March 2027: Monthly average compute price of NVIDIA's H100
$0.00
1
9
March 2027: Monthly average compute price of NVIDIA's H100
$0.00
1
9
AI Analysis
Trader mode: Actionable analysis for identifying opportunities and edge
About This Event
March 2027 If the average value of NVIDIA H100 compute per hour is above X in March 2027, then the market resolves to Yes. The market resolves based on the average value of NVIDIA H100 compute per hour in March 2027, 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 unless otherw
What Prediction Markets Are Forecasting
Traders on Kalshi currently give roughly a 56% chance that the average hourly compute price for an NVIDIA RTX 5090 will exceed $0.50 in March 2027. That's barely better than a coin flip. The market is signaling genuine uncertainty about where GPU rental prices land two years out, not a strong conviction in either direction.
To put that in perspective: a 56% probability means the market thinks a $0.50 threshold is essentially a toss-up, with a slight lean toward higher prices. If traders were confident in a clear trend, you'd expect odds closer to 70% or 30%. Instead, we're looking at a market that acknowledges multiple plausible futures.
Why the Market Sees It This Way
The RTX 5090 is NVIDIA's flagship consumer GPU, and its "compute price" refers to what it costs to rent one for an hour of processing power, typically through cloud providers or distributed computing networks. Several forces are pulling this number in opposite directions.
On the supply side, the 5090 launched in early 2025 with a $1,999 MSRP, but demand for AI inference workloads has kept prices elevated. The card's 32GB of VRAM makes it attractive for running local AI models, and hobbyists and small businesses have been snapping them up. If that demand persists through 2026, rental prices could easily climb above $0.50 per hour.
But there's a countervailing force: NVIDIA's release cadence. By March 2027, the RTX 6090 or whatever follows will likely be out, and the 5090 will be a generation old. Older hardware tends to get cheaper to rent as newer options flood the market. The question is whether the 5090's compute becomes a commodity that drops in price, or whether AI demand keeps growing fast enough to absorb all available capacity.
Electricity costs and cloud provider pricing strategies also matter. GPU rental markets have been volatile, with prices swinging based on crypto mining booms, AI hype cycles, and data center buildouts.
Key Dates and Events to Watch
NVIDIA's GPU announcements typically happen at CES in January or GTC in March. Any hints about the next generation's specs and pricing will directly affect expectations for the 5090's residual value. Watch for Blackwell Ultra or Rubin architecture news through 2026.
Cloud providers like AWS, Google Cloud, and specialized GPU rental services (Vast.ai, RunPod) regularly adjust their pricing tiers. Major enterprise AI deals could soak up supply and push prices up, while a slowdown in AI investment would have the opposite effect.
How Reliable Are These Predictions?
Prediction markets have a solid track record on discrete events with clear resolution criteria, like elections or policy decisions. This market is trickier because it depends on a specific data source (Ornn's index) and a specific price threshold. Two years is a long horizon for hardware markets, and small changes in AI adoption rates or NVIDIA's product roadmap could shift prices dramatically.
The 56% figure is less a prediction and more a reflection of balanced uncertainty. Markets are good at aggregating what people know today, but nobody has a crystal ball for the AI hardware landscape of 2027.
Current Market Outlook
Kalshi traders currently price a 56% chance that NVIDIA's RTX 5090 compute will average above $0.50 per hour in March 2027. That's essentially a coin flip with a slight lean toward the higher threshold. The market is telling you this: two years out, nobody has strong conviction about GPU rental pricing, and the 56% figure reflects genuine uncertainty rather than a confident directional bet.
The RTX 5090 launched in January 2025 at a $1,999 MSRP. On cloud platforms like Vast.ai and RunPod, Blackwell-generation GPUs initially commanded premium rates, often $0.80 to $1.20 per hour during the early supply crunch. But GPU rental prices historically follow a steep depreciation curve. The RTX 4090, for comparison, started near $1.00 per hour in late 2022 and settled into the $0.30 to $0.50 range within 18 months.
Key Factors Driving the Odds
The 56% price reflects two competing forces. First, AI inference demand keeps climbing. Enterprises and independent developers rent consumer-grade GPUs for fine-tuning, image generation, and small-scale inference workloads. The RTX 5090's 32GB of VRAM makes it attractive for running 70B-parameter quantized models locally, and that demand floor supports prices.
Second, supply keeps expanding. NVIDIA shipped massive Blackwell volumes through 2025 and 2026, and the RTX 6090 (or whatever the next generation is called) will likely launch before March 2027, pushing 5090s down the rental food chain. Historical patterns show flagship cards lose 40-60% of their rental value within two years of launch.
What Could Change These Odds
Watch for the RTX 6090 announcement, expected late 2026. If NVIDIA delays next-gen consumer cards, 5090 prices stay elevated and this market climbs toward 70% or higher. Conversely, an aggressive launch with strong supply could push prices below $0.40, making the Yes side look generous at 56%.
The Ornn index methodology matters too. This market uses a specific USD-denominated index with two-decimal rounding. Thin liquidity on the underlying index could produce volatile monthly averages that don't match what spot markets show. A single large rental contract priced above $0.50 for a sustained period could skew the arithmetic mean even if typical spot rates sit lower. If you believe the index captures institutional rental agreements rather than spot pricing, the Yes side deserves more weight than the 56% suggests.
AI-generated analysis based on market data. Not financial advice.
Overview
This prediction market concerns the average hourly cost of compute using NVIDIA's RTX 5090 graphics card in March 2027. The RTX 5090, announced in January 2025, is part of NVIDIA's Blackwell architecture, featuring 32GB of GDDR7 memory and a boost clock of 2.41 GHz. The market resolves based on the arithmetic mean of hourly values reported by Ornn, a data provider that tracks GPU rental prices on cloud platforms. The 'USD' index is used, with values rounded to two decimal places. This metric reflects the dynamic pricing of GPU compute in the cloud, influenced by supply, demand, and technological shifts. Compute pricing for GPUs has become a critical economic indicator as AI workloads expand across industries. The RTX 5090, while consumer-focused, is also used in small-scale AI inference and rendering tasks, making its cloud rental price a proxy for mid-range GPU demand. Ornn aggregates prices from major cloud providers and marketplaces, offering a transparent index that traders can use to speculate on future trends. The March 2027 date is far enough out that significant market changes could occur, such as the release of next-generation GPUs or shifts in energy costs. Recent developments, such as the AI boom and the increasing adoption of GPU cloud services, have driven volatility in compute prices. The RTX 5090's launch saw high demand, leading to shortages and elevated prices on secondary markets. However, by 2027, the market may stabilize as manufacturing scales and alternatives emerge. Traders in this market are essentially betting on the balance between sustained AI demand and potential oversupply. Interest in this market stems from its practical implications for businesses and researchers who rely on GPU compute. A high resolution value indicates expensive compute, which could signal supply constraints or robust demand, while a low value suggests oversupply or technological obsolescence. This market also attracts those interested in forecasting hardware lifecycle and cloud pricing dynamics.
Historical Context
The pricing of GPU compute has evolved significantly over the past decade. In the mid-2010s, cloud GPU rental prices were relatively stable, with offerings like AWS's G2 instances starting at around $0.65 per hour. The surge in cryptocurrency mining in 2017 drove up GPU prices, with consumer cards like the GTX 1080 Ti selling at premiums of 50% or more. This period highlighted the sensitivity of GPU prices to external demand shocks. The AI boom, particularly after the release of ChatGPT in November 2022, transformed GPU compute into a strategic resource. NVIDIA's A100 and H100 data center GPUs became scarce, with rental prices soaring to $4-8 per hour for H100s. In contrast, consumer GPUs like the RTX 30 series saw price drops after the crypto bust, but the RTX 40 series maintained higher prices due to AI demand. The RTX 5090, launched in early 2025, continued this trend, with initial cloud rental prices around $0.50-1.00 per hour. Historically, GPU prices tend to decline as newer models are released. For example, the RTX 3080's cloud price fell by about 30% within a year of its successor's launch. By March 2027, the RTX 5090 will be two years old, and NVIDIA may have released a 60 series, which could depress prices. However, if AI demand continues to outpace supply, prices might remain elevated. This historical pattern of decline and disruption is central to the market's uncertainty.
Why It Matters
The price of RTX 5090 compute per hour is a microcosm of the broader GPU economy, which affects countless businesses and researchers. For startups developing AI models, cloud compute costs are a significant operational expense. A high average price in March 2027 could signal that AI innovation is becoming more expensive, potentially slowing progress and consolidating power among well-funded companies. Conversely, low prices would democratize access to AI development, enabling smaller players to compete. This metric also reflects the state of the global semiconductor supply chain. If prices remain high, it suggests sustained demand and potential supply constraints, which could have geopolitical implications as countries vie for technological supremacy. Additionally, the price influences the secondary market for GPUs, impacting consumers and gamers. For investors, the prediction market offers a way to hedge against or speculate on these trends, making it a bellwether for the tech sector's health.
Current Status
As of early 2025, the RTX 5090 has just launched, with initial stock shortages and high demand from AI enthusiasts and researchers. Cloud providers are beginning to offer RTX 5090 instances, with prices starting around $0.75 per hour on some platforms. The market for this prediction is likely to be influenced by upcoming announcements from NVIDIA about the RTX 60 series, expected in late 2026 or early 2027, which could reduce demand for the 5090. Additionally, the broader economic environment, including energy prices and AI investment trends, will play a role. Traders are currently assessing these factors, with early indications that the market expects a moderate decline in prices by March 2027.
Frequently Asked Questions
What is the RTX 5090's compute price per hour?
The compute price per hour for the RTX 5090 varies by cloud provider, but as of early 2025, it ranges from $0.50 to $1.00 per hour. This price is influenced by factors like hardware cost, electricity, and demand.
How is the average compute price calculated?
The average is calculated as the arithmetic mean of hourly values reported by Ornn, a data provider that aggregates prices from various cloud platforms. Revisions after expiration are not considered.
Will the RTX 5090's price drop by 2027?
Historically, GPU prices decline as newer models are released. With the RTX 60 series likely coming in late 2026, the RTX 5090's price could fall by 20-40% from its launch price, but sustained AI demand could keep it stable.
What affects the cloud rental price of GPUs?
Key factors include hardware supply, electricity costs, competition among cloud providers, and demand from AI workloads. Shortages or surges in demand can cause price spikes, while oversupply leads to drops.
What is Ornn?
Ornn is a data provider that tracks GPU compute prices across cloud platforms, offering an index that reflects average hourly costs. This index is used as the resolution source for this prediction market.
Educational content is AI-generated and sourced from Wikipedia. It should not be considered financial advice.
