So the latest buzzword — or buzzphrase — is that “Compute is an investable asset class”.
It’s a line that Nvidia CEO Jensen Huang is touting and it’s throwing up all kinds of red flags for me. Firstly, I’m reminded of very similar wording around crypto near the top, just at the time the Wall Street investment banks piled in. It was ultimately a huge bait-and-switch as mom & pop investors were lured in to be exit liquidity for early investors.
Now, we have Huang announcing a partnership with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR in what’s basically the formation of the Avengers for capital raising in order to get $500 billion to build out AI infrastructure (with NVDA chips of course).
Again, the announcement is filled with buzzwords and the aim is to “establish independent financing platforms” of “third party capital”.
To put this all into plainer wording, they’re trying to say that compute is equivalent to financing highways, airports and bridges with the idea that if the company owning it fails, you get the chips/datacenter and lease it to someone else. It’s a pitch aimed to get into the deep pockets of pensions and insurance companies.
One of the reasons this is such a big red flag for me is because it’s another sign that the money is running out. Venture capital has been tapped hard on AI — and did very well — but isn’t interested in financing data centers. Companies have scaled up debt where they can but it’s expensive and their CDS are rising. That’s since led to Google raising equity and just this week Intel announced the same.
So the entire essay by Huang is an effort to raise money from people who normally don’t lend for this kind of thing. It’s an attempt to reframe the risks, and mitigate them. The stakes are abundantly clear as with the announcement, which included a CNBC panel that surely had the largest ability to deploy capital of any in history: BlackRock’s Larry Fink, Goldman Sachs’ David Solomon, BlackStone’s Jon
Gray, KKR’s Waldemar Szlezak, Brookfield’s Bruce Flatt, and Nvidia CEO
Jensen Huang.
Notably, all of them went into a hard sell on “AI factories”.
Not surprisingly, the market is loving it as they’ve combined to soak pensions and take the risk away from the companies that are promising a brave new world (and will be the ones that profit from it). The bamboozle they’re trying to pull is that a GPU is no longer an asset with a quickly-depreciating 3-year asset life as the next generation makes them obsolete but the equivalent of a power plant.
Moves today:
The brazen contradiction here is assuming the long useful life of these assets goes directly against the dominant investment theme in markets right now: That recursive self-improvement is coming via AI and will be deployed in chip design (and everything else). So we value Anthropic at $1 trillion plus because it’s going to change everything but it won’t find ways to do inference cheaper or design better chips?
That folds into my #1 sale red flag around AI, the idea that it’s going to cure diseases. Whenever one of the AI titans writes anything or describes the use case of AI, one of the first things that comes up — usually the first — is that it’s going to cure diseases.
Just yesterday, Mark Zuckerberg leaned into it writing:
Invention, not automation, will be the greatest contribution of
superintelligence. Early AI could answer questions and do routine work.
Soon it will increasingly help discover new knowledge — ranging from
discovering new drugs to cure a family member’s disease to finding new
ways to improve your business.
Dario Amodei has been selling this idea for years.
Surely you’ve heard a version of this before. The thing is, no one touting any of it has ever developed a drug or cured anyting. They haven’t the slightest idea how long drug development takes. To me, it’s like them saying that AI will help them find literal gold mines. The thing that any mining investor will tell you is that you don’t actually ‘discover’ gold mines, you have to prove them with years of drilling. Drug development if fraught with an infinite amount of issues and necesserily long testing timelines.
The AI hype-seller may now be starting to realize this and also realize that their investments need to pay off in a shorter timeline than curing cancer. But their regulatory capture runs so deep that Zuckerberg isn’t saying that they will find cures via the traditional route, instead he’s saying they need to ‘move fast and break things’ on human drug testing.
“We should accelerate society’s ability to develop new cures and
inoculate against new issues as they arise. This includes streamlining
how the FDA and other regulators test and approve new treatments. As AI
increases the pace of drug discovery, we will need to update these
processes to keep up with the pace of innovation anyway.”
Who is he to weigh in on drug testing protocols? Have we worshipped the tech gods so much that we’re going to let them re-write the rule book on drug testing so they can justify AI spending? How about before we launch accelerated human experiments you cure just one disease with AI the traditional way?
This article was written by Adam Button at investinglive.com.