The Vaporware Revolution

The Credit Strategist Blog

One day after announcing a multi-hundred billion dollar increase in AI-related backlog, ORCL announced that it signed a $300 billion deal to provide OpenAI with computing power over roughly five years starting in 2027. This confirmed where a significant part of that backlog comes from (previously OpenAI disclosed it would pay $30 billion annually to ORCL for these services). The deal will require 4.5 gigawatts of yet-to-be-built power capacity, roughly comparable to the output of two Hoover dams or the consumption of four million homes. Building that capacity will add significantly to the cost of this contract.

OpenAI is currently losing about $1 billion every month while generating around $10 billion in annual revenue, a fraction of the cost of the new contract. It is growing very rapidly but has to keep growing at an extraordinary rate to meet the demands of this contract. The company told investors it will burn more than $100 billion in cash by 2029, which would take it through the first three years of this contract. One has to extraordinarily optimistic about OpenAI’s business prospects to believe it will be able to pay $60 billion a year for any contract. AI proponents will argue that questioning this shows a failure of imagination. Perhaps they are correct but techno-optimists have the burden of proof with numbers on this scale.

OpenAI and ORCL appear to be taking a page out of Elon Musk’s playbook. Musk mastered the art of announcing vaporware - products that lure investors to TSLA stock but had little chance of appearing on any reasonable time schedule (or appearing at all). Vaporware played a key role inflating TSLA’s market cap to a level that bears no relationship to its financial performance. ORCL’s stock rose by 40% based on its AI backlog while its actual results were less than stellar. OpenAI is not publicly traded but no doubt it will use the new contract to raise more money at higher valuations. At the rate it is burning cash, it’s going to need a lot more cash over the next few years.

ORCL has a strong business ex-AI. OpenAI has no business ex-AI. Right now a company could stick the words “AI” on virtually any kind of business and see its stock lift. We saw that happen during the early days of crypto (a New Jersey delicatessen actually tried it - the principals are going to jail I believe). This is what happens in a bubble. Investors buy first and ask questions later (especially if the investor is a AI-driven machines). The questions that should be asked concern the feasibility of elaborate AI plans that aren’t even close to fruition. For example, the extravagant Stargate project - which involves both ORCL and OpenAI along with the master of hype, Masayoshi Son’s Softbank - may break ground on one small facility this year but isn’t close to putting its $500 billion target to work. That number itself may well be vaporware.

Building data centers is an extremely difficult and expensive task that occurs in the material world, not in the intangible stock market. The success of the ORCL-OpenAI contract assumes massive data centers will be built with no interruption in funding from market dislocations or exogenous events and with easy access to the enormous energy required to power them. An awful lot must go right for the parties to satisfy this contract.

Hubris is a much underrated quality today but one timeless investment lessons is that it is better to buy what nobody wants rather than what everybody wants. Vaporware is the antithesis of hubris.