Twitxy Saved

Aethir Ecosystem

Verified account This account is verified. Learn more

@AethirEco

The official Ecosystem account of @AethirCloud - Powerful Partnerships, Endless Potential | @AethirCloud - All things #Aethir
Aethir aethir.com/ecosystem Joined April 2024
40 Following 247K Followers 4.7K Posts

Latest posts 5

Pinned post

Aethir Ecosystem Verified account @AethirEco
AWS raised GPU prices twice in 2026: 15% in January, another 20% in July.

That is more than 33% year-to-date on the same silicon.

H100 capacity on P5 instances now runs $5.191 an hour, and B300 sits at $14.04 per accelerator-hour馃У
9 replies 14 reposts 97 likes 5.1K views View on X
Aethir Ecosystem Verified account @AethirEco
Usage-weighted token prices fell roughly 40% between late June and mid-August, per Silicon Data.

In July, AI spend per employee at the heaviest-spending 1% of enterprises rose about 49% month over month, per Ramp.

Cheaper inference gets used more. Agents, longer contexts, and always-on workflows absorb every price cut and come back for more GPUs.

That demand lands on inference, where cost per GPU hour decides what ships. Aethir keeps its fleet at 95%+ utilization and prices inference workloads 40-80% below centralized alternatives.

Falling token prices are a demand signal.
2 replies 6 reposts 28 likes 1.3K views View on X
Aethir Ecosystem Verified account @AethirEco
JLL puts the average US wait for a large data center grid connection at four years, and up to seven for a 100 MW hookup in Northern Virginia. The models that need that power ship in quarters.

The bigger the load, the further back in the utility queue it lands, and a gigawatt campus asks for the biggest load of all.

Aethir ACCELERATE is focused on building mid-sized sites, and smaller power requests have far more places on the grid where they can land.

That is how B300 and GB300 clusters go from site selection to live capacity in months.

Right-sized load, shorter line.
3 replies 5 reposts 39 likes 1.7K views View on X
Aethir Ecosystem Verified account @AethirEco
Enterprises buying AI compute from a hyperscaler usually juggle two products: Reserved capacity for the steady load and on-demand instances for spikes, each with its own pricing and availability.

Aethir ACCELERATE plans to bring dedicated NVIDIA B300 and GB300 clusters of 64 to 256 nodes online, and those clusters will sit inside a network of 430,000+ GPU containers across 94 countries.

The baseline runs on dedicated hardware. Overflow, testing, and regional inference can route to the wider network without adding a second vendor.

Owned capacity works best with an orchestrated network behind it.
3 replies 8 reposts 72 likes 1.9K views View on X
Aethir Ecosystem Verified account @AethirEco
Hyperscalers are stretching how long their servers stay in service.

Meta moved to useful lives of about 5.5 years, which keeps older accelerators on the books, and in the fleet, deep into later GPU generations.

That helps an accounting line. It does less for a buyer who needs current inference performance today.

Aethir ACCELERATE plans to build new capacity around NVIDIA B300 and GB300 clusters, sized at 64 to 256 nodes for the workloads enterprises are running right now, and designed to go from site selection to live capacity in months.

Capacity coming online in 2026 should run on 2026 silicon.