Ancapex brings AI teams that need reserved GPU capacity to the operators that have it. You sign directly with the buyer, and our fee is due only when a contract is signed: nothing upfront.
As of 19 October 2026 we hold 11 open buyer requests. Regions asked most: US and Europe; also Asia-Pacific, Canada. Starts wanted: Q4 2026 to Q1 2027. Terms asked: 12 months, a few shorter blocks too.
Counted every week from our open requests, without buyer names. Sizes are bands of the total GPUs a buyer asks for (under 100, hundreds, thousands), never one buyer's exact number; one request can name two models.
We show capacity without your name, and only with your consent. A buyer learns the operator, the site and the price only after you have signed our NDA and referral agreement and we have had a call with the buyer. Our public pages show rates only as a median and a middle half that each mix several operators' offers, never one operator's price.
You pay our fee only when a buyer signs a contract with you. Its size is agreed per deal in the referral agreement, and there is nothing to pay upfront.
We show your capacity to buyers without your name until you agree. We reply within one business day from dima@ancapex.io.
Send the GPU type, quantity, region, ready date, term and an indicative price through the form on this page, or book a call. We reply within one business day.
The operator, and only when a buyer signs a contract. Nothing is due upfront; the size of the fee is agreed per deal in the referral agreement.
Not before you agree. We show capacity without your name and only with your consent. A buyer learns the operator, site and price after you have signed our NDA and referral agreement and we have had a call with the buyer.
Yes. Most buyers reserve months ahead, so a site with a fixed ready-for-service date, a term and an indicative price is enough to start.
Mostly reserved B300 and GB300 NVL72, usually on terms of one year or longer, a few shorter. The current requests are listed on this page, and the supply side is on GPU availability.