High-cost AI is potentially bullish
And: Latest hiring data does not inspire confidence
Welcome to Cautious Optimism, a newsletter on tech, business, and power.
Happy Thursday! We’re heading into a holiday weekend here in the States, so expect news from our end to slow as the day grinds along. But to start the day the government dropped June’s non-farm payroll hiring data. The good news: Jobs created in June came in at 147,000 ahead of expectations of a gain of around 110,000. Let’s get to it. — Alex
📈 Trending Up: Datadog, soon to be part of the S&P500 … paying bribes … indie media … white-collar labor worries … data center buildout … Synthetic equity positions via tokenized SPV ownership — what could possibly go wrong? … computational cognition…
📉 Trending Down: Free press in the United States … the chance of a rate cut … USD stablecoin dominance? … CEOs being the only people with several jobs …
Hiring data doesn’t inspire confidence
The U.S. economy added more jobs than expected in June. The latest data point will be refined in the coming months, but a flat unemployment rate of 4.1% during this period of rapid economic policy change is worth celebrating.
Still, if we look into the data, there’s weakness lurking just beneath the surface. Of the 147,000 new jobs in June, 73,000 came from government employment. Federal labor rolls contracted by 7,000, while state-level employment rose by 73,000, mostly driven by education-related hiring. Healthcare employment also performed well, raining 39,000 jobs.
Elsewhere, the data was lackluster. Here’s the BLS:
Employment showed little change over the month in other major industries, including mining, quarrying, and oil and gas extraction; construction; manufacturing; wholesale trade; retail trade; transportation and warehousing; information; financial activities; professional and business services; leisure and hospitality; and other services.
Apart from hiring more folks to teach our kids and keep us healthy, the economy was pretty static in June. Cross that with recent headlines, and I wonder how strong the H2 2025 labor market will prove.
Microsoft is a good example of changing workplace norms. The famed technology giant is executing yet another cull, this time impacting around 9,000 staffers. It’s the third cut at Microsoft this year, following several trims in 2024 and a huge cut back in 2023. All the while, Microsoft has been performing well as a company, with rising revenues and profits.
I’m a capitalist, yes, but I also recall mass layoffs at companies arriving when corporate performance suffered, cash flows were weak, or demand for produced goods and services falling. None of those scenarios applies to Microsoft; indeed, they are inverted.
You can point a finger at AI, and, yes, some workers in the economy today are being squeezed out today by new tooling making the single worker more productive. But there’s also a mindset shift regarding hiring that we need to consider, at least in technology:
In the past, hiring lots of smart people was considered a win; more smart people meant more smart work, and therefore tidy cash flows in time.
Today, hiring lots of smart people is considered a loss; not only do you have more mouths to feed, but the more staff you have the more layers between the CEO and the trenches the slower you can move, and the more decisions that are made — by necessity — outside the c-suite.
Perhaps this is the new normal and we should all anticipate seeing some of the most valuable and profitable companies in the American economy work every quarter to find staff that they can jettison. If so, the already-threadbare social contract between employers and their staff could degrade to farce.
Jason Lemkin of SaaStr Fund told CO yesterday that “every CEO” he knows “wants to do the same” as Microsoft “at scale.” Why? Because they do not think, Lemkin continued, that “30%+ of [their] team is going to add value in Age of AI.”
Notable Capital’s Jeff Richards has a good look at the trend in historical terms.
Corollary: Why do companies want to reduce the size of their employee base?
Because humans are expensive. Yesterday on the podcast, Felicis’ Astasia Myers made a good point concerning why the upside for AI in the tech sector is so potentially large. From our own transcript:
Here at Felicis, we really think [that AI startups] can go after labor budgets now. This is a $35 trillion market. If you look at the Nasdaq top 10% companies or the Fortune 500 budgets, only 10 to 15% [of their spend] is actually software. So the markets that these AI companies are playing in are larger than ever before. […]
If you look at AI software products and infrastructure increasing the S&P500’s net gross margins by 5%, that's essentially like $600 billion a year in profits. And you can imagine software can go after 20 to 30% of that.
Her argument, which I think is economically correct if a little worrying in kitchen-table terms, is that AI tools will be able to attack a far larger pool of corporate spend than just IT budgets. They are going after payroll spend. And that’s why VCs are betting that the fastest-growing AI startups will be worth a mint.
High-cost AI is potentially bullish
After Anysphere’s Cursor introduced a new, more expensive service tier, Perplexity followed suit. You can now spend $200 per month on the AI search app (a multiple of its lower priced, more consumer-friendly offering).
What do you get for all that scratch? Unlimited access to Perplexity’s Labs feature (“craft everything from reports and spreadsheets to dashboards and simple web apps”), first access to Perplexity’s upcoming browser (Comet), access to more powerful AI models, and “even more premium data sources and benefits from leading brands exclusive” to the new Max offering.
You can spend a lot of money on AI today:
Cursor: Up to $200 monthly
Perplexity: Up to $200 monthly
OpenAI: Up to $200 monthly
Google AI: Up to $250 monthly
Anthropic: Up to $100 monthly
You get the idea. Initially AI tools were priced like consumer SaaS. You got a lot for $10 or $15 per month. Despite falling inference costs — shoutout Nvidia — increasingly large and complex models found ways to consume ever more compute. That means that folks who want to use the most powerful models were effectively rate-limited by how much they paid for access to AI services. Enter the second generation of AI subscription pricing — $200 each month for the best stuff, with fewer or no rate limits.
The bad news is that it appears that AI’s era of dramatically undercharging for what was offered is likely fading.
The good news is that a good number of AI companies appear to think that they are offering enough value to charge their most active and demanding users a lot more. If the above service tiers prove popular, AI will have proved to the market that its TAM truly is enormous.
Media economics, part 47
Up top we linked to news that The Bulwark has reached 100,000 paying subscribers. It’s a huge feat, and one worth celebrating.
Elsewhere in digital media, things are looking somewhat grim. Here’s TechCrunch covering a recent dataset from SimilarWeb:
Since the launch of Google’s AI Overviews in May 2024, the firm found that the number of news searches on the web that result in no click-throughs to news websites has grown from 56% to nearly 69% as of May 2025.
Not surprisingly, organic traffic has also declined, dropping from over 2.3 billion visits at its peak in mid-2024 to now under 1.7 billion.
Google, perhaps in response, is putting in a little bit more effort to support publishers, The Information reports, including hiring more sales staff for Google Ad Manager.
On one hand, Google is closing the key avenue for the average person to reach online information, while also mining that information to power its AI models; on the other, Google Ad Manager will have some more staff. Which side of that equation do you think carries more weight?
Correct! The answer is Cloudflare and Fastly, which are making attempts to rewire the web to force a shift in economic value destination, away from aggregators and towards first-party sites. At this point, as a person who writes online, I fully expect Google to make my life worse to make its life better. And the search giant is only making me more confident through its net actions over time.