1. The GPU compute complex dominated price action across Kalshi, with H100 contracts the epicenter. The Aug 28 strike above $3.25 per hour surged 800% from 2% to 18%, the $3 strike jumped 155% from 20% to 51%, the $2.75 strike doubled from 39% to 78%, and the simple "H100 SXM price up in next week" question rose from 34% to 66%. Read together, the whole ladder implies traders expect the H100 spot price on Aug 28 to land near $2.75 to $3.25, roughly a 10% to 30% step up from the $2.5 to $2.75 range that current strikes imply, with the $2.26 and $2.31 Aug 31 strikes already pinned at 97% and 92%.
2. The move sits squarely on top of Nvidia infrastructure headlines. TechCrunch reported that Nvidia partnered with data center developer Cloverleaf and argued that the "harness, not the model," is now the real hero of its stack. Starcloud raised $250 million for orbital data centers, and Micro1 crossed a $500 million gross run rate on AI training data. None of those stories directly named H100 spot pricing, which is why the report's automatic news-overlap check flagged a gap, but the direction is consistent: supply-constrained GPU capacity, more data-center buildout commitments, and a market that prices a near-term uptick in spot rates.
3. H100 was the only GPU family with meaningful repricing. H200 moved modestly (the $5.5 strike tripled off a 1% base, but absolute movement was small), B200 strikes above $9.50 and $10 dropped 50% to 1%, and A100 contracts above $1.50 on Sep 04 fell from 4% to 1%. The spread between H100 repricing upward while B200 repricing downward is a clean read on the current GPU hierarchy: H100 capacity is tight and commands a premium, B200 supply is arriving faster than buyers expected, and A100 is fading out.
4. OpenAI product timing markets also moved in one direction. "Will OpenAI release GPT-6 before Sep 1, 2026?" tripled from 2% to 8%, and "Will OpenAI release Astra before Sep 18, 2026?" rose from 48% to 70%. Polymarket's "GPT-6 by Dec 31, 2026" held at 70% and "GPT-6 by Jan 1, 2027" on Kalshi sat at 72%, so the Sep 1 market tripling off a low base looks like traders front-running a rumor cycle rather than a confirmed date. No headline in today's feed confirms a GPT-6 launch, which is the main divergence to watch: the Sep 1 market is pricing in something the news flow does not yet support.
5. Model-identity markets on Kalshi pointed the same way. "Best AI on Dec 31, 2026" has Claude at 67%, ChatGPT down 24% to 10%, Grok down 25% to 11%, and Ernie effectively zero. "Top AI model this month" has claude-opus-5-high at 60% and claude-opus-5-max at 30%, with the thinking variant at 2%. The market is consolidating around Claude as the default 2026 frontier pick even as TechCrunch ran a piece calling Opus 4.6 "a smut-machine" and a separate story that Z.ai is 87% likely to be confirmed behind the stealth model Ox Alpha. The Z.ai result is the clearest price-vs-news divergence in the feed: the news names Z.ai as the stealth model, but the market still prices "no company confirmed" at 11% and xAI at 3%, so traders are waiting for the source agency to actually say the words rather than taking the headline at face value.
| Source | Market | Change | Time |
| Kalshi |
Will the NVIDIA H100 compute per hour price be above $3.25 at 4 PM ET on Aug 28? |
↑ 800% (2% → 18%) |
(vs last snapshot, 1.4 days ago) |
| Kalshi |
Will the NVIDIA H200 compute per hour price be above $5.5 at 4 PM ET on Aug 28? |
↑ 300% (1% → 4%) |
(vs ~24h ago) |
| Kalshi |
Will OpenAI release GPT-6 before Sep 1, 2026? |
↑ 300% (2% → 8%) |
(vs ~24h ago) |
| Kalshi |
Will the NVIDIA H100 compute per hour price be above $3 at 4 PM ET on Aug 28? |
↑ 155% (20% → 51%) |
(vs ~24h ago) |
| Kalshi |
What will be the top AI model this month? |
↑ 100% (1% → 2%) |
(vs last snapshot, 17.0h ago) |
| Kalshi |
Best AI on Dec 31, 2026? |
↑ 100% (0% → 0%) |
(vs ~24h ago) |
| Kalshi |
Will the NVIDIA H100 compute per hour price be above $2.75 at 4 PM ET on Aug 28? |
↑ 100% (39% → 78%) |
(vs ~24h ago) |
| Kalshi |
H100 SXM price up in next week? |
↑ 94% (34% → 66%) |
(vs ~24h ago) |
| Kalshi |
Will the NVIDIA A100 compute per hour price be above $1.5 at 4 PM ET on Sep 04? |
↓ 75% (4% → 1%) |
(vs ~24h ago) |
| Kalshi |
Will Claude App Downloads for August 2026 be above 1600 ? |
↓ 71% (14% → 4%) |
(vs last snapshot, 17.0h ago) |
| # | Question | Yes | No | Change | Volume |
| 1 | What will be the top AI model this month? (claude-opus-4-6-thinking) | 2% | 98% |
↑ 100% (was 1%) (vs last snapshot, 17.0h ago) | $38,213 |
| 2 | Will the NVIDIA RTX 5090 compute per hour price be above $0.35 at 4 PM (Above $0.35) | 99% | 1% |
| $7,996 |
| 3 | Will it be reported by a Source Agency that Z.ai is confirmed as the d (Z.ai) | 87% | 13% |
| $7,981 |
| 4 | Best AI on Dec 31, 2026? (Ernie) | 0% | 100% |
↑ 100% (was 0%) (vs ~24h ago) | $7,533 |
| 5 | Best AI on Dec 31, 2026? (Claude) | 67% | 33% |
| $5,250 |
| 6 | Best AI on Dec 31, 2026? (Kimi) | 1% | 99% |
| $5,000 |
| 7 | Best AI on Dec 31, 2026? (ChatGPT) | 10% | 90% |
↓ 24% (was 13%) (vs ~24h ago) | $4,900 |
| 8 | Will the NVIDIA RTX 5090 compute per hour price be above $0.25 at 4 PM (Above $0.25) | 99% | 1% |
| $4,300 |
| 9 | Will the NVIDIA RTX 5090 compute per hour price be above $0.2 at 4 PM (Above $0.20) | 99% | 1% |
| $4,300 |
| 10 | Will the NVIDIA RTX 5090 compute per hour price be above $0.15 at 4 PM (Above $0.15) | 99% | 1% |
| $4,296 |
| 11 | Best AI on Dec 31, 2026? (Grok) | 11% | 89% |
↓ 25% (was 15%) (vs ~24h ago) | $4,100 |
| 12 | Will OpenAI release Astra before Sep 18, 2026? (Before Sep 18, 2026) | 70% | 30% |
↑ 46% (was 48%) (vs ~24h ago) | $2,522 |
| 13 | Will Anthropic release next Mythos-Class model before Oct 1, 2026? (Before Oct 1, 2026) | 60% | 40% |
↑ 15% (was 52%) (vs last snapshot, 17.0h ago) | $2,507 |
| 14 | Will it be reported by a Source Agency that xAI is confirmed as the de (xAI) | 3% | 97% |
↓ 25% (was 4%) (vs ~24h ago) | $2,427 |
| 15 | What will be the top AI model this month? (claude-opus-5-high) | 60% | 40% |
| $2,034 |
| 16 | Will it be reported by a Source Agency that no company is confirmed as (No listed company) | 11% | 89% |
↑ 57% (was 7%) (vs ~24h ago) | $1,868 |
| 17 | Will the NVIDIA B200 compute per hour price be above $10 at 4 PM ET on (Above $10.00) | 1% | 99% |
↓ 50% (was 2%) (vs last snapshot, 1.4 days ago) | $1,668 |
| 18 | Will Open AI or Anthropic IPO first? (Anthropic) | 89% | 11% |
↓ 4% (was 93%) (vs ~24h ago) | $1,524 |
| 19 | Will the NVIDIA H200 compute per hour price be above $4 at 4 PM ET on (Above $4.00) | 91% | 9% |
↑ 6% (was 86%) (vs ~24h ago) | $1,438 |
| 20 | Will Anthropic release Mythos before Jan 1, 2027? (Before Jan 1, 2027) | 16% | 84% |
↓ 11% (was 18%) (vs ~24h ago) | $1,360 |
| 21 | Will the NVIDIA H100 compute per hour price be above $2.5 at 4 PM ET o (Above $2.50) | 94% | 6% |
↑ 18% (was 80%) (vs ~24h ago) | $1,283 |
| 22 | Will OpenAI have a top-ranked AI model before 2027? (OpenAI) | 26% | 74% |
↓ 13% (was 30%) (vs last snapshot, 17.0h ago) | $1,172 |
| 23 | Will the NVIDIA H200 compute per hour price be above $7.5 at 4 PM ET o (Above $7.50) | 1% | 99% |
| $1,150 |
| 24 | Will the NVIDIA H100 compute per hour price be above $2.75 at 4 PM ET (Above $2.75) | 78% | 22% |
↑ 100% (was 39%) (vs ~24h ago) | $1,138 |
| 25 | Will the NVIDIA H200 compute per hour price be above $3.5 at 4 PM ET o (Above $3.50) | 91% | 9% |
↑ 7% (was 85%) (vs ~24h ago) | $1,127 |
| 26 | Will the NVIDIA RTX 5090 compute per hour price be above $0.3 at 4 PM (Above $0.30) | 99% | 1% |
| $1,074 |
| 27 | Will the NVIDIA A100 compute per hour price be above $2 at 4 PM ET on (Above $2.00) | 1% | 99% |
| $1,050 |
| 28 | What will be the top AI model this month? (claude-opus-5-max) | 30% | 70% |
↓ 14% (was 35%) (vs ~24h ago) | $1,017 |
| 29 | Will the NVIDIA B200 compute per hour price be above $9.5 at 4 PM ET o (Above $9.50) | 1% | 99% |
↓ 50% (was 2%) (vs last snapshot, 1.4 days ago) | $1,000 |
| 30 | Will OpenAI release GPT-6 before Sep 1, 2026? (Before Sep 1, 2026) | 8% | 92% |
↑ 300% (was 2%) (vs ~24h ago) | $976 |