for investors to assess valuation and future returns.

  "It's sort of like trying to hit a moving target. That's where some of the confusion is coming from."
  — Mark Vardabedian, 03:26

• Innovation or Obfuscation?:
  Patrick Coffey asks whether this is just the next frontier of business model innovation or reason for skepticism.

  "This is what Silicon Valley is famous for, trying new things, throwing ideas against the wall and seeing what sticks... They believe AI has a lot to offer. Now the trick is finding what the right revenue model is."
  — Mark Vardabedian, 03:58

• Potential Risks:
  There is growing concern over the potential for “creative accounting” in the evolving AI startup landscape.

  "It is very telling that investors are very unsure about how to gauge the revenue of some of these startups because that really feeds into the broader concerns about AI..."
  — Mark Vardabedian, 04:53

---

2. The Dream of AI Industrial Parks: SoftBank, Trump, and $1 Trillion Investments

The Vision: Crystal Land and Federal Collaboration (06:28–07:49)

• Masayoshi Son’s “Crystal Land”:
  SoftBank’s CEO is leading a plan to build “super industrial parks” focused on powering the U.S. AI and energy revolution.

  • Factories would produce components like fiber optic cables and chips vital for companies such as Meta and OpenAI.
  • The parks would be distributed nationwide, supported by hundreds of billions in investment.

  “They want to make these sort of super industrial parks scattered around the country... cranking out fiber optic cable and electric transformers and sort of things that go into this chunk of the economy that’s growing really quickly.”
  — Elliot Brown, 07:13

Key Challenges and Hurdles (07:49–09:21)

• SoftBank Outside Its Core:
  The company is primarily in tech, not physical manufacturing.

• Land and Regulations:
  Using federal land requires approvals, public notices, and coordination—a daunting bureaucratic hurdle.

• Demand Must Materialize:
  Building factories is only step one; someone must buy the output.

  “There’s a lot. So SoftBank...is a tech company, and this is in physical manufacturing... probably bigger is this would be using big chunks of federal land, which, you know, require a bunch of approvals and public notices...”
  — Elliot Brown, 07:56

The Japan-U.S. Investment Angle (09:21–10:49)

• Japanese Government’s Role:
  Japan has promised $550 billion in financing (in return for tariff relief as part of a trade deal), much of which could flow into these industrial projects.

• Public-Private Structure:
  The U.S. would own the parks and take 90% of profits; Japan retains some project veto power.

  “Japan...putting up money or loan guarantees to effectively help finance this...the terms...call for the U.S. to be able to pick the sites and uses for the funding.”
  — Elliot Brown, 09:11

• Risks of Political Volatility:
  With U.S. tariffs facing potential reversals (including a Supreme Court challenge), the entire financing scheme sits on a precarious foundation.

  “It seems like a potentially fragile deal because so much of the initial agreement focused again on reducing tariffs, and there’s so much uncertainty...”
  — Patrick Coffey, 10:33

Masayoshi Son’s Motivations and Market Timing (11:11–11:57)

• Son’s History:
  Known for bold bets during “hot” markets, particularly in AI, Son’s previous over-optimism sometimes led to losses. Now, market enthusiasm and his companies’ growing value have emboldened him to double down.

  “He gets really excited when markets get really hot in tech...he’s sort of been punished by the market by wasting a lot of money prematurely making that bet. But now the stars seem to be aligning...”
  — Elliot Brown, 11:14

• Bigger the Plan, Harder the Execution:
  Brown closes with skepticism about the project's feasibility but acknowledges its unprecedented scale and backing.

  “The general rule is the bigger the plan, the less likely it is to happen. On the other hand, there is a piece of paper out there saying Japan needs to commit to $550 billion...”
  — Elliot Brown, 10:56

---

Notable Quotes

• “They are using usage and performance based outcomes to measure revenue. That’s totally different than the previous model and... confused and even frustrated venture capital investors.”
  — Mark Vardabedian, 01:22

• “It’s sort of like trying to hit a moving target. That’s where some of the confusion is coming from.”
  — Mark Vardabedian, 03:26

• “They want to make these super industrial parks scattered around the country... cranking out fiber optic cable and electric transformers and sort of things that go into this chunk of the economy that’s growing really quickly.”
  — Elliot Brown, 07:13

• “The bigger the plan, the less likely it is to happen. On the other hand, there is a piece of paper out there saying Japan needs to commit to $550 billion, and most projects don’t have that.”
  — Elliot Brown, 10:56

---

Key Timestamps

• 00:49: Discussion begins on AI startup revenue models
• 02:53: Example of Prosper AI’s multiple revenue pivots
• 03:58: Perspective on business model “innovation”
• 04:53: Risks and investor uncertainty
• 06:28: Introduction of SoftBank/Trump “AI industrial parks” plan
• 07:13: Details of “Crystal Land” and proposed manufacturing clusters
• 07:56: The practical hurdles of building massive new infrastructure
• 09:11: Financing via Japan/US trade deal and public-private structure
• 10:33: Uncertainty over tariffs and legal challenges
• 10:56: Skepticism regarding large-scale execution
• 11:14: Masayoshi Son’s motivations and market context

---

Summary

This episode explores the chaotic search for viable business models among AI startups, which poses serious challenges for investors trying to assess real value in a sector awash with hype. It then shifts to a bold and unconventional blueprint for U.S. industrial policy—SoftBank’s scheme, floated in collaboration with the White House and the Japanese government, to build sprawling, AI-focused factory complexes (“Crystal Land”) across the country. The discussions reveal both excitement for AI-driven transformation and deep skepticism about the practical and political realities blocking these grand ambitions.

---

Transcript

[A]

AI is changing the world.

[B]

We're securing it.

[A]

Crowdstrike.

[C]

We stop breaches.

[A]

Welcome to Tech News briefing. It's Friday, December 5th, 2025. I'm Patrick Coffey for the Wall Street Journal. There's no shortage of new AI startups and they're looking for funding. But in this era, venture capitalists say figuring out how valuable these upstarts are isn't a straightforward proposition. Then imagine massive industrial parks rising around this country to house the factories that will fuel our AI revolution. You're skeptical, I get it. But SoftBank CEO Masayoshi San is all in to the tune of $1 trillion or more.

But first, venture capitalists are no strangers to fuzzy math and financial risk. But the biggest challenge they face in sizing up the many AI businesses now scouring Silicon Valley for funding may lie in figuring out whether their revenue claims are anywhere close to accurate. Journal reporter Mark Bartabedian talks about the new models that founders have adopted to tell investors how much they're bringing in. So Mark, at the core of this story is really pricing strategies. What are the benchmarks that they're using today in the AI boom?

[B]

In the AI era, startups have really shifted to a new model. They're using usage and performance based outcomes to measure revenue. That's totally different than the previous model, and that has, in a lot of cases, confused and even frustrated venture capital investors who are trying to suss out how much money these companies are making as they're trying to decide whether to invest or not. The new models in the AI era are really primarily based on usage. That offers investors a snapshot of adoption in the immediate moment. But in the AI era, that can abruptly change, sometimes overnight. And the other big model nowadays is outcomes based. So AI startups charging customers as their technology solves issues and successfully helps customers. That can be pretty subjective, that can be measured in different ways, and can offer difficult ways to measure the company's financials. And then there's the traditional simply charging per seat. They say that's basically charging per user a subscription fee. Nowadays in the world of AI, that is also tricky because they're so called power users.

[A]

Why are startups using these new models and how are they able to go back and forth with what seem like jarring pivots between approaches?

[B]

Startups are really experimenting with all these different models and oftentimes using a combination of all the models at once. Which is where venture investors say things really start to get confusing when they're trying to untangle multiple revenue models within the same company. And startups are throwing multiple things against the wall and seeing what sticks. In one case, I spoke to startup Prosper AI and they offer voice AI agents for the healthcare sector. And since June 2024 they've had three different pricing models. So experimenting with one and then pivoting to another and then pivoting again. And this all happened as they were raising a $5 million seed round. So that can give you a sense of what investors are having to deal with raising a funding round. At the same time, the startup's revenue model is changing. It's sort of like trying to hit a moving target. That's where some of the confusion is coming from. Investors, it's not clear how they should be evaluating these companies. And once they decide how to evaluate them, that could change the next quarter.

[A]

Some founders seem to argue that this could just be another venue for innovation, another way that AI is helping us do things differently. How skeptical should we be?

[B]

Well, investors did say that it is a brave new world out there that they are trying to navigate. You know, this is what startups do. This is what Silicon Valley is famous for, trying new things, throwing ideas against the wall and seeing what sticks. So it wasn't all couched to me in a bad way. Yes, investors are going to was through a lot of the uncertainty, but they're all in agreement that they believe AI has a lot to offer. Now the trick is finding what the right revenue model is.

[A]

So how are investors responding generally and how long until we see what looks like maybe some sketchy or creative accounting grow into something darker or ultimately prove that these companies have been truly great all along?

[B]

One thing that's important to note is much of the AI bubble concerns have focused on the largest AI startups out there in the public markets as well. This is an interesting case of very early stage startups grappling with an unclear environment about the return on investment. And it is very telling that investors are very unsure about how to gauge the revenue of some of these startups because that really feeds into the broader concerns about AI and when some of these companies are going to start to return on the massive amounts of investment that are getting pumped into them.

[A]

That was WSJ reporter Mark Vardabedian. Do you think we're in an AI bubble? If you're a listener on Spotify, be sure to let us know in this episode's poll or leave us a comment with your thoughts. Coming up, the dream of Trump industrial parks may become a reality. Only time and potentially a few hundred billion dollars from the Japanese government will tell. That's after the break.

[C]

Security tools miss what matters. Identity. They can't see who is accessing what, when, where, why, or at what risk level until it's too late. Sailpoint adaptive identity unifies identity, security and data detecting threats in motion and responding in real time. Because static security can't protect a dynamic enterprise. See what others Stop threats before they spread. Transform risk into resilience. Welcome to the new era of adaptive identity driven by Sailpoint.

[A]

Last year, presidential candidate Donald Trump proposed the idea of Freedom Cities to host manufacturing infrastructure on federal land. Now SoftBank CEO Masayoshi San wants to help him make it happen in partnership with the Japanese government, which this year agreed to invest $550 billion in the US in exchange for lower tariffs. Journal reporter Elliot Brown writes that the current plan is to build clusters of factories that will make the fiber optic cables, data center equipment, and ultimately computer chips that companies like Meta and OpenAI need to power their products. So, Elliot, let's talk about exactly what Masayoshi san is proposing here with what he calls Crystal Land.

[D]

Yeah, so this started as this dream of this mega industrial city a number of months back when SoftBank CEO pitched the President on it. But since it's evolved in concert with the Trump administration, where they want to make these sort of super industrial parks scattered around the country. And the idea is to have hundreds of billions of dollars of investment go in to make giant factories aimed at the AI and energy space where they're cranking out fiber optic cable and electric transformers and sort of other things that go into this chunk of the economy that's growing really quickly.

[A]

This sounds like a huge undertaking. Can we talk about the major hurdles of which there are many.

[D]

Yeah, there's a lot. So SoftBank, the main company sort of pushing this, is a tech company and this is in physical manufacturing of goods. So there's one, but probably bigger is this would be using big chunks of federal land, which, you know, require a bunch of approvals and public notices. You have the actually building these factories, finding things that are actually going to work there, and then ultimately having a demand for the product, because it's easy to say, let's build a factory, but ideally you have to have somebody buying the mounds of additional fiber optic cable and electric transformers that come with this.

[A]

It also made me think whether our infrastructure can even support something like this. I mean, I feel like we're already stretched out.

[D]

Yeah, I mean, I think it would be spread around the country, but we don't really know details about what, where and how, which is another challenge to the plan, is figuring out exactly what this is. But at the broad level, Masayoshi son from SoftBank is just a sort of big dreamer, a big tech optimist, and likes to take these big swings. And, you know, he sees space to get in big early on, the sort of physical side of the AI revolution, and also sees an opportunity to try to use a large chunk of the $550 billion that have been earmarked as part of the Japanese trade deal, where essentially Japan or putting up money or loan guarantees to effectively help finance this.

[A]

Yeah, that gets to the next big question, because we're talking hundreds of billions of dollars. That, that wouldn't come directly from San himself. The idea is that it would come from the Japanese government and Japanese tech companies, but that the US Federal government would ultimately own the project and get like 90% of the profits.

[D]

Yeah. So this goes to the details of the trade deal were negotiated to give Japan sort of relief on tariffs from the rate that had been set by Trump. And so as part of that deal, the government agreed to $550 billion of financing for something. And the way that the administration is currently looking at it, as we understand it, is to then direct a lot of that money toward this. And then, yeah, the terms of that agreement call for the US to be able to pick the sites and uses for the funding. And Japan has the ability to say no, but if they say no, then the US has the ability to ratchet the tariffs back up and then precisely how it would be financed or these grants or these loans. Is there additional money that needs to come from SoftBank or any of the companies it's working with? That's all yet to be seen.

[A]

It seems like a potentially fragile deal because so much of the initial agreement focused again on reducing tariffs, and there's so much uncertainty around those tariffs still with them scheduled to go before the Supreme Court. So how much could that play into the future of this proposal?

[D]

I've spent a lot of years writing about really big plans, and the general rule is the bigger the plan, the less likely it is to happen. On the other hand, there is a piece of paper out there saying Japan needs to commit to $550 billion, and most projects don't have that. This is still a pretty conceptual stage, so I wouldn't go expecting it to pop up tomorrow.

[A]

So why is Sohn trying to do.

[D]

This at a broad level? He gets really excited when markets get really hot in tech. He also, for the past 10, 15 years, has been really excited about AI, constantly seeing it as like the AI revolution on the Horiz, and he's sort of been punished by the market by wasting a lot of money prematurely making that bet. But now the stars seem to be aligning in that the market has also embraced this idea that there's an AI revolution imminently. And so his stock has tripled just since April. A lot of his investments are going up in value, and so he sees it as a really good time to make a huge bet again, something on a big scale.

[A]

That was WSJ reporter Elliot Brown. And that's it for Tech News Briefing. If you're a listener on Spotify, be sure to take this episode's poll or leave us a comment. Today's show was produced by Julie Chang. I'm your host. Patrick Coffey. Jessica Fenton and Michael Lavalle wrote our theme music. Our supervising producer is Katie Ferguson. Jessica Fenton is our technical manager. Our development producer is Aisha Al Muslim. Chris Zinsley is the deputy editor and Falana Patterson is the Wall Street Journal's head of News Audio. We'll be back later this morning with TNB Tech Minute. Thanks for listening.

[C]

So many organizations choose outsystems because it's an outstanding way to quickly deploy apps and AI agents and deliver results. A top US bank deployed apps for their customers to easily open new accounts on any device. We helped a leading global insurer quickly deliver a portal and app for their employees, while a global brewer developed an app to automate tasks to clear Bottlenecks. OutSystems, the 1 AI powered low code platform.