It is the bartering of despair

Openai has just announced that you will buy 6 AMD GPUS gigawatts. Not with cash, but with Warrants: AMD issues Openai the right to buy up to 160 million of its shares if it meets staggered purchase milestones. A company that loses 2.5 billion every six months has just committed to a massive infrastructure deployment. And it will, at least for the moment, without releasing a hard.

OpenAi has titled The statement of this blog operation like a ‘Strategic ParnerTeship ‘but rather it is A barter of mutual survival disguised as business alliance.

1. AMD desperately needs to exist as a credible alternative to Nvidia. Although that means selling its own capital instead of only chips.
2. OpenAi needs to escape Its total dependence on a single supplier that already has too much negotiation power.

None can afford what is promising in traditional terms. But Both need the market to believe that they can.

The story tells the structure of the Warrantsthat are unlocked by sections:

1. The first, with the initial deployment of a gigavatio.
2. The following, as scale to the sixth.

But there is more: the VARING (Acquisition of rights) is also linked to AMD reach “certain objectives of action prices” since OpenAi achieves “necessary technical and commercial milestones.” Translation: They only work if both companies grow exponentially. Both have just become hostages of mutual success.

He Timing It is no accident: two weeks ago Openai signed 100,000 million with Nvidia. Now this. AMD values ​​it in “tens of billions of dollars in revenue” counting drag effect. A billion dollars in infrastructure commitments in fifteen days. By a company with negative cash flow. The joke is counted alone.

The innocent spectator will see this news and assume that it is only a diversification of suppliers. He who sees beyond his noses will understand that andStán accumulating promises that cannot pay but that the market values ​​as if they were real effective. AMD has risen more than 30% in the stock market after the announcement. A climb that does not come from more chips sold, but of marketing: the chatgpt manufacturer has validated its product.

And as always, what they don’t say is as important as what they say. The first Gigavatio does not reach the second half of 2026. One year. The agreement is “definitive” but the milestones are “flexible”, conditioned to technical, commercial and stock market objectives. No one signs rigid calendars in this industry because nobody knows if there will continue to exist to the current rate of spending.

And above all: no press release clarifies how Openai will finance this without issuing more debt or dilute more capital.

We are seeing an industry financing itself with optionality and future:

1. Nvidia finances OpenAi to buy Nvidia chips.
2. AMD gives him Warrants A OpenAI to buy AMD chips.
3. OpenAI becomes a potential shareholder of its suppliers.
4. Suppliers are revalued because Openai buys them.
5. OpenAi can only exercise those Warrants If suppliers grow while complying with huge orders.

Meanwhile, The only real money entry is chatgpt subscriptions. That do not reach to cover current operating costs, much less for a billion infrastructure.

AMD has just exchanged certainty for optionality. Openai has just exchange liquidity that does not have the possibility of having part of its supply chain. Both bet in the same box: that the exponential growth of the demand for AI will make economic fundamentals irrelevant. It is an understandable, but very high bet.

They may be right. Or that in five years this agreement appears in business schools as a case study on Chow an entire industry decided that mathematics were optional. The market, meanwhile, rises 30%. Because nothing says “solid foundations” how to finance yourself selling shares of yourself to buy things that you cannot pay.

In Xataka | The engineer who does not need spotlights: smooth his AMD at the edge of bankruptcy and ten years later has made her an empire

Outstanding image | Xataka