GPU Price Increase 2026: Hyperbolic H100 More Than Doubles

Hyperbolic's H100 SXM listing was $1.50/hr in June. It's $3.19/hr now. That's a real, checkable gpu price increase 2026 story, not a rumor: three independent trackers caught the rate at three different points along the climb, and the math works out to roughly a 113% increase, about 2.1x, in under three months. It's a big jump. It's also not the flat "tripling" a quick mental rounding of $1.50 to $3.19 might suggest, and getting that number right matters if you're using it to plan a renewal or a migration.

Quick Answer: Hyperbolic's GPU Price Increase 2026, What Changed and When

Hyperbolic's H100 SXM on-demand rate rose from $1.50/hr (verified June 9, 2026) to roughly $2.89/hr (reported August 1, 2026) to $3.19/hr (confirmed on Hyperbolic's own marketplace page and independently by ComputePrices.com as of August 26, 2026). That's a roughly 113% increase, not the tripling the raw numbers might imply at a glance. Hyperbolic runs a marketplace model that sources capacity from third-party suppliers and reprices weekly to track those supplier rates, so the increase reflects tightening upstream GPU supply rather than a one-time policy change. It also isn't isolated: CoreWeave disclosed a roughly 25% across-SKU increase in July 2026, on the same underlying HBM and lead-time pressure.

Hyperbolic's H100 Rate Timeline: $1.50 to $3.19 Since June 2026

Three Independent Price Checks, Three Dates

Here's what each snapshot actually showed, in order:

Hyperbolic's own marketplace page confirms the current number directly: "GPU rentals on Hyperbolic start at $3.19/hr for an NVIDIA H100 SXM. H200s are $3.99/hr and B200s are $5.99/hr, with pricing refreshed weekly based on the best available rates from suppliers." Three separate sources, three separate methodologies, and they land within a few cents of each other by the end of August. That kind of agreement across independent trackers is what makes this a real, dated event rather than a stale or mis-scraped number.

How Big Is the Jump, Really: It's Not a Triple

Do the division and $1.50 to $3.19 is a 2.13x multiple, a 113% increase. That's genuinely large for a ten-week window, and it's worth stating plainly rather than rounding it up for effect. Calling it a "triple" would require the rate to hit $4.50; it hasn't. Calling it "more than doubled" is accurate, since anything past 2.0x clears that bar, and 2.13x does.

The distinction matters for anyone budgeting off this number. If you assumed a tripling and planned around $4.50/hr, you'd be padding your estimate by more than 40% above what Hyperbolic is actually charging today. If you assumed the increase was smaller because "it's just gone up a bit," you'd be underestimating a genuine 2.1x cost multiplier on the same workload. Get the actual number, not the vibe of the number, before it feeds into a budget line.

The middle data point matters too. Thunder Compute's August 1 comparison caught Hyperbolic mid-climb at roughly $2.89/hr, which means the increase wasn't a single overnight repricing. It moved in steps across the summer, consistent with a marketplace that adjusts weekly rather than a provider that announces one hike and holds it.

GPU Cloud Pricing Comparison: How Hyperbolic's New Rate Stacks Up

A single number means little without the rest of the market next to it. Here's where Hyperbolic's $3.19/hr H100 SXM rate sits against hyperscalers and Spheron's live marketplace, pulled the same day this post was published:

Hyperbolic's new rate is still nowhere near hyperscaler pricing. Against AWS's roughly $6.88/hr, $3.19/hr looks cheap, and against Azure's roughly $12.29/hr it looks like a different market entirely. But the comparison that actually matters for someone choosing a provider today isn't "Hyperbolic vs AWS," it's "Hyperbolic vs the other marketplace and neocloud options quoting a live rate this week." On that comparison, Hyperbolic's H100 SXM has moved from being one of the cheapest options on the board to sitting above Spheron's H100 SXM5 on-demand floor and well above Spheron's PCIe and spot tiers.

If your workload doesn't need SXM's NVLink bandwidth, our H100 NVL vs SXM5 vs PCIe form factor guide walks through when PCIe is a perfectly good substitute at a meaningfully lower rate.

What's Driving the Increase: Capacity, Demand, or Repricing

Hyperbolic's Marketplace Model Means Prices Track Supplier Rates Weekly

Hyperbolic doesn't own the GPUs it rents out. It's built on what the company describes as a decentralized operating system that orchestrates a globally distributed network of underutilized GPUs from data centers, mining farms, and personal machines, and its own marketplace page states plainly that pricing is "refreshed weekly based on the best available rates from suppliers." That structure means Hyperbolic's rate card isn't a policy decision made once a quarter. It's a rolling reflection of what the suppliers behind it are charging, updated on a weekly cadence. When those suppliers raise rates, Hyperbolic's listed price moves within days, not months.

That's the mechanism. It's a fundamentally different pricing model from a provider like CoreWeave, which negotiates multi-year contracts and only occasionally touches its public rate card. Hyperbolic's number moves faster and more often, in both directions, because it's built to.

The Broader HBM and GPU Supply Squeeze Behind Every Provider's Numbers

The mechanism explains how the price moves. The macro story explains why it's been moving up. HBM memory costs rose roughly 30% in Q4 2025 alone, and data center GPU-server lead times were running 36-52 weeks as of an April 2026 industry analysis, against an estimated backlog of roughly 3.6 million units. That's not a Hyperbolic-specific pressure. It's the same input-cost story CoreWeave's CFO cited on the company's Q2 2026 earnings call, where he disclosed that pricing changes in July included an approximately 25% increase across SKUs in response to the current demand environment and the increasing ROI customers are observing. Our CoreWeave price increase breakdown covers that hike in full, and it's worth reading alongside this one: two very different pricing models, CoreWeave's negotiated-contract structure and Hyperbolic's weekly-repriced marketplace, both moving up on the same underlying supply story in the same quarter.

CoreWeave's Agrawal put the reasoning behind the pass-through directly: "the increase in the value of output of the CoreWeave cloud has outpaced the value of the input increases" in the supply chain. Whether or not that logic applies cleanly to Hyperbolic's supplier network, the direction of travel is the same across the market right now: tight memory, long lead times, rising rates. Our GPU shortage 2026 guide has the fuller picture on what's driving the squeeze and how teams are working around it.

What This Does to Your Monthly Bill at Typical Usage

Run the numbers at typical utilization and the abstraction turns into a real line item fast. A single H100 SXM run continuously for a 720-hour month:

• At Hyperbolic's June rate ($1.50/hr): about $1,080/month
• At Hyperbolic's current rate ($3.19/hr): about $2,297/month
• Difference: roughly $1,217 per GPU, per month

Scale that to an 8-GPU node and the gap alone is close to $9,700 a month, before you've paid for anything else the job needs. That's the real cost of a marketplace repricing that most teams never notice until the invoice lands, because nothing about the workload changed, only the rate underneath it.

Pricing fluctuates based on GPU availability. The prices above are based on 27 Aug 2026 and may have changed. Check current GPU pricing → for live rates.

Where to Find H100 Capacity at the Old Rate

If your budget was built around H100 at $1.50-2/hr, that specific number is gone from Hyperbolic's board, but capacity in that range hasn't disappeared from the market. A few places to look:

1. Spheron's aggregated marketplace. Spheron pulls capacity from 5+ providers on the backend, which keeps on-demand H100 pricing competitive without locking you into a single supplier's weekly repricing. H100 GPU rental on Spheron currently starts lower per GPU on PCIe than Hyperbolic's new SXM rate, and spot pricing runs lower still for workloads that can tolerate reclaim.
2. Check PCIe if your workload doesn't need SXM's NVLink bandwidth. A lot of inference and mid-size fine-tuning jobs never touch the interconnect bandwidth SXM is built for. Our form factor decision guide (linked above) walks through how to tell the difference before you pay for hardware you don't need.
3. European neoclouds, if data residency allows it. Providers like DataCrunch have historically undercut US-based marketplaces on raw GPU-hour cost. See our DataCrunch (Verda) alternatives roundup for the current field.

None of this is a one-time check. Marketplace rates move weekly on Hyperbolic's own admission, and the rest of the market is repricing on the same HBM and lead-time pressure, so whatever number you land on today is worth rechecking before you sign anything longer than a month.

Full documentation for standing up an instance on Spheron is in the Spheron docs.

Pricing fluctuates based on GPU availability. The prices above are based on 27 Aug 2026 and may have changed. Check current GPU pricing → for live rates.

Watching Hyperbolic's rate climb and wondering what H100 costs elsewhere this week? Spheron's marketplace aggregates capacity from 5+ providers with per-minute billing and single-GPU granularity, so you're not locked into one supplier's weekly repricing.

H100 GPU pricing → | Get started on Spheron →