Week 25: Price of NVIDIA's B200 compute at end of week?
Short Answer
Both the model and the market overwhelmingly agree that the price of NVIDIA's B200 compute at the end of Week 25 will be Above $3.00, with only minor residual uncertainty.
1. Executive Verdict
- Blackwell B200 GPU supply appears severely limited by packaging capacity.
- Major AI model initiatives are expected to significantly boost B200 demand.
- Hyperscalers appear to advertise significantly higher hourly rates for B200 instances.
- Cloud provider rates and reports indicate B200 compute rental prices likely to surge.
- NVIDIA production schedules and hyperscaler demand fluctuations may influence B200 pricing.
Who Wins and Why
| Outcome | Market | Model | Why |
|---|---|---|---|
| Above $7.00 | 1.0% | 13.0% | Research does not highlight strong supporting evidence. |
| Above $6.50 | 1.0% | 13.0% | Research does not highlight strong supporting evidence. |
| Above $6.00 | 1.0% | 13.0% | Research does not highlight strong supporting evidence. |
| Above $3.00 | 97.0% | 97.0% | Research does not highlight strong supporting evidence. |
| Above $3.50 | 97.0% | 95.0% | Research does not highlight strong supporting evidence. |
Current Context
NVIDIA B200 does not have an official list price. As of June 2026, the B200 is not sold as a standalone unit to the public. However, the effective implied market price for a B200 GPU is approximately $40,000, with a reported range between $30,000 and $55,000 [^]. Supply for the Blackwell B200 remains constrained as of mid-2026 due to a large order backlog. Despite these constraints, volume shipping is expanding, which is contributing to stabilizing pricing as cloud providers bring more capacity online [^][^][^].
Cloud rental prices for B200 compute vary significantly. At the end of Week 25 (June 19, 2026), cloud rental pricing varied widely based on the provider and access model. Spot prices were observed as low as approximately $2.14 per hour, while on-demand rates typically ranged from approximately $4.60 per hour to $14.00 per hour, depending on specific instance configurations [^][^][^][^].
2. Market Behavior & Price Dynamics
Historical Price (Probability)
Outcome probability
Date
This prediction market has experienced a dramatic and rapid upward trend. The contract began trading at a very low probability, starting at 1.0% on June 13 and briefly touching 0.0% on June 15. This indicates initial market skepticism or uncertainty about the resolution criteria being met. However, the market saw an explosive move to a 97.0% YES probability by June 18. This sharp spike suggests that new information became available during the resolution week, aligning with reports that the effective market price for a B200 GPU is approximately $40,000, well within the commonly reported range of $30,000 to $55,000. The market's rapid repricing reflects a sudden consensus forming around this reported price stability.
The total volume traded is extremely low at only 20 contracts, which suggests that market conviction may be thin despite the high price. The sample data points show zero volume during the most significant price swings, indicating that the move from near 0% to 97% may have been driven by a small number of trades or even a single market maker adjustment on very little liquidity. Due to the swift, one-directional movement, clear support and resistance levels have not been established. The key price points are simply the floor near 0% and the current ceiling near 97%.
Overall, the chart illustrates a profound shift in market sentiment from extreme doubt to overwhelming confidence. Initially, participants saw little chance of the B200's price meeting the market's condition. By mid-week, sentiment completely inverted, and the market now overwhelmingly expects a YES resolution. This change reflects traders processing real-world pricing data for the B200 as it became more concrete during the resolution period.
3. Significant Price Movements
Notable price changes detected in the chart, along with research into what caused each movement.
📈 June 17, 2026: 90.0pp spike
Price increased from 0.0% to 90.0%
Outcome: Above $4.00
What happened: The primary driver of the 90 percentage point spike was traditional market news indicating significant increases in NVIDIA B200 compute rental prices. Reports circulated that NVIDIA B200 GPU prices had surged 114%, with some providers issuing notices of impending lease rate increases that could potentially double prices [^][^]. These reports reinforced the expectation that B200 compute prices would remain robustly above $4.00 per hour, aligning with mainstream on-demand instances already ranging from $4.63/hr to $5.98/hr in June 2026 [^][^][^]. Social media activity was not a primary driver, as the available research did not identify specific posts from key figures or viral narratives directly causing this movement.
📉 June 16, 2026: 83.0pp drop
Price decreased from 83.0% to 0.0%
Outcome: Above $4.00
What happened: The primary driver for the prediction market's 83.0 percentage point drop was likely traditional news reports indicating a significant decline in NVIDIA's share price, such as an "Nvidia Earnings Lead to Share Slump" [^] or "Nvidia’s Shock Selloff" [^]. This negative financial outlook would have prompted a rapid downward revision of B200 compute price expectations, especially since spot cloud rental prices were already observed as low as $2.139/hr as of June 16-17, 2026 [^][^]. This movement corrected the market from earlier projections that showed little belief in prices dropping below $4.00/hr [^]. Social media was not a primary driver, as no relevant activity was identified in the provided research.
📈 June 15, 2026: 84.0pp spike
Price increased from 0.0% to 84.0%
Outcome: Above $4.00
What happened: The provided research does not indicate a specific social media post, news announcement, or market event on June 15, 2026, that directly caused the 84.0 percentage point spike in the prediction market for NVIDIA B200 compute prices. While B200 GPU rental prices have generally exhibited volatility and past surges in late May and early June 2026, often stabilizing between $4 and $6 per hour, the available information does not pinpoint the primary driver for this particular movement on the given date [^][^][^][^][^][^][^]. No specific social media activity from key figures or viral narratives were identified. Therefore, social media was irrelevant as a primary driver based on the provided sources.
4. Market Data
Contract Snapshot
A "Yes" resolution occurs if NVIDIA B200 compute per hour is above $4.5 on June 19, 2026, at 4 PM ET; otherwise, it resolves "No". The market closes at this time, with the outcome verified from Ornn (dashboard.ornnai.com). If Ornn publishes no data for the specified date and time, the most recently available published data will be used, and revisions made after expiration will not be accounted for.
Available Contracts
Market options and current pricing
| Outcome bucket | Yes (price) | No (price) | Last trade probability |
|---|---|---|---|
| Above $3.00 | $0.99 | $0.02 | 97% |
| Above $3.50 | $0.99 | $0.02 | 97% |
| Above $6.00 | $0.01 | $1.00 | 1% |
| Above $6.50 | $0.01 | $1.00 | 1% |
| Above $7.00 | $0.01 | $1.00 | 1% |
| Above $4.00 | $0.94 | $0.07 | 0% |
| Above $4.50 | $0.31 | $0.70 | 0% |
| Above $5.00 | $0.03 | $0.98 | 0% |
| Above $5.50 | $0.02 | $0.99 | 0% |
Market Discussion
NVIDIA B200 cloud hourly rental rates currently vary widely, from approximately $2.14/hr to over $14.00/hr, with a market-observed average between $4.50–$6.00/hr as of June 2026 [^][^][^][^][^]. Prediction markets for the end of June 2026 B200 compute price have largely centered their consensus around the $5.00–$6.00/hr range, reflecting a balance of ongoing supply constraints and high demand [^][^][^]. Industry commentary also notes that effective compute costs (price per million tokens) have been significantly reduced, sometimes by up to 5x, due to software optimizations rather than solely changes in hardware rental prices [^][^].
5. How might the market penetration and performance of AMD's Instinct MI300X series in Q2 2026 influence cloud providers' pricing for NVIDIA's B200?
| NVIDIA Market Share | 80–85% revenue share (Q2 2026) [^][^][^] |
|---|---|
| AMD Instinct Market Share | 5–7% market share (Q2 2026) [^][^][^] |
| NVIDIA B200 Average On-Demand Cloud Pricing | $4.71–$5.89/hr (mid-June 2026) [^][^][^][^] |
NVIDIA maintains market dominance, but AMD's MI300X series offers a viable alternative. As of Q2 2026, NVIDIA continues to hold a dominant position in the AI accelerator market, accounting for approximately 80–85% of revenue share. Meanwhile, AMD's Instinct series, including the MI300X and MI325X, has secured roughly 5–7% of the market share [^][^][^]. The AMD Instinct MI300X has emerged as a credible, cost-effective solution for inference and fine-tuning workloads, particularly where its 192GB of HBM3 memory provides an efficiency advantage over NVIDIA's 80GB H100 configurations [^][^].
AMD's market presence is expected to pressure NVIDIA B200 cloud pricing. The growing market penetration and demonstrated performance of AMD's Instinct MI300X series are anticipated to influence cloud providers' pricing strategies for NVIDIA's B200 [^][^]. This competitive environment will compel providers to re-evaluate GPU cost and capacity. Such pressure encourages a focus on metrics like cost-per-token, especially for inference-heavy workloads, rather than solely relying on NVIDIA's historical pricing power [^][^].
NVIDIA B200 cloud pricing in Q2 2026 remains highly variable. As of mid-June 2026, NVIDIA B200 cloud pricing is characterized by volatility and supply constraints [^][^][^][^]. On-demand rates for the B200 average approximately $4.71–$5.89 per hour. However, prices show significant variation, ranging from about $2.14 per hour for spot or reserved instances up to $14–$16+ per hour for on-demand access at major cloud providers [^][^][^][^].
6. What do Q2 2026 production reports from NVIDIA's key manufacturing partner, TSMC, indicate about the supply volume of Blackwell B200 GPUs?
| B200 Unit Backlog | Approximately 3.6 million [^][^] |
|---|---|
| TSMC CoWoS-L Capacity Expansion | Approximately 40,000 wafers/month by year-end [^][^][^] |
| Enterprise Lead Times | 8–16 weeks [^][^] |
Blackwell B200 GPU supply remains severely limited by packaging capacity. NVIDIA's Blackwell (B200/GB200) GPU production for Q2 2026 is critically constrained by TSMC's CoWoS-L advanced packaging capacity [^][^][^]. This limitation results in significantly constrained supply volumes, with demand for these advanced GPUs considerably outpacing the available output [^][^][^].
Despite expansion, a large backlog persists for B200 GPUs. Despite TSMC's ongoing efforts to expand its CoWoS-L advanced packaging capacity, targeting approximately 40,000 wafers per month by year-end, a substantial backlog for B200 units exists [^][^][^]. The current B200 unit backlog is approximately 3.6 million, with hardware reportedly sold out through mid-2026 [^][^]. Consequently, enterprise lead times for priority buyers are estimated to be between 8 and 16 weeks [^][^]. Prediction markets for late June 2026 further reflect these persistent supply constraints, projecting B200 rental prices to stabilize around $5–$6 per hour [^].
7. How do advertised hourly rates for B200 instances on major hyperscalers like AWS and Azure compare to those on specialized GPU clouds like CoreWeave in mid-2026?
| AWS B200 Hourly Rate | $14.13–$14.24/GPU/hr (mid-June 2026) [^][^] |
|---|---|
| RunPod B200 Hourly Rate | $4.99–$5.89/hr (mid-June 2026) [^][^] |
| Prediction Market B200 Rate | $5.00–$6.00 per hour (June 2026) [^][^][^] |
Major hyperscalers advertise significantly higher hourly rates for B200 instances. In mid-June 2026, a notable price disparity exists for NVIDIA B200 instances, with major hyperscalers charging significantly more than specialized GPU cloud providers. AWS on-demand NVIDIA B200 instance pricing is reported between $14.13–$14.24/GPU/hr, and Azure's rates range from approximately $13.52–$27.04/GPU/hr [^][^]. These rates from hyperscalers are estimated to be 2x to 5x higher compared to those offered by specialized providers [^][^].
Specialized GPU cloud providers offer substantially lower B200 instance pricing. RunPod lists rates between $4.99–$5.89/hr, and Lambda's rates are around $5.50/hr [^][^]. CoreWeave advertises NVIDIA B200 instance pricing for inference at $8.60/GPU/hour, with total hourly costs for HGX B200 clusters, such as $68.80/hour for 8x B200 configurations, reaching higher figures [^][^][^].
Prediction markets consistently indicate lower B200 compute pricing for June 2026. These markets for B200 compute pricing in June 2026 consistently identify the $5.00–$6.00 per hour range as the modal outcome for end-of-June rental rates [^][^][^]. This reflects a market consensus that prices are likely to remain bounded well below the list prices currently advertised by major hyperscalers [^][^][^].
8. Which public pricing aggregators or cloud provider dashboards offer the most reliable data for tracking B200 spot and on-demand compute prices for Q2 2026?
| Lowest B200 spot price (June 2026) | $2.14/hr [^] |
|---|---|
| Typical B200 on-demand price range | ~$5.00/hr to over $14.00/hr [^][^][^][^][^] |
| Providers compared by Mercatus | Over 30 [^] |
Dedicated GPU-cloud aggregators reliably track B200 spot and on-demand prices. For reliable, real-time tracking of NVIDIA B200 spot and on-demand pricing in Q2 2026, dedicated GPU-cloud aggregators are the most effective. Platforms such as GPU Tracker, ComputePrices, GetDeploying, and Mercatus normalize pricing across numerous providers and clearly differentiate between marketplace (spot/community) and secure (on-demand/dedicated) instances [^].
B200 pricing in Q2 2026 shows significant disparity across providers. As of June 2026, NVIDIA B200 pricing exhibits substantial differences. Spot instances are commonly available around $2.14/hr [^][^]. During Week 25 (June 15–19, 2026), the lowest observed spot price for a B200 instance consistently remained at approximately $2.14/hr, a baseline maintained throughout the first half of June 2026 [^]. On-demand pricing typically ranges from approximately $5.00/hr to over $14.00/hr, with variations dependent on the provider's service model and commitment level [^][^][^][^][^].
These aggregators facilitate comparison across numerous B200 compute providers. Dedicated GPU-cloud aggregators are instrumental in navigating these price differences. For example, Mercatus compares over 30 providers daily [^], GetDeploying compares more than 26 providers [^], and ComputePrices.com offers B200 GPU information from over 18 providers [^].
9. What major AI model training initiatives from firms like OpenAI, Google, or Anthropic are anticipated in Q2 2026 that could significantly increase demand for B200 compute?
| B200 Lead Times | 8–16 weeks (standard enterprise buyers) [^] |
|---|---|
| B200 Cloud On-Demand Pricing | $5.50 to $14.24/hr [^][^][^] |
| AI XPV Platform Investment | $35 billion [^] |
Major AI model initiatives in Q2 2026 will significantly boost B200 demand. Anticipated releases include OpenAI’s GPT-5.5 on April 23, the ongoing restricted preview of Anthropic’s Claude Mythos, and the imminent release of Google’s Gemini 3.2 [^]. This increased demand is further amplified by massive infrastructure expansions, notably Anthropic's 300MW compute arrangement with SpaceX announced on May 6, and a $35 billion AI XPV Platform initiative supporting both Anthropic and OpenAI expansion [^][^]. The power demand for AI compute is growing exponentially, contributing to an acknowledged compute shortage [^][^].
B200 hardware faces severe supply constraints and high costs due to bottlenecks. Lead times for standard enterprise buyers of B200 hardware currently range from 8 to 16 weeks [^]. Cloud on-demand pricing for B200 compute varies from $5.50 to $14.24 per hour, with prediction markets anticipating hourly rates above $5.07 by June 30, 2026, as of June 2026 [^][^][^]. These elevated cloud prices are a direct result of severe supply chain bottlenecks affecting advanced packaging and High Bandwidth Memory (HBM), leading some providers to mandate large-scale, long-term commitments for access to Blackwell hardware [^][^].
10. What Could Change the Odds
Key Catalysts
Key Dates & Catalysts
- Strike Date: June 19, 2026
- Expiration: June 26, 2026
- Closes: June 19, 2026
11. Decision-Flipping Events
- Trigger: Key catalysts that could influence B200 pricing volatility include Nvidia's production ramp schedules and fluctuations in hyperscaler demand [^] [^] [^] [^] .
- Trigger: Other significant factors that could change market probability for B200 pricing are U.S.
- Trigger: Commerce Department export control rulings and the availability of next-gen Blackwell Ultra (B300) hardware [^] [^] [^] [^] .
13. Historical Resolutions
No historical resolution data available for this series.
