The following article is part of a series on the software industry available on Semi Insider.
Palo Alto Networks (PANW) CEO Nikesh Arora made an offhand comment this week on X (the best platform to access offhand comments from important people that should not be used in investment analysis), seemingly calling for a collapse of the “neocloud” euphoria in two years:
“In two years from now you will be able to buy a neocloud for less than they raise at today. Just saying.”
The risks facing investors in neocloud businesses are abundant. We wrote about it (again) here: Updated Guide to Investing In Capital Intensive Growth Businesses – Neo-Clouds CoreWeave and Nebius
But Arora’s comment deserves a closer look, as it hints at a roadmap of what’s to come for investors, illustrates the tug-of-war between capital for infrastructure vs. capital for product and services (software), and how neoclouds and pressured software businesses (SaaS-pocalypse) can chart a path forward.
We’ll delve into all these topics in the coming weeks. But first, an explanation of what Arora meant, and a bit more detail on what would need to happen to make his statement true (our interpretation is he means buying neoclouds at a lower market cap than the cash they’re raising from financing today).
The current situation in IT
Investors often commingle technology with finance and economics. So let’s reframe what’s happening right now in IT, not as high-tech (AI data centers), but instead as simple finance.
Perhaps you’ve heard, a new type of computing (GPUs, accelerated computing, like from Nvidia) has made data, code – and crucially – customized software and digital services, plentiful and abundant. As a result, demand for these accelerated compute systems (and the real estate and power needed to operate them, aka. a data center) has outpaced supply for a few years now.
This reverses the trend of the prior decade (capital-light 2010s) when enterprise compute (largely based on CPUs) was abundant to rent at low cost from a provider like Amazon AWS. So abundant and cheap, in fact, that many businesses didn’t rent it at all; they rented (via seat-based subscriptions) the final software products.
With demand and supply scarce, pricing on compute (and the underlying power and real estate) has gone up. A lot. Enter the neoclouds – data centers offering GPUs (and some more simply, real estate with contracted power) to customers in need of accelerated computing. These companies are raising money from investors to build more compute infrastructure to meet demand.
Two ways to fund a business
In simple terms, there are two basic ways to fund a business:
- Debt financing, borrow cash in exchange for interest payments (a cut of future cash flows)
- Equity financing, sell ownership in the business in exchange for cash (a cut of future business valuation)
Put another way, as a colleague impressed on me years ago regarding startup business costs to the owner: “Debt is expensive up front but cheap long-term; equity is cheap up-front but expensive long-term.”
Both halves of that statement assume the business startup is successful.
Debt is expensive up front because interest payments are a burden when the new business doesn’t have adequate cash flow. As the business grows and matures, though, those interest payments become increasingly manageable (cheap). Equity, on the other hand, is cheap upfront because it requires no cash outflow. But as the business grows and matures, the equity stake can become very valuable as it represents a claim on all future excess cash flows (after all expenses, including debt interest, is paid).
Hold this thought.
The neocloud cash-raise extravaganza
Early on in this current bull market, a certain batch of companies that had been “mining” Bitcoin (solving complex algorithms with GPUs to earn new Bitcoin) and other cryptocurrencies, discovered they could repurpose their data center and GPU computing hardware assets.
With data center compute and contracted data center power in high demand, the economics of renting out these GPUs and data centers surpassed crypto mining. So much so that many of these companies have been able to secure additional equity financing (a capital raise, as Arora mentioned) via IPOs, subsequent secondary stock issuance once publicly traded, and venture capital/private equity.
Some of the early entrants into the neocloud business – CoreWeave (CRWV), Nebius (NBIS), and IREN (IREN) in particular, as well as other neocloud entrants – have additionally secured debt financing to fund their expansion too. Albeit at high interest rates, or with a future equity-linked component built in.
As a result, neoclouds are fetching progressively higher enterprise valuations (the sum of equity value and debt, or market capitalization plus debt and minus cash & equivalents). The argument for these higher valuations is that elevated demand for accelerated computing will persist. But what if the demand moderates? Valuations could collapse. If there’s a future crash, investors and businesses in need of cheaper computing assets can swoop in in a couple years and get some good deals. Right? Well, maybe…
The future debt vs. equity financing problem
Will a company or investor really be able to purchase a neocloud at less than their single capital raises (via equity financing) are at today? Perhaps, but contingent on another item being true: The capitalization of the neocloud must be stilted towards equity, not debt.
Remember the statement that equity financing is expensive long-term, if the business is successful. If neocloud business economics change for the worse (as Arora was predicting for the neoclouds), the inverse becomes true. In the future, the purchaser of the equity investment is left holding an asset worth a fraction of what was originally paid. Would-be buyers of remaining assets can get pennies-on-the-dollar buying out the distressed equity.
As for debt financing, if the business is successful, interest payments and debt payoff/refinancing are cheap. But if not, servicing that debt remains really expensive. The equity in such a business may very well be cheap too (cheaper than capital raises happening at neoclouds right now), but what about the lenders demanding repayment first? The only way to expunge such debt is via restructuring, like via bankruptcy (swapping out debt for equity and diluting existing equity owners, renegotiating debt terms, or closing the business and selling off assets).
If there is a neocloud crash, say in two years from now, investors and prospective business buyers of distressed assets are not automatically guaranteed a value on the remaining equity. The increasing amount of debt financing for neoclouds, not just equity financing, is changing who will need to absorb a loss before the data center assets can be sold off at a better value. More debt load, if mispriced today, would necessitate a round of bankruptcies, asset value write-downs, and other such destruction of capital, in order to make Arora’s prediction true.
And that type of capital destruction ultimately hits equity valuations too, even for those investors and businesses that might be waiting to swoop in and purchase neocloud assets on the cheap. See the problem? Waiting for a cheaper price in the future doesn’t guarantee it can be purchased at a better value than today.
Important side point: Arora’s statement, as CEO of a company that monetizes via product/service, and a customer of data center infrastructure, serves a purpose for his company and various stakeholders. Ie. desire to build more future vertical integration into the business, at potentially even better terms.
How to avoid future bagholding as an investor
In this piece, I’m neither defending nor supporting any claim about the future direction of all neocloud businesses. The discussion requires a great deal of nuance. Disclosure: As of this writing, Chip Stock Investor owns equity in the hyperscalers, as well as CoreWeave, Nebius, and DigitalOcean. We also own shares of Palo Alto Networks. Our future stake in such businesses is subject to change.
Rather, this is meant to help think about the neocloud investment premise (demand for accelerated computing being higher than supply), and how not to be a future bagholder when the demand environment changes.
If you are buying equity in a neocloud (via the stock), there’s another half to the company’s capitalization you need to consider: Debt. What are the terms of the debt? And is the business on track to be able to service it? If not, your equity value could get wiped out.
If you own the debt, infrastructure investments are ideally made this way – assuming the current supply-demand and resulting economics of the infrastructure is sustainable long-term. If the money-plus-interest owed to you by a neocloud comes with an equity clause, or subsequent funding rounds are being made via equity, it’s time to ask hard questions as to why the managers of that (supposedly) stable long-term infrastructure business are selling long-term expensive equity for cash.
Subsequent articles will delve into evolving business models of neoclouds, but as a primer, we did this video about vertical integration in computing (as well as evolving business models of enterprise software companies) here back in April: $750 Billion in AI CapEx — Who Wins and Who Gets Left Behind
As always, the Chip Stock Investor team is here to help you uncover investing opportunities in technology, but also to help you make sure you don’t blow up your portfolio. You know, “the first two rules of investing” and such… www.chipstockinvestor.com
