The vibes are bad
Also: The return of the IoT
Welcome to Cautious Optimism, a newsletter on tech, business and power.
Wednesday! Yesterday’s news that the Chinese government planned electricity subsidies to data centers that used domestic chips was one-upped by news that the CCP proffered “guidance requiring new data centre projects that have received any state funds to only use domestically-made artificial intelligence chips.”
That’s bad news for Nvidia, at least in the near term, as well as for Chinese companies that might have wanted to access the global silicon market. To work! — Alex
📈 Trending Up: Openness on Android … MENA data centers … capex hedging … Google-Wiz … Reevo … humanoid robots … NYTimes, after earnings … soybean tariffs … parallels with Maoism … Japan-India relations
📉 Trending Down: Amazon-Perplexity relations … hyperscaler dominance … birth rates … climate progress in Europe … the Russian grid
Things That Matter
Democrats sweep elections: Yesterday, Democrats won in Virginia (governor, lieutenant governor, AG and House of Delegates), New Jersey (governor), and California (redistricting). And in New York City, Democratic Socialist Zohran Mamdani won the mayoral election.
Efforts by the city’s elite (non-pejorative) to elect former New York governor Andrew Cuomo, part of the state’s political elite (pejorative), failed. Cuomo wound up winning a single borough (Staten Island), which also happens to be its smallest.
Mamdani, in contrast, won in the Bronx, Brooklyn, Manhattan and Queens. The final tally saw Mamdani winning a majority of votes (per current counts).
Reactions vary. POTUS thinks his party lost because he wasn’t on the ballot. Folks who were mad about Mamdani before the election remain so: Lux Capital’s Josh Wolfe expects the city to descend into a crime-and-housing-cost doom spiral, and other investors are striking similar notes. (CO is broadly critical of Mamdani’s housing policies, recall.) Still, the man won, and by a large margin.
There were calls after Trump’s second win to support the president; he won, the argument went, so we as citizens should want him to succeed. Fair enough. I do demand, however, that the folks who think along those lines to apply their own logic to Mamdani. Here’s investor Bill Ackman doing just that.
What makes me worry is how completely Democrats won last night. I’m not fretting over a particular policy here, but how free and fair elections will be during the next major election cycle. If you live in the White House, are training your own shock troops, are trying to poison public confidence in elections themselves, and see yourself as much more than the head of one of three equal branches of government, you might try to turn up the heat heading into 2028.
And given the record amounts of corporate money flowing into POTUS’ coffers, it might be time to refresh our memory of the Business Plot.
Return of the IoT: Samsara had one of the best and most interesting IPOs in recent years. Despite its share price ups and downs, the company was fascinating not only because it was an impressive business from the world of venture, but because it was an IoT company.
I recall Vegas convention room keynotes back in the 2010s discussing how IoT was the future; that future never seemed to arrive in our day-to-day lives. Years later, Samsara showed that the IoT predictions were at least partially true. For corporations, at least. The company makes its bread by selling IoT hardware and software for fleets and assets, like trucks and forklifts. It’s a great business, with rising ARR and cash flow.
And now it’s not alone! Armis, which secures a company’s IoT footprint, just raised a $435 million round that it is dubbing “pre-IPO.” With more than $300 million in ARR (and thus a roughly 20x ARR multiple today), Armis says it has a path to $1 billion in ARR and is “undertaking preparations for an initial public offering.”
Hell yeah. Growth Equity at Goldman Sachs Alternatives led the round, and CapitalG put in “major” participating funding.
Yep, we’re doing data centers in space: My giddiness at Starcloud putting a modern GPU into orbit is being amplified by news that Google intends to follow suit. Yes, the search giant intends to launch “two prototype satellites by early 2027” that will “test how [its] models and TPU hardware operate in space and validate the use of optical inter-satellite links for distributed ML tasks.” (Google’s house chips are TPUs, or tensor processing units, which compete with GPUs and similar from other companies.)
Why doesn’t Alphabet have its own space program? Microsoft? Seems like an untapped opportunity.
The vibes are bad
Call me out of the loop if you must, but I had no idea that Sequoia’s Roelof Botha intended to step down from his role atop the venture capital firm, and hand the reins to two former deputies.
Given that he’s moving to an ‘advisor’ role, perhaps the man just wanted a change. A WSJ report indicates Botha got voted off the island, but it’s odd that he went on a media tour recently smack-talking venture capital’s current model before hitting the bricks.
Botha said venture capital is not an asset class, that much of venture offered return-free risk, and that the number of truly exceptional companies built each year doesn’t scale with a greater venture capital base. (He said it several times, in fact.)
The news, however, does fit into my growing opinion that the vibes are bad.
Tech sentiment has shifted from a broad ebullience around AI in 2024 to a crisis of confidence today. Both market observers and participants are worried about how quickly AI models are improving, too much investment in data centers, competition with China, inflated valuations, the viability of many ZIRP-era unicorns, and an exit market that hardly generated enough DPI in 2025 to support prior levels of venture investment.
Hell, inflation is ticking higher, the labor market is hardly upright, housing costs remain brutal on young families, insurance costs are racing higher, the government is shut down, and we’re stuck with historically high trade barriers.
Shares of SoftBank recently got clipped “amid a broader drop in AI-linked companies as investors turned wary of stretched valuations in the market’s most crowded trade,” as CNBC put it. Meanwhile, AI investment has been a critical driver of GDP this year.
Consumers are stretched (See: Chipotle, Papa Johns) as tech companies also work to shed staff like a dog ridding itself of fleas. Amazon’s cuts ran deep, IBM is planning layoffs, and it seems other tech giants can’t excise staff quickly enough.
And no, it’s not to free up cash for new GPU shipments. Tech companies expect more fruitful workers thanks to AI and are acting on that belief.
Meanwhile, freight data points to an economy in turmoil while loan defaults rise amongst consumers and companies alike. Bitcoin fell below the $100,000 mark.
There’s still plenty of good news, if you want. Mag7 performed pretty well in the third quarter, and current investments into server hardware will continue for at least a few quarters. Many startups are seeing rapid uptake of their AI-powered services and software, and the stock market is trading around all-time highs.
An eVTOL company went public; Pinterest showed growth, even if investors hated the results. Palantir’s growth was impressive, even if its shares are overvalued.
That’s why the vibes matter here. It’s the shout before the echo; the changing light that may presage a change in the weather.
It’s not that AI fears are turning into a broader sense of gloom. Instead, the AI halo is no longer bright enough to keep away the shadows rising elsewhere in the economy.
Think I’m being too harsh? Ask yourself this: With the most recent tech IPO, Navan, trading far below its offer price, how many more IPOs do you expect this year? In the first half of 2026? Three months ago, how many did you expect for those periods?
That’s what I am talking about.