jakke

Policing in big blue cities in the reactionary backlash era has been incoherent. To summarize:

This police chief was a holdover from the previous mayor. He didn't really live in Seattle full-time and he'd kind of become more of a frequent podcast guest than a particularly active or visible bureaucrat. No one understands why the mayor had kept him around when they disagreed on everything.

Then there was a highly visible public mass shooting. The remote-work police chief was out of town, and news about the situation came out slowly and awkwardly, and he didn't seem to know how to respond. When he made it back to Seattle, a reporter for a neighbourhood news website asked him about his travel plans at a press conference, and he could fully have ignored the question, but he decided to explain that no one needs to know whether the police chief is planning to be in Seattle at all this summer. In the video, a staffer guides him out basically saying "shhh, shhh, it's okay" that suggests that this staffer has had to play the same role before. At this point, the travel schedule was a bigger news story than the highly visible public mass shooting, which was wild, because several people died and a toddler was shot and at least one shooter is still at large.

The mayor tried to have him resign, and then when he didn't resign she fired him, but for some reason tried to do so silently, because the mayor is not great at talking about police-related topics. Firing the police chief of a large city in private is not a thing, so a press release and then eventually a video did dribble out, but not until a ton of confusion had happened.

Various parts of the local Democratic Party machine are very upset at how this all happened. Not at the whole mess of the police chief's weird defensiveness about his own travel schedule in the wake of a highly visible public mass shooting, nor at the mayor's weird attempts at a secret midnight firing, but at the lack of personal consultation from the mayor. This is candidly a little confusing, because once a police chief's frequent flier miles become a larger story than a shooter at large, there is no conceivable way that police chief is remaining employed.

Because the mayor's office has put in zero effort at replacing the most powerful bureaucrat in the whole municipal apparatus, whom she's kept on the payroll despite their many public disagreements and his increasingly public itinerant lifestyle, the new police chief is the deputy who always filled in for the previous police chief while he was travelling, who agrees with the previous police chief and disagrees with the mayor on pretty much everything.

A big part of what's always fascinated me about real estate development is the huge grinding friction between the masters of the universe sitting in plush boardrooms in Manhattan and Palo Alto pointing at PowerPoint slides and viewing themselves as "capital allocators" when they make big abstract decisions on one hand and the actual messy confusing process of actually allocating resources to projects by physically attaching big complicated metal things to the ground on the other. Most financial decisions in 2026 really do match that abstract model where rich powerful people can say the words and then their minions scurry off and do the thing, because transportation is cheap and money can freely move between countries and people can be hired and fired in huge numbers without regulations or unions slowing down the process. But actually physically building things remains really complicated, partially because reshaping land is still really expensive and partially because local land-use regulations have gotten stricter even as macro-scale labour and capital regulations have gotten weaker, and this is going to pose a significant obstacle to the massive data centre buildout, because the new electricity demand necessarily means a ton of new power plants and transmission lines.

Such documents, which were reviewed by Futurism by way of Google Search, included what appeared to be a detailed medical report of a real patient, clinical trial results that included patient names, documents sharing the names and phone numbers of primary school-aged children, company documents marked for internal use only, and employee reviews that included personal information about workers. (We’re opting not to share any links to the content out of concern for users’ privacy.)

The charitable interpretation is that people want to compare the modern tech industry to the late 90s. And I get that temptation, because there's this wild optimism about technology that isn't really matched in reality, and companies are issuing a ton of debt to pay for big capital improvements, and the circular capital flows between a few very large companies make it hard to tell whether any of the revenue is real. That's pretty much the same situation that did take down a bunch of tech companies when the dotcom bubble burst, and the NASDAQ lost the vast majority of its value, and I can see why this feels like that.

But the modern tech industry is so much bigger and is sitting on so much more cash and has so much more monopoly power in their long-term contracts with every big bank and retailer and manufacturer than anything that existed during the dotcom era. Companies like Google and Meta and Amazon are just able to tank tens of billions of dollars in loss in a way that no previous generation of tech companies was able to. Remember how Meta wrote off $70 billion in losses on their Metaverse project and it didn't really have any meaningful long-run effect on the company's financial performance? They're all operating on these giant cash reserves of hundreds of billions of dollars, and it's completely warped any sense of the risk-reward tradeoff that e.g. Elizabeth Warren will tell you is the good part about capitalism. They can spend years pursuing projects that wouldn't make sense as part of any conventional corporate finance spreadsheet just because the C-suite thinks it's the right way to look like a brilliant visionary in front of the other companies' C-suites, and they have so much money that a compete failure is actually totally fine.

So I used to work with a former Big Tech product management guy and he had a sign by his desk with a picture of Scrooge McDuck swimming in gold that said OPTIMIZE FOR MONEY NOT PROXIES FOR MONEY. And he was just as difficult to work with as that made him sound. But his point was that successful tech companies can't just say "oh we have so much activity" or "oh we're generating buzz" but actually need to generate revenue and eventually positive cash flow. You often see less-successful companies like Pinterest using dicier metrics like Monthly Active Users to confuse investors about their actual cash flow and generate a sense of hype. Anyway, hard not to think of that when you're looking at these earnings highlights:

For the actual usable products Google offers (search ads, cloud computing, YouTube ads) they're showing revenue numbers. For their LLM API, they're showing "tokens per minute". And for Gemini, much like Pinterest, they're focusing on Monthly Active Users.

It's really bizarre to see a huge incredibly successful tech company like Google with a bunch of really lucrative product lines using these silly fake metrics and pretending they're somehow relevant to earnings! If these products were useful, people would be displaying a willingness to pay to use them, and then Google would be generating revenue from them, and would tell us about that revenue. Telling us about their "tokens per minute" feels a lot like an airline trying to look successful by telling us how much fuel they're burning per minute.

@brostateexam Here's a very solid explainer of how this off-balance-sheet accounting works. Also some very funny comparisons in here about how the AI infrastructure buildout is like ten times the size of all major infrastructure projects plus the new Moon landing project combined. Sorry to quote an article at length, but I think this is the part that's most relevant to the accounting going on here:

To finance Meta’s Louisiana data center, dubbed Hyperion, both Meta and Blue Owl invested equity into a separate legal structure. That entity took on $27 billion in debt while another related legal entity will serve as the landlord and a Meta subsidiary will be the sole tenant of the project, according to details provided by ratings agency S&P Global. Meta determined it shouldn’t bring billions in debt from the Louisiana project onto its own balance sheet because it isn’t responsible for finding tenants to replace or join it at the nearly 4,000-acre campus — a critical job that impacts the entity’s economic performance, the social media company said in its most recent quarterly SEC filing. Meta said its role is limited to construction management, along with administrative and property management services.

(Here's the unpaywalled archive version.)

It's really hard to tell whether this is an existential threat or whether we're just entering an era where Big Tech is using what looks like creative accounting just to issue overly optimistic press releases. On paper, these "hidden debt" numbers are getting too high for these companies to survive if the AI boom doesn't work out perfectly. Like, here's the money shot from the article:

And these are big huge companies that could realistically tank $100 billion in losses and have that be an embarrassing news cycle that's bad for their stock price but ultimately wouldn't effect their ability to operate as going concerns. But as that climbs to $250 billion, or $300 billion, that's getting into the range where these companies would run into trouble. I don't think that would cause any of these companies to collapse entirely (which is fucking wild! that's so much money! especially for Meta which is pretty much just an ad company) but it would probably require some amount of restructuring, and bad news for the overall stock market, and an end to the long-term tech boom.

It's also not clear to me from this article, though, how guaranteed these various obligations really are. Like, if they've promised to lease a bunch of data centre capacity, but they can choose to terminate the lease or arrange for the data centre construction to be cancelled, that's not a real obligation, right? That's just a news release that they intend to do a thing, with no legal requirement that they follow through. The article mentions an Oracle/OpenAI collaboration called Stargate, and we know from reporting elsewhere that Stargate has made some big announcements about data centre leases and then the data centres never got built and the obligations kind of evaporated. And that's a pretty standard hype cycle in tech, and calling that "hidden debt" is way too pessimistic, because the company was never really under any obligation.

But from this article, it looks like that's not how all these agreements work? Some of them are more solid. Quoting directly from the article:

Meta has a contract guaranteeing investors' losses if the data center becomes unnecessary and the lease is terminated. The company announced on July 13 that its total investment in the facility is now expected to exceed $50 billion.

If that is now the standard these companies are operating under, that would be a whole different situation. That's a huge shift of the risk from the external investors who would historically put their money on the line to the Big Tech shareholders who basically treat their investments as equally safe to e.g. a bank. It's impossible to tell from this article whether this is becoming standard or whether most data centre obligations are still easy to opt out of.