The Great Crypto Liquidity Push
And: Why I still think the U.S. will win the AI war
Welcome to Cautious Optimism, a newsletter on tech, business, and power.
Monday! Want some good news to start the week? The United States may have successfully bullied the U.K. into giving up on its efforts to force Apple to break encryption so that the nation could tap into protected consumer iCloud accounts. We have to fight the same fight every few years/months regarding encryption. Thus far, attempts to turn the math that keeps our data and online transactions secure into an insecure mess have failed. May they continue to do so.
The updated Figma IPO filing dropped after we put a lid on the newsletter today. So, we’ll do a full pricing dive of the design software giant’s $25 to $28 per share price range and roughly $16 billion valuation tomorrow morning. — Alex
📈 Trending Up: Cowardice … European M&A? … the iPhone copying Android, years later … TSMC … Supabase … video podcasting … propaganda … Angel Reese …
Round of the Day: Friend of the blog Ryan Lawler covered Eton Solution’s $58 million Series C. The software company offers tooling for family offices, including 2025-requisitie AI enhancements. Looking into Eton, it reports more than 800 families using its code and more than $76 billion worth of assets under management. Given a growing share of American wealth settling in the accounts of the 0.1%, there should be plenty of market for Eton to sell into.
📉 Trending Down: Legacy software … Musk-France relations … the free press … Apple’s technologist market share? … China-India water relations …
Remember the Google-Windsurf deal that came together after the OpenAI-Windsurf deal died, and the search giant bought out the founders of the startup, reportedly leaving the rest of the asset flapping in the wind? Well, it turns out that that’s precisely what happened. Per Jeff Wang, Windsurf’s erstwhile replacement CEO before its deal with Cognition, Google airlifted the company’s core engineering team while leaving everyone else in the lurch. On a scale of one to yuck, that’s a puke out of ten.
The Great Crypto Liquidity Push
American cryptocurrency exchange Coinbase traded down to the low $30s as 2022 came to a close and 2023 got its start. Today, Coinbase is worth $419.78 per share with a market cap over $100 billion. For the United States’ most prominent web3 name, the last few years have proved a glow-up of epic proportions.
The company is nota alone in riding a wave of growth and rising investor sentiment. After listing at $31 per share, stablecoin giant (and Coinbase partner) Circle is worth $223.78 per share today. It’s worth north of $50 billion. (eToro’s own IPO has generated less spectacular returns.)
With the value of bitcoin at an all-time high, the value of ethereum’s token finally on the ascent, and the domestic regulatory landscape coming good for the decentralized crew at last, a lot of companies want to snag their own slice of crypto enthusiasm by tapping the public markets:
Grayscale, which offers a number of crypto-focused investing products (ETFs, trusts, and some publicly traded funds), filed to go public a week back. With a yet-private IPO filing, we don’t have our hands on its guts yet, but I am desperately curious to see how low-cost crypto funds convert to business outcomes. As a big fan of low-cost investing vehicles (viva la zero-cost index fund wave), I do worry that the race to zero bips makes it hard to stay afloat. Presumably I’m flat wrong, given that Grayscale is listing.
Bullish, a crypto exchange and media company (CoinDesk) filed to go public on Friday. We’ll do a full dive into its numbers tomorrow, but when you parse through miles of data, the company is smaller than I had expected. Its S-1 includes some fascinating media notes, and indication of pricing pressure in the crypto exchange game. Regardless, Bullish wants to list because what better time is there than now?
The corollary to that point: If the IPO waters are so warm that even less-than-market-leading crypto concerns are listing, does that mean that the SaaS unicorns are still afraid, or simply unable to list at an attractive price?
Gryphon Digital Mining’s merger with the Trump Sons-backed American Bitcoin Corp is moving ahead, per an early July release. The idea is to reverse-merge American Bitcoin (a Hut 8 subsidiary) onto the public markets, subsuming Gryphon in the process. This is a weird one, but, again, shows more crypto assets and activity heading to the normal public markets.
And breaking this morning, Dynamix (SPAC) will merge “with another entity to create a new company known as the Ether Machine,” the WSJ reports. The two companies intend to “manage over $1.5 billion in ether, the largest cryptocurrency behind bitcoin.” Bitcoin treasury companies are popular in certain circles; so, why not ether treasury concerns? Especially with the value of ethereum’s token rising in recent weeks? Sure.
Update: After sending today’s newsletter to its editors, Trump Media and Technology Group announced that it has “accumulated approximately $2 billion in bitcoin and bitcoin-related securities as part of its previously announced bitcoin treasury strategy.”
I remain largely skeptical of companies trying to tie their future to the price of an asset that is easily accessible sans middleman, but I’m also boring. So, bet as you will. And expect another few shots at public capital by the crypto world, which more than ever appears to be a new rail for traditional finance with certain technological advantages (and disadvantages) that will in time become as regular and boring as whatever software powers ATMs.
Why I still think the U.S. will win the AI war
Over the weekend, OpenAI’s impressive result in the International Math Olympiad (IMO) this year was the topic of choice on tech twitter. Precisely how the company managed the feat, and if it actually won or simply did very well, kept the thumbs of folks avoiding human contact busy.
What matters more than the precise details of the win is the simple fact that AI models are still improving at a quick clip sans a new breakthrough. If you are at all exposed to AI-related assets — and if you own any index fund, you are — this is good news.
But the IMO saga was just one of several things putting a smile on Sam’s face. He also bragged that after waiting up “early on a [S]aturday,” he used a new OpenAI “model for a little coding project” that it finished in “5 minutes.” In the new dictionary definition of a humblebrag, Altman added that he was “not sure how [to] feel about” the outcome.
Sure.
To close off his news making, Sam closed off the weekend by announcing that his company will cross the one million GPU threshold this year. And that he wants to “100x” the number.
One million GPUs? 100 million GPUs? Why not ten billion? I’m not kidding; with Zuck planning datacenters that ape the outline of Manhattan, we’re heading for a world in which datacenters are the largest structures in existence.
But while leading American AI giants, hyperscale cloud providers, and neoclouds race to build the necessary global infra to power their future plans, one country is worried about overcapacity and misspent spending. Here’s the FT:
“When it comes to projects, there are a few things — artificial intelligence, computing power and new energy vehicles. Do all provinces in the country have to develop industries in these directions?” Xi told the Central Urban Work Conference, a rarely held high-level Communist party meeting on urban development. […]
Local governments across China, including in remote and sparsely populated regions such as Xinjiang and Inner Mongolia, have rushed to construct data centres in a bid to capitalise on the boom in AI usage.
But the Financial Times has reported that many of these projects suffer from a lack of technical knowhow and have wasted critical chip resources.
This is the risk of central industrial planning. China has encouraged and subsidized certain sectors in its economy. For example:
China spent tens of billions of dollars subsidizing its domestic EV industry, helping it grow from nothing to a global behemoth. Less expensive efforts to bolster EVs like generous Beijing-area license plate allocations were also employed. Now, as a result, China’s EV market is the world’s largest.
China made AI infra a “national priority” after the launch of ChatGPT, leading to hundreds of new data centers, the MIT Technology Review reports. Hundreds of startups followed in their wake, hoping to win a piece of the AI future.
China deemed domestic AI chip sovereignty a goal worth subsidizing, leading to both supportive industrial policy and a huge AI fund.
While those efforts did bolster domestic investment in key economic sectors, it doesn’t appear to have done so very efficiently. Price wars between Chinese EV companies are so fierce that they are helping drive domestic deflation, some data center capacity is sitting idle, and China is still desperate to get its hands on even dumbed-down Nvidia gear.
Meanwhile, the leading AI shops in the United States are able to secure, and lever more or less all the GPUs that they can get their hands on. Sure, supply constraints aren’t great, but if you want a sustainable AI industry, it’s better to allocate scarce resources to those best fit to put them to use than, say, another dozen data centers in the boonies of China where they will languish.
I don’t want to get too political, but as a capitalist and general skeptic of central economic planning, none of what we’re seeing from China is too surprising. Heavy spend to build capacity works well in the industrial sector until oversupply leads to a global trade environment that is becomes mercenarily mercantilist (Enter Trump 2.0). Then you have a host of unprofitable companies and falling export prices to content with.
OpenAI is perhaps the first company to the seven-figure GPU mark. It won’t be the last. But in time if current trends hold, I don’t see too many Chinese companies challenging it or its Western peers. Despite subsidy.