Trending
Written by: Sean Peche | Ranmore Fund Management
From $100bn “strategic partnerships” to companies funding customers to buy their own products, Sean Peche questions the financial logic behind a wave of mega-AI announcements — and wonders whether the hype is outpacing the cash flow.
On 22 September, there was the $100bn OpenAI & NVIDIA deal, “to build and deploy at least 10 GW of AI data centres with NVIDIA systems.” You’d think, for a $100bn deal, you’d “finalise the details” before announcing it, because the last sentence in the press release read: “NVIDIA and OpenAI look forward to finalising the details of this new phase of strategic partnership in the coming weeks.”
Well, during those “coming weeks”, “AMD and OpenAI announce strategic partnership to deploy 6 gigawatts of AMD GPUs.” Now, maybe this doesn't change things, but the accountant in me was a little uncomfortable about Nvidia using all their current $57bn of cash plus $43bn of future profits to “invest” $100bn in a single customer — to use to buy Nvidia chips — on which the “profit” would be valued at 50x by investors.
It’s like me setting up a lemonade stand and giving money to a friend to buy my lemonade, getting my own money back, and declaring a “profit”. The only difference is my lemonade stand isn’t 8% of the S&P 500 index — not that the passive industry seems worried (or looking).
Now, with the AMD deal, OpenAI gets a warrant to buy 160m AMD shares at a knockdown price, which then vest based on deployment and “AMD achieving certain share-price targets”… Huh? OpenAI can’t influence the AMD share price, can they? Well, I guess they could pay full price for those chips, which might mean AMD “beats the estimate” and things get hot enough for OpenAI to sell their shares to fund operations.
Pure speculation on my part, but they told us in September they expect to burn through $115bn of cash in the next few years. And it’s hard to fund $115bn from their current $13bn of annualised revenue.
OpenAI’s legal team have been working hard, because in July they partnered with Oracle, committing investment “that exceeds $300bn over the next 5 years”. Call it $60bn a year — which Oracle can't fund, because they only generate $20bn per annum in operating cash flow and already have $95bn of debt from buying back stock.
What about Softbank, also partners in the $500bn Stargate venture? Well, they already have $120bn in net debt. So is this a bubble? You tell me.
Microsoft was first worth $500bn in Q1 2020 when they were generating $2bn of FCF a quarter — not en route to burning $115bn. So my conclusion is: lots of commitments, lots of hype, not much cash flow. Unless you work there.
The latest Form 144 shows Nvidia’s CEO has sold shares for $736m in the past three months, and last week the latest OpenAI funding round allowed staff and former employees to sell $6.6bn worth of shares, valuing OpenAI at $500bn. Nothing like imputing a ginormous private market value based on 1.3% of shares changing hands…
Related: Why the Market Keeps Rising—Even With Sky-High Valuations