Are you a GENIUS?
Welcome to Cautious Optimism, a newsletter on tech, business, and power.
Happy Wednesday! The GENIUS Act passed the Senate, giving the crypto industry another win. The bill provides a comfortable regulatory framework for stablecoins here in the United States, with fewer corruption protections than Democrats wanted — the bill passed 68-30, with just two GOP defections — and I believe language restricting stablecoins from remitting earned deposit income with holders. For Circle, Stably, 1Money, and other stablecoin-focused technology companies, it’s a win. The GENIUS Act now heads to the House.
The stablecoin market is crowding quickly. Sure, the USDC-Circle-Coinbase axis is the best-known American play, but with major companies contemplating their own stablecoins for either internal or limited external use, we could see a bevy of similar digital assets rise. — Alex
📈 Trending Up: Onshoring, crypto style … AMD’s gaming chops … Middle Eastern capital flowing into China … no shit … AI for good … Whatnot … Harvey … mid-life crises … the ‘negative wealth effect’ …
📉 Trending Down: Following the law … South Korean chip giants? … corporate transparency … mortgage demand … de-escalation between Iran and Israel … EU-US trade relations …
Quick Hits
Fed Watch: Today, the Fed is expected to announce that it has no intention of cutting rates. Investors have currently priced in a zero bips cut today, and just a 14.5% chance of a cut in July. By the September Fed confab, expectations rise to about a two-thirds chance for a 25 or 50 bips cut (0.25%-0.5%). That’s far later than many want, and the labor market is showing some signs of stress. Expect mean tweets when JPow stays pat.
AI influence: Today’s you don’t say piece of reporting comes from the FT, which writes that lobbyists “acting on behalf of Amazon, Google, Microsoft and Meta are urging the Senate to enact a decade-long moratorium on individual states introducing their own efforts to legislate AI.” It’s good to confirm, but who else would be pushing for such a law? Recall that states are taking diverging approaches to AI regulation, which tech companies don’t like. States can get things done more quickly than Congress today, so if states get blocked from writing AI rules, there might simply be none passed at all. Ten years is long enough for jobs-related impacts from AI to be felt, and IP use rules sorted out by the courts, I suppose.
The slow boil: Amazon CEO Andy Jassy sent his staff a memo this week that was mostly a pep talk and salespitch for the company’s AI tools (Amazon knew that by releasing the document publicly, we’d all read it). But what made headlines was the point from Jassy that total employment at Amazon should decline over time:
As we roll out more Generative AI and agents, it should change the way our work is done. We will need fewer people doing some of the jobs that are being done today, and more people doing other types of jobs. It’s hard to know exactly where this nets out over time, but in the next few years, we expect that this will reduce our total corporate workforce as we get efficiency gains from using AI extensively across the company.
Yeah, that sounds about right. There’s concern in the market today that AI tools will cause a collapse in employment for certain roles. I doubt it. Some jobs will get replaced by AI more quickly than others, mostly we’re going to see AI make workers more productive over time, lowering the total need for human capital at major companies.
This is the take that Anthropic’s CEO is taking, more or less. And what Nvidia’s Jensen Huang disagrees with. Given my slowly rising use of AI products in my daily workflow on TWiST, I’m more Team Jassy than Team Huang on this one, while making no moral judgement on the future.
AI spend reaches the stratosphere
News that Musk’s AI company xAI is in the market to raise more equity funding and debt underscored how quickly the company has invested its earlier raised capital. As we didn’t expect that the comparatively upstart AI foundation model company was getting its compute hardware for free, we knew that xAI was spending quickly.
But perhaps not this quickly:
Elon Musk’s artificial intelligence startup xAI is burning through $1 billion a month as the cost of building its advanced AI models races ahead of the limited revenues, according to people briefed on the company’s financials.
In total, xAI expects to “burn through about $13 billion, as reflected in the company’s levered cash flow,” according to Bloomberg, which got its hands on details that were shared with investors considering putting more funds into the company.
For folks chasing wealth and status, especially in places like Vegas, there’s a riff that there’s always another level. The point of the quip is that no matter how high you may climb, someone is doing better, more, and having more fun. Enjoy what you have, in other words.
By loose analogy, when it comes to AI spend, we keep learning that there’s always another level. If you thought it was eye-popping that OpenAI burned through $5 billion last year while growing its revenues to a year-end total of around $3.7 billion.
Well, the next level of spend is $13 billion worth of burn for what Bloomberg reports is an anticipated revenue result of around $500 million this year for xAI. The AI company expects revenue to quadruple to $2 billion in 2026, and presumably it will need to eat less cash to build out its compute backbone next year than what we’ve seen thus far in 2025, but still. Wow.
What is xAI getting for its spend? The company’s founder wrote a few days back that he’s “increasingly confident that Grok 3.5 will be the smartest AI by a significant margin.” We’ll see. That’s a big claim.
Going up one more level, the outlays from indie AI foundation model companies are a fraction of what the cloud hyperscalers are spending at present. Google, Microsoft, Amazon, and Meta are pushing mountains of earned cash into compute, and it appears that there’s still room to accelerate their spend.
Maybe not on more Nvidia gear — if there is any spare chips around to buy — but certainly on humans. On a podcast appearance that TechCrunch was kind enough to transcribe, here’s how OpenAI’s Sam Altman discussed the scale of pay Meta is offering to try and poach the smaller company’s leading talent:
“[Meta has] started making these, like, giant offers to a lot of people on our team,” Sam Altman said on the podcast. “You know, like, $100 million signing bonuses, more than that [in] compensation per year […] I’m really happy that, at least so far, none of our best people have decided to take him up on that.”
I, for one, will accept any nine-figure employment offer. Zuck, call me.
Is Substack venture-backable again?
Substack is a weird duck. It raised a bunch of money back when capital was cheap and a16z was interested in investing in media companies that might wrest influence from traditional fonts. Hence the venture firm backing Substack and Clubhouse. Then Substack failed to raise another round, eventually collecting community capital and perhaps a smidgen more in a quiet round.
The money tap didn’t stop altogether. The a in a16z backed The Free Press, a center-right-ish publication built on the Substack platform, for example.
Things have changed. In February of 2024 Substack announced more than three million paid subscribers on its platform. That reached four million in November of the same year, and five million this March.
Newcomer reports that on the back of that quick growth, Substack is in the market for $50 million to $100 million at a valuation that will at least match the company’s last private mark — around the $650 million mark.
Backing its fundraise is around $450 million worth of platform GMV, and thus around $45 million worth of annual revenue for Substack itself. The company takes a 10% cut of platform spend, so the numbers make sense. And they provide an interesting loose indication of the average Substack monthly price: About $7.50, which you can quickly calculate by dividing total GMV by the number of paid subscribers, and then by the number of months in a year.
Therefore, for Substack to reach the $100 million annual recurring revenue milestone, it will need total subscriber revenue to reach $1 billion, which would require around 11.1 million paid subscribers at current prices. Not impossible, but also not a milestone that Substack should reach before 2027 at current growth rates.
That’s a long time to wait for a company to reach just a fraction of the presently needed revenue to go public. Then again, Substack almost certainly has big plans internally to grow faster, and we do not know what they are.
But why pick at the potential deal? The more money Substack has, the more shelter hacks of all sorts may be able to find in today’s failing media economy. I mean, here I am? Here we are?