Every gambler knows that the secret to survivin’
Is knowin’ what to throw away and knowing what to keep
‘Cause every hand’s a winner and every hand’s a loser
And the best that you can hope for is to die in your sleep

Kenny Rogers, The Gambler.

I don’t know Masayoshi Son, and I don’t need to. But my guess is that the SoftBank founder and CEO can’t be having a good start to his week. The stock of his flagship, SoftBank, is deflating faster than a balloon stuck in powerlines after a New Year’s party. He has bet big on OpenAI as his all-in wager. Whether he is right or wrong remains to be seen.

SoftBank’s shares dropped as much as 12.5 percent. Reports emerged that OpenAI and Oracle had scrapped plans to expand a flagship data center in Texas. Bloomberg reported the expansion fell apart over a combination of financing difficulties and shifting demand. The news undermines two core assumptions behind the entire Stargate Project. That alone should be reason to dig into every data center announcement and follow the money trail.

SoftBank’s credit default swaps widened. In plain English, the bond market is now charging more to insure against the possibility that SoftBank cannot pay its debts. The people who lend SoftBank money are getting nervous. Not a big surprise.

S&P, the credit rating agency, which already had SoftBank at junk, cut its outlook to negative earlier this month. A downgrade could raise borrowing costs precisely when SoftBank needs to borrow more than ever. These are legitimate problems. It also makes you wonder how SoftBank will meet its commitments to OpenAI in April 2026. I wrote about this last week, but even I didn’t know how much was up in the air.

I often refer back to my essay, The Announcement Economy, mostly because that is what we are living in. The details get lost in the bombast of headlines, and the media herd moves on to the next thing. Meme is the news, if news itself isn’t the news.

Let’s turn back the clock to January 2025. Stargate was announced with pomp and show that could rival Mardi Gras. President at the podium, Son and OpenAI’s Sam Altman flanking him. Big words, bigger promises. The beautiful future. Elon Musk, seemingly a sore loser for not being invited to the party, didn’t like it a bit. He said SoftBank “has well under $10 billion secured.” Altman fired back, called him wrong, and invited him to the Texas site. That Texas site is now the one being abandoned. Musk had obvious self-interest in undermining the project. He was also, turns out, right.

This is a good time to ponder the overall data center announcement frenzy. There is a level of insanity in all the news that keeps coming, and it deserves its own moment of skepticism.

I just read this New York Times piece about Nscale, a UK-based data center startup founded in 2024 that just raised $2 billion at a $14.6 billion valuation, with Nvidia backing and Sheryl Sandberg, Nick Clegg, and Susan Decker joining the board. The company barely existed two years ago. Its founder previously sold health supplements and worked in coal mining. Now he’s building “the engine of superintelligence.” Haven’t we seen this movie before? Different cast, similar theme with a charming, charismatic fundraiser and a board with more white shoes than an NBA star. Read the story and decide for yourself.

The irony of the name should not be lost on anyone. In model railroading, N-scale means 160 times smaller than the real thing. The real thing being a hyperscaler like Google, whose CEO told investors the company plans to spend up to $185 billion on infrastructure in 2026.

The data center buildout echoes the late 1990s fiber frenzy. Companies strung cable across continents for bandwidth nobody needed, on the assumption that demand would only go up and that the underlying technology would not change fast enough to matter. Both assumptions proved wrong. The chips will get faster. The models will learn to do more with less. The Excel chart pointing up and to the right rarely survives contact with Moore’s Law.

These are physical products, as much as they might seem like digital ones. You can’t put up walls faster than a query on ChatGPT. Neither can you get energy sources revved up on demand. Physics is physics, and atoms are atoms. It just doesn’t make for a good announcement.

Nscale is a UK company. Like everyone else, the UK is feeling left out of the AI buildout frenzy. Just like they did in the fiber buildout days. Don’t be surprised if you see more such announcements about upstarts raising billions from other parts of the world. No one wants to be left out of the announcement party.

The scale of the announcements is staggering. US data center construction starts hit $77.7 billion in 2025, a 190 percent increase over the prior year. The four largest hyperscalers are on track to spend north of $500 billion on infrastructure in 2026 alone. But announcements and reality are two different things.

The numbers bear this out. I recently pointed to a report by Sightline Climate that is tracking 190 gigawatts across 777 large data centers announced since 2024. As I noted in a recent post, of the 16 gigawatts slated to come online this year, only 5 gigawatts is actually under construction. Last year, 26 percent of expected capacity slipped. Sightline’s estimate is that 30 to 50 percent of the 2026 pipeline won’t materialize. Look, if hyperscalers are building their own data centers, we know they can. They have the money. They have customers. They have the domain expertise. They can even spin up their own power sources. Others feel more like Milli Vanilli. They sound good. But is it for real?

I am an AI believer. But boy, the green gas coming out of the announcement engine makes me blanch.

If you see skepticism in my recent writing about physical infrastructure, it’s because sometimes you have to follow the dollars. In my earlier pieces on the $110 billion funding round and the announcement economy, I laid out how the headline numbers didn’t really add up. The structure hasn’t changed. SoftBank is seeking a new $40 billion loan, the largest dollar-denominated borrowing in its history, to meet its OpenAI obligations. Son is doubling down.

To fund the earlier rounds, Son sold SoftBank’s entire Nvidia stake for $3.3 billion. Those shares would be worth well over $150 billion today. He traded one of the great unrealized gains in modern investing history for a 13 percent stake in a company that remains unprofitable, whose flagship data center partner just backed out citing weak demand, and which S&P now considers a liability on SoftBank’s own balance sheet.(1)

Son may still be right. Yahoo Japan, Alibaba, ARM. He has been early and right before. He has also been early and spectacularly wrong — see WeWork. But the structure underneath, borrowed money, illiquid assets, a portfolio more than half locked up, means the margin for error is far thinner than any announcement headline ever suggested.

It isn’t money until it’s money. And it isn’t infrastructure until someone actually needs it. Uses it. And pays for it. As Kenny Rogers put it:

You got to know when to hold ’em, know when to fold ’em
Know when to walk away and know when to run
You never count your money when you’re sittin’ at the table
There’ll be time enough for countin’ when the dealing’s done

Footnote #1 (added on 3/10/26): I can see how my awkward sentence gives an impression that Son’s Nvidia sale funding OpenAI. The point I was making was that he has hyperactive style of a a gambler, and this is how he has ended up with OpenAI, and not with $150 billion if he knew how to hold them. Clearly my sentence structure could have been better.

Why I wrote this piece

• SoftBank’s bet on OpenAI is starting to weigh SoftBank’s shares have nearly halved in four months as questions grow over the scale of its OpenAI exposure. The FT’s Lex column has some strong words. (FT)
• SoftBank Seeks Record Loan of Up to $40 Billion for OpenAI Stake The largest dollar-denominated borrowing in SoftBank’s history, a 12-month bridge loan underwritten by JPMorgan and others, to fund commitments already made. (Bloomberg)
• The AI Data Center Boom’s Newest Billionaire Nscale, founded in 2024 by a former health supplements seller and coal miner, just raised $2 billion at a $14.6 billion valuation with Sheryl Sandberg, Nick Clegg, and Susan Decker on the board. The Times loves an outsider-looking-in story. (New York Times)

March 9, 2026

Pop some popcorn. Put some butter. Add some salt. Because opportunists (politicians) are pointing their muskets at villains (tech bros), using children’s welfare as the ammunition. In case you were wondering, I am talking about the battle between New Mexico AG and Silicon Valley’s villain in chief.

The next bout is on May 4. So mark your calendars. Why?

This week two verdicts came in quick succession. First, a New Mexico jury ordered Meta to pay $375 million for knowingly enabling child predators on Instagram and Facebook. Then, a Los Angeles jury found Meta and YouTube negligent for designing platforms that addicted a young woman who first used YouTube at age six and Instagram at nine.

On the surface this is big win for ambulance chasers. However it could be much bigger if politicians actually have the best intentions that go beyond winning the next elections. History tells me, they are what I said they are, opportunists.

Let’s be honest about what New Mexico AG is. He is an elected official who identified the most despised man in American technology, a category that has never been more unpopular, and filed suit. Zuckerberg is the perfect patsy. Ian Fleming couldn’t have created a more perfect caricature of a villain for our post social age. Weird. Awkwardly dressed to appear cool. Black AR glasses.

Rich, arrogant, visibly indifferent to the damage his products caused, and now on record in two courtrooms defending decisions his own safety teams told him were harming children. The internal documents that came out in both trials showed the same pattern: engineers and safety researchers raising alarms for years, recommendations made and ignored, Zuckerberg choosing growth. None of this is new. I wrote about Facebook’s growth-at-any-cost DNA back in 2018. The pattern has not changed. Only the courtroom has.

New Mexico AG smelled blood. A politician at the end of the day, doing what politicians do. Scoring points against the current villain in chief, building a reputation on the wreckage of someone else’s failure, and calling it justice. The “independent monitor” he is asking for is the tell. And that is not how change happens. That is a patronage job dressed as accountability. That said, underneath the political theater, the structural demands are real. And if a judge grants even a portion of them, they could change daily life for two billion people.

Think about what your daily experience on Instagram or Facebook or TikTok or YouTube actually is. You think you made some choices. Pat yourself on the back. You have been fooled into thinking that you were the one making the choices. Now, let’s get real.

You open the app and a feed appears. Like I did. This morning. I went to Instagram to see what my “pen friends” were doing. I didn’t see a damn thing. What I saw, I didn’t choose. Neither do you. You see what they want you to see.

An algorithm shows you what it wants to show you, trained on everything you have ever paused on, liked, or watched past the halfway point. It is the ultimate illusion, where perception is conflated for reality. It has one job: keep you there. The most reliable way to do that, as the internal research showed and as the companies knew, is to surface content that provokes anxiety, outrage, desire, or envy. Those are the emotions that generate engagement. Engagement brings in the moolah. The loop is the product. I called this trap back in 2018. The feeds have only gotten more sophisticated since.

Layer in the design features built around that loop. Infinite scroll, so there is never a natural stopping point. Autoplay, so the next thing starts before you have decided you want it. Notifications calibrated to pull you back at the moment your attention drifts elsewhere. Variable reward, the unpredictable appearance of a like, a comment, a new follower, built like a slot machine. None of this happened by accident. All of it was tested, measured, and optimized.

This is what the courts are now calling a design defect. And to understand why that matters, you need to understand the wall that has protected these companies for thirty years.

Section 230 of the Communications Decency Act, passed in 1996, is the founding legal pillar of the modern internet. It says platforms cannot be held liable for content posted by their users. It is why Facebook is not responsible for a defamatory post, why YouTube is not liable for a radicalization video, why the whole architecture of user-generated content was able to exist at scale. Every time someone tried to sue a social platform for harms caused by what users posted, Section 230 was the wall. Case dismissed.

The Los Angeles plaintiffs found a door in that wall. The jurors were specifically instructed not to consider the content Kaley saw on the platforms. Not because it was irrelevant to her harm, but because content is where Section 230 lives. As NPR reported, the lawyers pursued a case of defective design precisely to get around the high bar set by Section 230.

The entire case was built around the design. Infinite scroll, autoplay, algorithmic feeds, notification systems. Those are not user posts. Those are the company’s own engineering decisions, made in their labs, tested on their servers, optimized by their employees. CBS News put it plainly: this case centered around how the apps are designed, not the content itself. Section 230 does not protect you from your own product choices.

The jury answered yes to seven specific questions per defendant. Were they negligent in the design or operation of their platforms? Was that negligence a substantial factor in causing harm? Did they fail to adequately warn users? And did they act with malice, oppression, or fraud? That last finding triggered punitive damages.

The pipe (aka the Social Media platform), not what flows through it. That argument won and opened new legal ground that can change everything that comes next.

If the design is the defect, the remedy is redesign. Not a warning label. Not a terms of service update. The product itself has to change.

Not surprisingly, I have some suggestions. Actually pretty simple ones that even politicians can grok.

Start with defaults. Right now every addictive feature ships on by default. Flip that. Ship the product in its least addictive form. Chronological feed. No autoplay. Scroll that ends. Notifications off until you choose otherwise. Every addictive feature requires an explicit, informed choice to activate.

That is consent architecture. And it is exactly what these courts have been asking. Did you warn users? Did you get meaningful consent? The answer, established now by two juries, is no.

Real age verification is the other demand with teeth. But it is also a hairball that will unwind a lot of the internet, security and privacy. That is for experts with more gravitas and knowledge than me.

Instagram’s Adam Mosseri knows what is coming. His year-end memo was full of language about authenticity, credibility signals, and platform responsibility. The same language is now showing up in court filings. At trial, he testified that there is always a tradeoff between “safety and speech.” That is a telling formulation. Safety as a constraint on the product, not a foundation of it. When I pushed on what Instagram is really positioning itself to become, he pushed back. But the architecture he described, who posts rather than what is posted, is exactly the consent architecture these courts are now demanding.

The practical problem for the platforms is simple. You cannot run two different algorithmic systems in one product without everyone noticing. Once you build the less addictive version for minors, the question becomes impossible to avoid. Why is that version not the default for everyone? Why is the more addictive version what adults get, when they never consented to it either?

The legal fight is about children. The logic does not stop at eighteen.

The tobacco parallel is everywhere in the coverage this week. It is the right one, but people are using it loosely. The important part of the tobacco story was not the $206 billion settlement. What changed behavior was what came alongside the money: mandatory disclosure of internal research, advertising restrictions, forced redesign. Joe Camel did not die because of a fine. Joe Camel died because a court said you cannot market this product to children, and you cannot keep making it more addictive than it already is.

The companies held out for decades. Denied the science. Attacked the researchers. Then they did not.

But here is where the parallel gets uncomfortable. Tobacco did not go away. It became Juul. I wrote about Juul back in 2018, when Silicon Valley was celebrating a $15 billion e-cigarette company as a tech startup. I called it Camel 2.0 then. Same addiction, new delivery mechanism, new investors, new PR. As a former smoker who nearly lost his life to cigarettes, I was livid. Tobacco companies figured out how to sell the same addiction in a cuter package, and Silicon Valley helped them do it.

Zuckerberg will do the same. He is too good at survival not to. The feeds may change form. The addiction machine will find a new name. Addiction is a lucrative business. It always finds a way.

Meta and YouTube have already announced appeals. They have the money and the lawyers to fight this for years. And they will. But discovery in each new case keeps producing documents. The internal research keeps surfacing. The bill keeps growing. Eventually it is cheaper to change than to fight.

Zuckerberg’s doctrine is simple: two steps forward, one step back, and make it appear as if he is changing. His superintelligence memo last July was the latest example. Talk about the future, avoid the present, and let the lawyers handle the rest.

That is the logic of the crowbar.

All eyes on Santa Fe on May 4. There just might be enough structural change. And if consent architecture is put in place, then we can all feel the glow, for a moment. Billions of people deserve a different default. Till the battle starts again. Like Juul.

March 25, 2026, San Francisco

PS: I want to apologize for sending two long emails in a single day. I had no intention of doing so. I was definitely not going to write about Meta or social media. I was thrilled to just nerd out about broadband today. However, I can’t help myself when it comes to Facebook and Zuckerberg and their shenanigans. I have been covering them since they were a tiny startup.

I have been a WebPass customer for years. Fast, reliable, founder-run. When there was a problem, you could reach someone who gave a damn. It was the kind of internet service that made you forget you were dealing with a utility.

Then Google bought it.

That should have been the warning sign. Google buys things it should have no business touching. It launches products without a plan. Its decisions to get into new markets are a map of some mediocre executive’s ambition.

I watched the service slowly get worse. More outages, stagnant speeds, the kind of indifference that sets in when a product becomes a line item rather than a mission. Like it was for Charles Barr, founder of WebPass. When Barr was running the show, the speeds went from 50 Mbps to 100 Mbps to 200 Mbps. All for less than $50 a month. Eventually the speed went to a gig per second. It was one of the fastest in the country.

Since then, as competitors like Sonic were pushing 10 Gbps across San Francisco, my WebPass connection sat frozen at 1 Gbps. The founder used to pick up the phone. Now I was just another entry in a spreadsheet at Alphabet, buried inside “Other Bets,” a division whose name tells you everything about how seriously Google took it.

Google is a company where nobody has real product conviction anymore, just bonus targets and quarterly metrics. It wears a veneer of innovation, but in reality it is a financially optimized extraction machine run the McKinsey way.

Anyway, now comes the news I didn’t want to hear, but knew was going to happen anyway.

Google has sold the whole thing, Google Fiber (including WebPass), to Astound Broadband, owned by Stonepeak, a private equity group. They are calling it a “merger.” It is not a merger. A merger implies two equals coming together to build something. This is a flip. Google wanted off the hook for a money-losing division, and they found a buyer. Customers are just assets in the transaction, transferred like furniture.

Here is the most telling detail. So far, Alphabet has not filed an 8-K with the SEC as part of the announcement. An 8-K is what public companies file when something material happens. Google’s lawyers apparently concluded that losing your internet provider did not rise to that level. GFiber sat inside “Other Bets,” a segment that in 2025 generated $1.54 billion in revenue across everything Alphabet deemed non-core, while losing $16.8 billion. GFiber itself was less than half a percent of Alphabet’s total sales. To their investors, this was a rounding error. To you, it’s your internet connection. That gap tells you everything about where you stood in the relationship.

Then there is Dinni Jain, the CEO of GFiber who will now lead the combined company. His story is presented as reassurance, a steady hand, a cable industry veteran, a man with a Vermont farm who talks about patience and humility. A banker turned cable guy turned Google, and humility. Lolz.

But look at the actual career. He started at NTL, a UK cable company that filed Chapter 11 in New York in 2002. At the time, its $10.6 billion bond default was the largest in US corporate history. He left a year before the collapse. Moved to Insight Communications, a cable operator largely owned by the Carlyle Group, which Time Warner Cable bought for $3 billion in 2012. He became COO of Time Warner Cable, which in turn was acquired by Charter. Every company he has run has been consolidated, absorbed, or flipped. He made enough money to retire to a farm in Vermont in search of humility. And pigs have wings too.

He was hired as Google’s third CEO “out of retirement” to run the fiber unit in 2018.

To stabilize a business in freefall.

He did real work. Under Jain, GFiber cut support calls in half and moved to ten-minute technician appointment windows. In 2024, he hired GFiber’s first-ever CFO. That tomTomed the real plan. You don’t hire a CFO to run an internet service. You hire a CFO when you are preparing to sell. He stabilized GFiber well enough to make it sellable. And now it is being sold.

Let me tell you what Astound actually is, because the name was invented to obscure it. And while we are on the subject of names, notice that Google rebranded “Google Fiber” to “GFiber” before this deal closed. Scrub the Google brand, hide the Astound connection, give customers nothing to search for. That is a classic private equity move. Polish the turd before you sell it.

The only reason I know about Astound is because when I lived in New York I signed up for an upstart called RCN. It wanted to take on Time Warner Cable. And Verizon. And AT&T. I like upstarts. My whole life is betting on the underdog. Monopolies suck, destroy innovation, and throttle progress. RCN convinced me that my apartment needed them just by existing. Sucker, that was me, because in a few months I found that what seemed like startup gold was spray tan on a tin can.

I moved to San Francisco. And because I am like a jilted lover, I kept tabs on that shit show. Being a broadband nerd made it simple. So now you know why I am not “astounded” by Astound. I know the genesis.

In February 2017, private equity firm TPG bought RCN, a cable operator in New York, Boston, Chicago, and DC, for $1.6 billion, and simultaneously bought Grande Communications in Texas for $650 million. A year later, TPG added Wave Broadband on the West Coast for $2.36 billion. Then came the bolt-ons. enTouch in Texas, Digital West in California, WOW!’s Chicago and Maryland markets for another $661 million, Harris Broadband in central Texas. In January 2022, they stapled all of it together under a new name, Astound. The name came from a small Bay Area provider Wave had previously acquired.

In August 2021, before the rebrand, Stonepeak bought the entire platform from TPG for $8.1 billion, $3.6 billion in equity plus $4.5 billion in assumed debt. That is a leveraged buyout. The debt sits on the balance sheet and has to be serviced from operating cash flow, which means it has to be extracted from customers.

Last July, Astound had to refinance. Stonepeak injected another $400 million, pushing debt maturities out to 2029 and 2030. They called it good news. What it actually was: a company under pressure, buying itself more time.

Now add GFiber and WebPass to that debt stack.

The playbook is not complicated. Low introductory pricing, step-function rate increases after twelve months, retention discounts for customers who call to cancel, and underinvestment in everything that doesn’t show up on a cash flow statement. The BBB and Trustpilot reviews are a graveyard of billing complaints and broken promises.

Astound makes the old Comcast look almost saintly. At least with Comcast you knew what you were getting. Astound is as dodgy as there are angles in a tapeworm. Google Fiber’s reviews on Yelp are already failing grades on billing, customer service, support, and performance. Under a PE regime, it is only going to get worse.

Game is the game.

And debt is the game PE plays. Stonepeak has billions in debt to service. That debt requires predictable and increasing cash flow. The easiest way to get more cash flow from a captive broadband customer is to raise prices. There is no mystery here.

What angers me most is the language. The GFiber email I received, and I suspect you got one too if you’re a customer, promised that “nothing is changing about your service.” Not the speed. Not the price. Not the “extraordinary customer experience.” This is the kind of sentence that only gets written when everything is about to change. Companies actually keeping their promises don’t need to make them this explicitly.

“Nothing is changing” is corporate for: we need you to stay through the deal close.

The regulatory picture offers little comfort. The FCC under Chairman Brendan Carr views consolidation favorably. The current commission is not going to save you. California’s Public Utilities Commission is the one real lever. The CPUC has actual teeth. It forced Verizon to make concrete commitments before approving the Frontier acquisition.

Citizens can engage that process. Advocacy groups can push for price locks, service standards, build-out requirements. It won’t block the deal, but it might constrain the worst impulses of a PE-owned operator trying to wring cash from a captive customer base. Given the demographic profile of WebPass customers, none of the consumers are going to complain. They will just pay-up.

Here is the worst part. There is a better option in San Francisco. Sonic.net. And no, it’s not available in my building. A lot of tech people are soon going to find out that Google has sold their internet experience to a terrible company, with a patchwork infrastructure and a poor customer track record.

Let me be specific about that infrastructure. Astound is not one network. It is five or six different networks, built at different times, by different operators, using different equipment, stitched together under a single brand with a management layer on top. As of today, 86% of Astound’s footprint is still HFC, the old hybrid fiber/coax cable architecture from the 1990s. Only 14% is genuine fiber-to-the-home.

Their broadband upgrade path runs through a Harmonic virtualized platform that can theoretically be migrated to fiber someday. Notice the word “theoretically.” Their mobile service rides T-Mobile’s network. Their TV runs on TiVo, distributed through a DirecTV streaming partnership. Their WiFi hardware is Amazon eero. None of these are bad choices individually. Together they describe a company that has outsourced every layer of its stack to someone else.

Google Fiber changes the mix somewhat. GFiber brings about 2.8 million locations passed across 15 states, approaching 1 million subscribers, all on genuine fiber. Markets include Kansas City, Austin, Atlanta, Nashville, Salt Lake City, and a recent expansion into Las Vegas. WebPass adds fixed wireless and ethernet connections in dense urban markets. That is a real fiber asset being dropped into a predominantly HFC platform. Whether the combined company upgrades the HFC side to match, or lets the GFiber advantage erode to fund debt service, is the question that determines what kind of company this becomes. History suggests the latter.

The founding promise of WebPass was simple. Fast internet, honest pricing, people who answer the phone. Google broke that promise through neglect. Astound will break it by design. These are two different kinds of failure, but the customer ends up in the same place.

In the shit.

March 27, 2026, San Francisco