October 2026: Monthly average compute price of NVIDIA's RTX 5090
$0.00
1
7
October 2026: Monthly average compute price of NVIDIA's RTX 5090
$0.00
1
7
AI Analysis
Trader mode: Actionable analysis for identifying opportunities and edge
About This Event
October 2026 If the average value of NVIDIA RTX 5090 compute per hour is above X in October 2026, then the market resolves to Yes. The market resolves based on the average value of NVIDIA RTX 5090 compute per hour in October 2026, calculated as the arithmetic mean of hourly values reported by Ornn. Revisions to the underlying made after expiration will not be accounted for. The "USD" iteration of the index will be used unless explicitly stated otherwise. Values are rounded to two decimal places
What Prediction Markets Are Forecasting
Traders on Kalshi currently give a 65% chance that the average price of renting an NVIDIA RTX 5090 GPU for compute will exceed $0.50 per hour in October 2026. That's roughly a 2 in 3 chance, which is notable confidence but far from a sure thing. The market is saying that while higher prices are the more likely outcome, there's still a real possibility that the price stays below that threshold.
To put $0.50 per hour in context: that's about the cost of a small coffee per hour of GPU time. For AI developers, researchers, and small startups, this price point often determines whether they can afford to run their own models or must rely on cloud providers.
Why the Market Sees It This Way
Several forces are pushing prices upward. First, the RTX 5090 is one of the most powerful consumer GPUs ever made, with 32GB of VRAM that makes it attractive for running local AI models. Demand for consumer-grade AI hardware has exploded since 2023, and that trend shows no sign of slowing.
Second, NVIDIA has a history of supply constraints. The RTX 5090 launched in early 2025 to immediate shortages, and scalpers drove prices well above MSRP. If NVIDIA continues to prioritize data center chips for AI companies, consumer cards like the 5090 may remain scarce, keeping rental prices high.
Third, the broader AI compute market has seen persistent price inflation. Cloud GPU prices have risen steadily as demand outpaces supply, and consumer cards often follow similar trends when used for distributed compute networks.
However, the 35% chance of prices falling below $0.50 reflects real counterforces. By October 2026, the RTX 5090 will be nearly two years old, and NVIDIA will likely have released newer cards that could pull demand away. More competition from AMD and Intel could also pressure prices downward.
Key Dates and Events to Watch
Watch for NVIDIA's next GPU generation announcements, typically teased in the fall. If a strong successor to the 5090 is announced for late 2026, prices for the older card could drop quickly. Also monitor NVIDIA's quarterly earnings reports for signals about production capacity and whether they're shifting more silicon toward data center products.
How Reliable Are These Predictions?
Prediction markets have a solid track record for technology price forecasts, though they're better at short-term predictions than 18-month horizons. The further out the date, the more uncertainty creeps in. A lot can change in the AI hardware world between now and October 2026, and the market's 65% figure reflects that genuine uncertainty rather than false precision.
Current Market Outlook
The market is pricing a 97% probability that the monthly average compute price for NVIDIA's H100 will stay above $2.00 per hour through October 2026. That is near-certain territory. Kalshi traders are effectively betting that demand for H100 compute will remain robust enough to prevent a price collapse below that threshold over the next two years.
To put this in context: $2.00 per hour is roughly where H100 spot pricing sat in late 2023 when supply was tightest. Current spot rates on major cloud providers like AWS and Lambda Labs range from $2.50 to $4.00 per hour depending on commitment length and provider. The market is saying there is only a 3% chance that average pricing drops below the $2.00 floor by October 2026.
Key Factors Driving the Odds
The dominant factor is supply constraints. NVIDIA cannot manufacture H100s fast enough to meet demand from hyperscalers, AI startups, and sovereign cloud projects. TSMC's CoWoS packaging capacity is the bottleneck, and while NVIDIA is expanding supply, the H100's successor (B100/B200) will likely absorb much of the new production. H100s will remain the workhorse for inference workloads even as Blackwell ships.
The second factor is the stickiness of long-term contracts. Most H100 capacity is locked into 1-3 year deals at prices above $2.00 per hour. The spot market is small relative to reserved instances. Even if spot prices dip temporarily, the monthly average will be pulled upward by these contractual floors.
The third factor is the Ornn index methodology. Ornn tracks a weighted average across multiple providers. The index includes both spot and reserved pricing, but the weighting favors larger providers like AWS and Azure where H100 pricing has been stable. A price crash would require simultaneous discounting across all major cloud providers, which is unlikely given each is still capacity-constrained.
What Could Change These Odds
The most obvious catalyst is a macroeconomic downturn that cuts AI capex. If the Fed triggers a recession in 2025 or 2026, enterprise cloud spending could freeze, and hyperscalers might cancel GPU orders. But this is the scenario the market sees as only 3%.
A second risk is NVIDIA's own product cycle. If Blackwell GPUs ship in volume by mid-2025 and offer 2-3x the performance per dollar, H100 demand could crater as customers migrate. However, history suggests older NVIDIA architectures retain value for inference workloads, and the transition takes 12-18 months. The October 2026 date is early enough that H100s will still be in heavy use.
The biggest unknown is the Ornn index itself. If Ornn changes its methodology or provider list, the average could shift mechanically. But the market appears comfortable that the index will remain consistent through 2026.
AI-generated analysis based on market data. Not financial advice.
Overview
The NVIDIA H100 GPU is a high-performance computing chip designed for artificial intelligence and machine learning workloads, particularly large language model training and inference. Since its launch in 2022, the H100 has become the dominant hardware for AI compute, with demand far outstripping supply. The compute price per hour for an H100 is a market-driven metric that reflects the cost to rent or access this GPU in cloud data centers. This price is influenced by factors such as chip supply, data center capacity, electricity costs, and the overall demand for AI compute from companies like OpenAI, Google, and Meta. The prediction market question asks whether the average hourly compute price of the H100 in October 2026 will be above a specific threshold, as measured by the Ornn index, a third-party aggregator of GPU rental prices. This index tracks spot pricing from major cloud providers and specialized GPU rental platforms, providing a real-time benchmark for the cost of AI compute. As of 2024, the H100 compute price has been volatile, with initial scarcity pushing prices above $3 per hour, followed by a decline to around $1.50 to $2.00 per hour as supply increased and alternative chips like NVIDIA's own H200 and AMD's MI300X entered the market. The future price trajectory depends on several unknowns: whether AI demand continues to grow exponentially, how quickly new data centers come online, and whether competing hardware or more efficient software reduces the need for H100s. This market is of interest to investors, AI researchers, and cloud pricing analysts because it serves as a proxy for the health of the AI industry and the profitability of NVIDIA's data center business.
Historical Context
The concept of renting GPU compute by the hour emerged with the rise of cloud computing in the 2010s, but it became mainstream with the AI boom starting in 2022. Before that, GPU rental was niche, primarily used for cryptocurrency mining and academic research. The launch of OpenAI's ChatGPT in November 2022 triggered a surge in demand for AI training and inference, which in turn drove up demand for NVIDIA's A100 and later H100 GPUs. By early 2023, H100s were in such short supply that cloud providers allocated them by lottery or waitlist, and spot prices reached as high as $4 per hour on some platforms. In 2023, NVIDIA increased H100 production capacity by securing long-term supply agreements with TSMC, its chip manufacturer. This led to a gradual easing of supply constraints, and by mid-2024, H100 rental prices had fallen to around $1.50 to $2.00 per hour for on-demand instances. However, prices remained above historical levels for comparable GPUs (e.g., the A100, which cost about $1 per hour at its peak). The Ornn index was established in 2023 to provide a transparent, aggregated benchmark for GPU rental prices, filling a gap left by opaque cloud provider pricing. Historical data from Ornn shows that H100 prices peaked in Q1 2023 and have since declined roughly 40% through Q3 2024.
Why It Matters
The price of H100 compute is a direct indicator of the cost of doing AI research and development. For startups and academic institutions, high GPU rental costs can be a barrier to entry, limiting who can train large models. If prices remain above $2 per hour in October 2026, it would suggest that demand for AI compute continues to outpace supply, which could mean NVIDIA maintains its dominant market position and pricing power. This would affect the profitability of cloud providers and the viability of AI startups that rely on rented GPUs. Conversely, if prices fall below $1 per hour, it would indicate a glut of compute capacity, possibly due to increased competition from AMD, Intel, or custom chips from Google (TPU) and Amazon (Trainium). This could signal a maturing AI market where hardware is commoditized. Investors in NVIDIA stock watch these prices closely because they correlate with the company's data center revenue. For policymakers, the cost of compute influences which countries and companies can lead in AI development, as cheaper compute democratizes access. The downstream effects include the pace of AI innovation, the concentration of AI power in a few large companies, and the energy consumption of data centers, since lower prices might encourage more widespread use.
Current Status
As of late 2024, the H100 compute market is in a state of gradual normalization. Supply has improved significantly since the 2023 shortages, with NVIDIA ramping production and cloud providers expanding data center capacity. The Ornn index for October 2024 shows an average price of $1.72 per hour, down from $2.30 in January 2024. However, demand remains strong, driven by continued investment in generative AI from companies like OpenAI, Anthropic, and Meta. New competitors like AMD's MI300X and Intel's Gaudi 3 have entered the market, but they have not yet captured significant market share. NVIDIA's next-generation Blackwell GPU (B100) is expected to launch in 2025, which could reduce demand for H100s as customers upgrade. The key uncertainty for the October 2026 price is whether the AI industry's growth will sustain enough demand to keep prices elevated, or if the combination of increased supply and competition will drive them down.
Frequently Asked Questions
What is the Ornn index and how is it calculated?
The Ornn index is a third-party benchmark that tracks the average hourly rental price of NVIDIA H100 GPUs across multiple cloud providers and specialized GPU marketplaces. It calculates the arithmetic mean of all reported hourly prices for a given month, using the USD value of the index.
Why does the H100 compute price matter for AI companies?
The H100 compute price directly affects the cost of training and running AI models. For a startup, renting H100s can cost hundreds of thousands of dollars per month, so even a $0.50 per hour change can significantly impact budgets and profitability.
Will the H100 be obsolete by October 2026?
Probably not obsolete, but it may be superseded by newer GPUs like NVIDIA's Blackwell or Hopper Next. However, H100s will still be used for inference and less demanding training tasks, so rental demand may persist, though at lower prices.
Educational content is AI-generated and sourced from Wikipedia. It should not be considered financial advice.
