Kebz retweetledi
Kebz
2.1K posts
@lightofrahim @primisprotocol @khouuba You might be able to get access to their closed beta and test it out i know they were opening up slots
English
Français
Kebz retweetledi
@blknoiz06 @metaversejoji Primis
2DfBjrPFZjDTiCY6pxchS6aSdUdEpkm7PdqpovHjBAGS
Kebz retweetledi
Kebz retweetledi
the last few days in compute markets have been insane.
gpu availability collapsing. on-demand prices repricing higher. compute futures announced.
compute etfs filings starting to appear.
there couldn’t be clearer signs that compute is becoming a real financial market.
and once compute becomes a market, builders need more than provider lists and spot screenshots.
they need:
pricing they can trust
routes they can use
rates they can plan around
one interface instead of fragmented chaos
that’s why i’m more excited than ever about what we’re building at @primisprotocol (open beta closer each day)
not another cloud. not another gpu marketplace.
but the pricing layer for compute.
primis mode.
English
@tassadaro22393 Look into @primisprotocol @khouuba they’re building the pricing layer for compute, open beta should be in a couple weeks. Very promising project
English
Imagine beeing highly liquid. What to buy quick ?!
Help me !
English
Kebz retweetledi
Nebius raising H100 and B200 rental rates ~30% with no new supply is the clearest signal yet that AI compute remains structurally constrained and that reliable capacity now carries real pricing power.
This is exactly why a dedicated pricing layer for compute is essential.
@primisprotocol is building precisely that: reserved, workload-specific rates that protect AI teams from post-reservation volatility.
When spot routes move 25-30%, Primis users continue reconciling at the rate they locked in delivering the cost predictability centralized providers increasingly struggle to offer.
Open beta is loading at the right time. The future of scaling AI won’t just be about who secures the GPUs it will be about who can price and manage that compute intelligently.
Primis mode.
Yiannis Zourmpanos@yianisz
$NBIS sent customers an email saying GPU rental prices will jump roughly 30% starting June 1. - H100 goes from $2.95/hr > $3.85/hr. - B200 from $5.50/hr > $7.15/hr. - Even B300 pricing jumps almost 29–30%. That matters because cloud GPU pricing normally falls over time as supply improves. Nebius just did the opposite.. You don’t raise GPU pricing 30% overnight unless customers are willing to pay it, especially after $NVDA confirmed supply stays tight through Vera Rubin. More importantly, this is margin expansion on already deployed infrastructure. No new GPUs. No new customers. Just higher revenue/hour.. And honestly, I think this weakens the “AI cloud commoditizes” bear case. Premium infrastructure with NVIDIA alignment, networking quality, and reliable capacity is starting to show real pricing power.
English
Kebz retweetledi
this is exactly the shift.
compute is moving from “cloud resource you rent when needed” to a constrained market where short-term access is no longer guaranteed.
if providers can sell long-term capacity to frontier labs, public on-demand supply disappears first.
so the builder problem becomes:
what can i run?
where can i run it?
at what rate?
for what window?
with what confidence?
that’s why the next layer is not just another GPU marketplace.
it’s pricing, routing, reservation, and access across fragmented supply.
this is what @primisprotocol is building for.
David Arnal@davidarngar
GPUs are running out worldwide In recent times, it's becoming increasingly difficult to rent servers with GPUs. For example, right now I couldn't find a public provider where you can rent even an 8xH100 (well, there are still a couple of machines on Vaste), let alone a larger cluster. Even getting a single video card has become challenging. The A100 now costs more than it did two years ago - a video card that, by the way, has been on the market for almost six years. The price of newer graphics cards has also risen by 1.5-2 times. No immediate improvement in the situation is expected. Cloud providers don't see the point in renting GPUs publicly or for short periods, since Anthropic will buy them anyway. And how do you solve your computer problems?
English
gpu availability collapsing.
on-demand pricing up 29%.
b200 still at 0%.
this is the primis thesis playing out in real time.
builders don’t just need access to compute.
they need predictable compute economics.
when availability leads pricing, the winner is not just whoever owns supply.
it’s the layer that helps builders access, route, price, and plan around compute.
that’s what @primisprotocol is building.
primis mode.
Warren Pies@WarrenPies
Nebius ($NBIS) raising on-demand pricing by 29%. The 3Fourteen GPU Availability Index went from 80% at the beginning of the year to 9%. GPU availability leads pricing. B200 still at 0%...GH200 tightening further.
English
Kebz retweetledi
fyi only 4 big sprints left !
primis mode
Khouba@khouuba
fyi only 6 big sprints left ! primis mode
English
Kebz retweetledi
primis is the pricing layer for compute.
we help builders access compute through one unified layer instead of dealing directly with fragmented providers, unstable pricing, and infra complexity.
this is our @BagsHackathon demo submission.
English
Kebz retweetledi
holyyyyyy!!
@OpenAI @sama basically saying what the market is starting to realize:
access to compute is not enough.
builders need guaranteed capacity and predictable planning for critical workloads.
that’s the @primisprotocol thesis too. but as a neutral layer across fragmented providers, not inside one model ecosystem.
predictable compute is becoming the product.
primis mode.
OpenAI@OpenAI
Introducing OpenAI Guaranteed Capacity: a new offering that enables customers to guarantee long-term access to OpenAI compute. We’ve made long-term investments in infrastructure, partnerships, and capacity planning to help customers scale reliably. Now, Guaranteed Capacity helps customers plan ahead for critical workloads in a compute-constrained world. openai.com/guaranteed-cap…
English
Kebz retweetledi
Kebz retweetledi
agree.
physical compute is too heterogeneous to collapse into one clean commodity.
but the pricing / financing layer still needs standardized risk surfaces.
so the market won’t be one compute price. it’ll be structured rates by workload, tenor, reliability, latency, and supply type.
that’s exactly where @primisprotocol fits. the pricing layer of compute.
English
compute financing is a commodity - lenders need to price and transfer risk
physical compute is not a commodity - a GPU fits into a larger BOM / RA that is highly heterogeneous
inevitable to me that multiple markets serving different parts of the value chain will exist
English
@primisprotocol Its a no brainer for builders
The recent GPU spikes definitely validated that
Primis mode
English
