GPU Mart

GPU Mart is the exchange where the AI era gets its hardware. We connect AI startups, university HPC labs, and enterprise buyers with verified GPU inventory.

[GPU Mart @GPUMart]

We're launching GPU Mart: the liquidity layer for global AI infrastructure.

Here's why it needs to exist, and why right now. 🧵

[GPU Mart @GPUMart]

@boardyai working on something unannounced in the compute space. Boardy Pro would be huge for this

[GPU Jack @gpujack]

Some personal news

[GPU Jack @gpujack]

it says here that H200s **might** be arriving @GPUMart next week...

who wants one?

[GPU Jack @gpujack]

. @GPUMart is rapidly becoming the premier exchange for AI infrastructure. While we handle massive deals (like last week’s $15M H100 order) we remain committed to supporting the "little guy."

For startups, owning your infrastructure is an uphill battle. Brokers and OEMs consistently prioritize large-scale orders, making it incredibly difficult to secure a single node.

And if that single node happens to be an H200? You have a better chance of borrowing Jensen's leather jacket.

Today, we met with an AI startup in desperate need of one H200 node to launch an ambitious project.

@GPUMart is searching far and wide to make it happen.

Who has the node? Full details at the link below.

[GPU Mart @GPUMart]

Last night GPU Mart had an executive offsite at @PubKey DC

$btc up roughly $2k since this photo was captured

[GPU Jack @gpujack]

🚨 AI Factory Alert 🚨

Now live on @GPUMart: Turnkey NVIDIA HGX H100 Cluster, 504 GPUs, full NDR InfiniBand fabric, US-based.

Use it for LLM pre-training, fine-tuning, large-scale inference, HPC, Neocloud operations, multimodal and video AI, or deep learning recommendation models.

https://t.co/zLrARrYiaN

[GPU Mart @GPUMart]

GPU Mart has some fresh ~~certified preowned~~ H100 servers available for $174k

each server delivers enterprise-grade AI power:
• 8x NVIDIA H100 80GB GPUs
• 2x Intel Xeon Platinum 8462Y+
• 2TB DDR5 RAM
• Fully tested, ready to ship from the USA

https://t.co/Cy4V2u5CaQ

[GPU Mart @GPUMart]

🥕🥕🥕 https://t.co/5lmKEtdz9K

[GPU Mart @GPUMart]

The Easter Bunny stopped by GPU Mart HQ a couple days early with a deal sweeter than a basket of peeps.

500 brand new H100 and GH200 nodes available at a steep discount

[GPU Mart @GPUMart]

The Great GPU Shortage

sellouts, sublets and slumlords

via @SemiAnalysis_ @dylan522p

[GPU Mart @GPUMart]

Claude serving up Cowork tokens to the early east coast shift

[GPU Mart @GPUMart]

@LLMJunky

[GPU Mart @GPUMart]

"Human-only code reviews are the new bottleneck in software development. The future of quality depends on dissolving these bottlenecks without sacrificing rigor." - @harjotsgill, CEO of @CodeRabbitAI

[GPU Mart @GPUMart]

"By almost any definition anyone could have written down let's say 5 years ago or 10 years ago... we basically have the vision of AGI that we had back then." -@realDanFu

explore the future of hardware and scaling with @Tim_Dettmers (@allen_ai) and @mattturck

https://t.co/awoXpqC6Fu

[GPU Mart @GPUMart]

burn the keys

[0x_Vivek @0x_Vivek]

you simply do not grasp the relief of zero inference tax until you actually burn the api key

buy gpus. own your compute.

[GPU Mart @GPUMart]

@mustafasuleyman https://t.co/boFl9NTaRs

[GPU Mart @GPUMart]

yc s26 batch waiting for an h100 rental

@garrytan @ycombinator

[GPU Mart @GPUMart]

yc s26 batch waiting for an h100 rental

@garrytan @ycombinator

[Solomon Tucker @babesandballers]

The situation is a lot worse than they want to show you at BWI. This is 430pm. Some told me they got in line at 1030am. Folks Find another airport ASAP @NFTC_News @juliedonaldson_ @Lokay @TravelLeisure @DCCelebrity @tomroussey7news #tsa #travel

[GPU Mart @GPUMart]

@Midnight_Captl The "compute event horizon" framing is right. When inference demand is throttling production AI and B200 spot is hitting ATH simultaneously, you're past the point where adding supply solves it near-term. The backlog absorbs new capacity before it can hit the market.

[GPU Mart @GPUMart]

What the normalized chart doesn't show: the divergence is accelerating.

B200 is up ~114 vs baseline. H100 at 98. A100 at 85. The gap between generations is widening faster than deployment cycles. Teams that locked in H100 contracts at trough pricing are sitting on real cost advantages right now.

[Steve Hou @stevehou]

Normalized GPU rental price appreciation. AI compute prices are going up across the board, especially the newest installed Blackwell B200 chips followed by H100 and A100. AI demand is surging, showing no sign of slowing down.

Credit: @Silicon_Data

[GPU Mart @GPUMart]

@lexfridman The supply chain section (39:23) is the most underrated part. Jensen on rack-scale co-design explains why you can't separate the software story from hardware constraints. B200 hitting ATH pricing this week isn't a coincidence — it's rack-scale demand in practice.